<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/">
  <channel>
    <title>AquaMesh — field notes</title>
    <link>https://aquamesh.ai/blog/</link>
    <description>What we find in industrial water plants, how the measurement works, and what it is worth.</description>
    <language>en</language>
    <lastBuildDate>Fri, 13 Nov 2026 01:12:18 GMT</lastBuildDate>
    <atom:link href="https://aquamesh.ai/feed.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Industrial pretreatment: the self-monitoring blind spot</title>
      <link>https://aquamesh.ai/blog/industrial-pretreatment-self-monitoring-gaps.html</link>
      <guid isPermaLink="true">https://aquamesh.ai/blog/industrial-pretreatment-self-monitoring-gaps.html</guid>
      <pubDate>Thu, 12 Nov 2026 09:00:00 GMT</pubDate>
      <description>Monthly or quarterly grab samples miss cleaning cycles and batch upsets. Under 40 CFR Part 403, industrial dischargers bear the compliance cost of that gap.</description>
      <content:encoded><![CDATA[<p>Industrial facilities discharging process water to a publicly owned treatment works (POTW) operate under a monitoring program designed around the administrative capacity of the 1970s. That program, the National Pretreatment Program, still forms the compliance backbone for thousands of manufacturers, food processors and industrial facilities in the United States. Its self-monitoring requirements are legally adequate. They are not designed to capture what modern batch and cyclic operations actually produce.</p>
<h2 id="what-the-regulations-require">What the regulations require</h2>
<p>The National Pretreatment Program is codified at <a href="https://www.ecfr.gov/current/title-40/chapter-I/subchapter-N/part-403" rel="noopener">40 CFR Part 403</a>. Its core requirement is that industrial users (IUs) discharging to a POTW must meet applicable pretreatment standards and report compliance to the control authority, which is usually the POTW itself.</p>
<p>Categorical pretreatment standards set technology-based discharge limits for specific industry categories. The reporting requirements at <a href="https://www.ecfr.gov/current/title-40/chapter-I/subchapter-N/part-403/section-403.12" rel="noopener">40 CFR Section 403.12</a> require significant industrial users (SIUs) to submit semi-annual self-monitoring reports. These reports cover the pollutants identified in the facility's permit, based on samples taken during a "representative period of daily operations."</p>
<p>That phrase carries more weight than it appears.</p>
<h2 id="the-sampling-gap">The sampling gap</h2>
<p>A food and beverage plant has at least two fundamentally different discharge profiles. During production, the facility generates process water with moderate organic loading. At the end of a production run, clean-in-place (CIP) cycles flush equipment with hot water and cleaning agents. The resulting discharge is concentrated: studies on food processing wastewater characterization, including EPA guidance on <a href="https://www.epa.gov/npdes/national-pretreatment-program" rel="noopener">the National Pretreatment Program</a>, document that CIP and sanitation flows typically carry significantly higher organic loads than steady-state production discharge, compressed into a window of one to three hours.</p>
<p>A quarterly grab sample taken at 10 a.m. during a production run will not capture that event.</p>
<p>The same pattern appears across industrial categories:</p>
<ul class="post-list"><li><b>Batch chemical processes</b>: end-of-batch vessel drains carry concentrated organics and total dissolved solids that differ sharply from steady-state discharge.</li><li><b>Metal finishing operations</b>: rinse-water dumps and plating bath replacements create intermittent peaks in metals concentrations that fall outside scheduled sample windows.</li><li><b>Dairy and food processing</b>: cleaning cycles and product changeovers generate BOD spikes that last hours but occur between collection points. The measurement gap in dairy processing operations, and what it costs before it surfaces, is described in <a href="/blog/dairy-foods-process-losses.html">why dairy plants miss their most costly process losses</a>.</li></ul>
<p>A facility with 250 operating days per year and quarterly sampling has four snapshots of its discharge profile. Those four points may or may not coincide with the conditions that drive actual compliance exposure.</p>
<h2 id="what-compliance-exposure-looks-like">What compliance exposure looks like</h2>
<p>The compliance structure amplifies the sampling gap in two directions.</p>
<p>First, if a grab sample catches a discharge above the local limit, the POTW bills surcharges. Surcharges for excess BOD, TSS or FOG are set by local ordinance and expressed in dollars per pound above the threshold. For a facility generating a BOD spike during a CIP cycle, a single high sample can produce a surcharge bill that reflects an event the facility cannot document or explain.</p>
<p>Second, if the facility does not know its own discharge profile, it cannot demonstrate the sample was atypical. It has no process record tying the high load to a cleaning cycle, no continuous data showing what the daily average actually was, and no basis for requesting a resample or appealing a surcharge.</p>
<p>Facilities with continuous monitoring face the same events, but they have a record. A plant with an online organic-load sensor can show the POTW that the high sample coincided with a CIP cycle, that the event lasted 90 minutes out of 24 hours, and that the daily average load was within permit limits. Some POTWs structure local limits around daily averages rather than instantaneous peaks. Continuous data is what allows that argument to be made.</p>
<p>The sampling gap is not confined to industrial dischargers under the pretreatment program. The same dynamic shapes compliance exposure for NPDES permittees: <a href="/blog/nitrogen-permit-compliance-sampling-gap.html">NPDES nitrogen permit compliance and the DMR sampling gap</a> describes how scheduled sampling intervals at municipal treatment plants create comparable exposure in discharge monitoring reports.</p>
<h2 id="why-periodic-sampling-persists">Why periodic sampling persists</h2>
<p>The case for grab sampling is administrative simplicity. A POTW with 100 or more industrial users cannot process continuous data streams from each one. The control authority receives reports, reviews them against permit limits, and takes enforcement action when limits are exceeded. This model scales well for the regulator.</p>
<p>The problem is that it scales on the assumption that grab samples represent the discharge they are drawn from. For a facility with a stable, continuous process, that assumption holds reasonably well. For a facility with batch operations, cleaning cycles or variable production schedules, it does not hold for the high-load events that determine actual compliance exposure.</p>
<p>The EPA pretreatment program regulations give control authorities broad authority to require more frequent monitoring when the existing program does not adequately characterize the discharge. The <a href="https://www.ecfr.gov/current/title-40/chapter-I/subchapter-N/part-403" rel="noopener">40 CFR Part 403</a> framework allows the POTW to tighten reporting requirements through the facility's control mechanism. Facilities already under augmented monitoring requirements face this directly. Most facilities are not, until an exceedance triggers review.</p>
<h2 id="a-practical-starting-point">A practical starting point</h2>
<p>Understanding actual discharge variability is the first step. A facility that has installed flow measurement and online organic-load measurement at its final discharge point can generate a profile of daily load variation over a quarter. That data answers two questions: how often do high-load events occur, and do they correlate with identifiable process events such as cleaning schedules, batch cycles or production changeovers?</p>
<p>If the high-load events are identifiable, they can often be managed operationally: equalization tanks, staggered CIP cycles or modified flush procedures. If they are infrequent, the facility has documentation that the quarterly grab sample is statistically likely to be representative. If they are common, the facility knows before the POTW does.</p>
<p>The same historian-based methodology that reveals these patterns in continuous sensor data is described in <a href="/blog/what-25-days-of-historian-data-shows.html">what 25 days of historian data actually shows</a>. The relationship between DO measurement precision and the compliance exposure that surfaces in permit reports is examined in <a href="/blog/intermittent-aeration-nitrogen-removal-do-control.html">intermittent aeration: why DO sensor precision matters</a>. For facilities evaluating continuous online monitoring as a complement to periodic grab sampling, what sensors can and cannot do for official compliance purposes is examined in <a href="/blog/online-sensors-vs-lab-sampling-wastewater-discharge.html">online sensors vs. lab sampling for discharge compliance</a>.</p>
<p>Whether your facility's current self-monitoring program is capturing the events that drive your compliance exposure is worth checking before the POTW does it for you. <a href="https://aquamesh.ai">See how AquaMesh maps discharge variability from operational historian data</a>.</p>
<p><em>Author: Alex Towfigh, AquaMesh</em></p>]]></content:encoded>
      <category>pretreatment</category>
      <category>industrial discharge</category>
      <category>compliance</category>
      <category>self-monitoring</category>
      <category>POTW</category>
      <category>40 CFR Part 403</category>
      <category>BOD</category>
      <category>TSS</category>
    </item>
    <item>
      <title>Ammonia sensor drift in wastewater aeration control</title>
      <link>https://aquamesh.ai/blog/ammonia-sensor-drift-wastewater-aeration-control.html</link>
      <guid isPermaLink="true">https://aquamesh.ai/blog/ammonia-sensor-drift-wastewater-aeration-control.html</guid>
      <pubDate>Sat, 07 Nov 2026 09:00:00 GMT</pubDate>
      <description>How ammonium sensor drift degrades ABAC aeration control, wastes blower energy and creates nitrogen permit compliance gaps.</description>
      <content:encoded><![CDATA[<hr />
<p>title: Ammonia sensor drift in wastewater aeration control slug: ammonia-sensor-drift-wastewater-aeration-control date: 2026-11-07 section: Field notes description: How ammonium sensor drift degrades ABAC aeration control, wastes blower energy and creates nitrogen permit compliance gaps. tags: wastewater, sensors, aeration, nitrogen, activated sludge</p>
<hr />
<p>A 1 mg/L positive drift in an online ammonium sensor running ammonia-based aeration control (ABAC) makes the blower work harder than the process needs, every hour the bias persists. A negative drift at a permit-sensitive plant does the opposite: the process under-aerates, nitrification stalls, and the next grab sample surprises the operator. Both are common. Neither shows up in the normal daily walk-through.</p>
<p>This post covers how ammonium sensors drift in activated sludge service, what that drift does to an ABAC control loop, and what operators can check to catch it before it costs money or compliance points.</p>
<h2 id="how-abac-uses-the-ammonium-signal">How ABAC uses the ammonium signal</h2>
<p>In a fixed-setpoint system, the operator sets a dissolved oxygen (DO) target and blower airflow modulates to hold it. In ABAC, the DO setpoint is a function of the measured ammonium concentration in the aeration basin. When ammonium is high and nitrification demand is high, the system raises the DO setpoint; when ammonium is low, it backs off. The logic trims aeration to the actual nitrogen load rather than the worst-case load.</p>
<p>The energy benefit is real. The U.S. EPA Nutrient Control Design Manual (EPA/600/R-10/100, Chapter 5) documents that properly tuned ABAC systems can reduce aeration energy 10 to 30 percent relative to fixed-DO operation by eliminating the over-aeration margin that fixed setpoints require when loading is low. That saving disappears when the ammonium sensor reading the setpoint signal is wrong.</p>
<h2 id="why-ammonium-sensors-drift">Why ammonium sensors drift</h2>
<p>Most online ammonium analyzers used in municipal and industrial wastewater use ion-selective electrodes (ISEs), the same measurement principle as benchtop ammonium probes. An ISE generates a voltage proportional to the logarithm of ammonium activity. Two mechanisms cause that voltage to drift in field conditions.</p>
<p><b>Reference electrode fouling.</b> The ISE measurement is a potential difference between the measuring electrode and a reference electrode. The reference electrode contains a fill solution that contacts the sample through a small junction. In a wastewater aeration basin, biological solids, oils, and sulfide compounds can clog or coat that junction over weeks. As the junction fouls, the reference potential drifts, and the ammonium reading drifts with it. <a href="/blog/ph-sensor-drift-wastewater-activated-sludge.html">pH sensors face the same fouling mechanism at the reference junction</a>, and ammonium ISEs show similar calibration drift timelines in mixed liquor service.</p>
<p><b>Ion interference.</b> Ammonium ISE membranes select for NH4+ based on ionic size, but potassium (K+) and sodium (Na+) interfere because they share similar hydrated ionic radii. At a municipal plant with a stable industrial load, K+ and Na+ concentrations are roughly constant and the interference is corrected during calibration. But when industrial contributions or stormwater dilution shift the background ion matrix, a calibration accurate two weeks ago introduces systematic bias without any physical change to the sensor. The <a href="https://www.standardmethods.org/" rel="noopener">Standard Methods for the Examination of Water and Wastewater</a> (Method 4500-NH3) documents the ionic strength adjustment and interference correction procedures that apply to ISE ammonium measurement in wastewater matrices.</p>
<p><b>Temperature effects.</b> ISE sensitivity follows the Nernst equation, which is temperature-dependent. A 10 degree C shift in sample temperature changes the theoretical electrode slope by roughly 3.4 percent. Most instruments include automatic temperature compensation, but if the compensation was characterized during summer commissioning, it may carry a fixed offset in winter operation.</p>
<p>None of these failure modes produce an alarm. The sensor stays online, the reading stays in a plausible range, and the ABAC loop responds to a signal that does not match the actual ammonium concentration in the basin.</p>
<h2 id="what-drift-does-to-the-control-loop">What drift does to the control loop</h2>
<p>Consider an ABAC plant where the target is 2.0 mg/L NH4-N at the end of the aeration zone. The control logic holds DO at a setpoint calculated as: DO setpoint = 1.5 + (measured ammonium / 2). At true ammonium = 1.0 mg/L, the DO setpoint is 2.0 mg/L.</p>
<p>If the ammonium sensor drifts positive by 1.5 mg/L, the measured reading is 2.5 mg/L where the true value is 1.0 mg/L. The control loop calculates a DO setpoint of 2.75 mg/L instead of 2.0 mg/L, a 37.5 percent increase in the target. The blower runs harder to hold that inflated setpoint.</p>
<p>In a modelled example for a 5 MGD activated sludge plant with 200 kW of installed blower capacity running at 65 percent load, a 37.5 percent increase in DO setpoint translates to roughly 75 kW of additional blower load. (This is a simplification; <a href="/blog/aeration-energy-cost-wastewater-plants.html">aeration energy scales non-linearly with airflow and diffuser backpressure</a>, and actual response depends on blower type and control architecture.) At $0.10/kWh operating continuously, that is approximately $5,400 per month in avoidable blower cost for every month the bias persists undetected. This is a modelled example; actual costs depend on plant size, blower efficiency, diffuser condition, and local electricity tariff.</p>
<p>A negative drift carries permit risk. If the ammonium sensor reads low when ammonium is actually elevated, the ABAC system backs off aeration while the actual nitrogen load requires more. <a href="/blog/nitrogen-permit-compliance-sampling-gap.html">NPDES nitrogen permit limits are evaluated against 30-day and 7-day average discharge values</a>, so a persistent bias can accumulate into a reportable violation even if no individual grab sample triggers an immediate alarm.</p>
<h2 id="where-the-problem-hides">Where the problem hides</h2>
<p>In a plant running <a href="/blog/do-sensor-calibration-interval-activated-sludge.html">continuous DO sensor monitoring</a>, a step change or sudden sensor failure is usually visible: the DO reading spikes or drops, the blower response changes, and the operator notices. Gradual ammonium sensor drift does not produce a visible step. The ABAC loop chases the drifting signal smoothly; blower energy creeps upward over two to six weeks; effluent quality degrades slowly.</p>
<p>The diagnostic signature is a divergence between the online ammonium reading and the periodic grab sample from the same location. Most plants collect a grab sample from the end of the aeration zone one to three times per week. If the online reading and the grab sample disagree by more than the expected diurnal and analytical variation, the sensor needs investigation.</p>
<p>In <a href="/blog/intermittent-aeration-nitrogen-removal-do-control.html">intermittent aeration systems running ABAC</a>, where the ammonium signal drives both the DO setpoint and the aerobic/anoxic cycle timing, a drifted sensor can simultaneously waste energy and distort the nitrogen removal cycle. The performance loss compounds across both dimensions.</p>
<h2 id="calibration-and-validation-intervals">Calibration and validation intervals</h2>
<p>Published guidance from the <a href="https://www.wef.org/resources/publications/" rel="noopener">Water Environment Federation Nutrient Removal Manual of Practice No. 34</a> recommends site-specific calibration schedules for online analyzers based on the fouling rate observed during commissioning, updated when the operating load changes significantly. The EPA Nutrient Control Design Manual (EPA/600/R-10/100) recommends verifying online analyzers against grab samples at least weekly during initial commissioning and recalibrating when deviations exceed 10 percent of the measured value.</p>
<p>A practical starting point for most activated sludge plants in continuous aeration service:</p>
<ul class="post-list"><li>Weekly: compare the online ammonium reading to a grab sample analyzed on-site or sent to a certified laboratory.</li><li>Every two weeks: inspect the reference electrode junction and replace fill solution if cloudy or discolored.</li><li>Monthly: run a two-point calibration using fresh standard solutions prepared from a traceable standard.</li><li>After any chemical slug load such as an industrial spill, septage acceptance, or a step change in industrial contribution: re-verify immediately.</li></ul>
<h2 id="detection-with-continuous-data">Detection with continuous data</h2>
<p>A sensor drifting gradually produces a consistent bias rather than random noise. If historical data is available from a SCADA historian or continuous monitoring platform, plotting the rolling difference between online readings and grab sample values over time reveals whether the divergence is growing. A trend in that difference is the signature of gradual drift; random scatter around zero indicates a calibrated sensor.</p>
<p>The same pattern-based approach applies across <a href="/blog/spectral-change-analysis-vs-calibrated-parameters.html">multiple sensors in parallel or sequential service</a>: when two ammonium sensors at different basin locations both drift in the same direction at the same rate, a shared environmental driver such as a temperature shift or ion matrix change is more likely than independent mechanical failure. Separating those two causes matters because the corrective actions are different.</p>
<p>Ammonium sensor drift in ABAC service is predictable, slow-moving, and detectable. That makes it preventable. The cost of a weekly grab-sample comparison and a monthly calibration is a fraction of one month of avoidable blower energy or a single permit exceedance.</p>
<hr />
<p>If your plant runs ABAC or is evaluating ammonium-based setpoint control, AquaMesh can show you how continuous ammonium monitoring integrates with existing SCADA and historian data to detect sensor drift before it affects blower operation. <a href="https://aquamesh.ai/#contact">Request a demo.</a></p>
<p><em>Alex Towfigh is a founder at AquaMesh.</em></p>]]></content:encoded>
      <category>wastewater</category>
      <category>sensors</category>
      <category>aeration</category>
      <category>nitrogen</category>
      <category>activated sludge</category>
    </item>
    <item>
      <title>pH sensor drift in activated sludge: operating cost</title>
      <link>https://aquamesh.ai/blog/ph-sensor-drift-wastewater-activated-sludge.html</link>
      <guid isPermaLink="true">https://aquamesh.ai/blog/ph-sensor-drift-wastewater-activated-sludge.html</guid>
      <pubDate>Fri, 06 Nov 2026 09:00:00 GMT</pubDate>
      <description>A pH electrode drifting 0.5 units in an activated sludge basin drives unnecessary chemical dosing and masks nitrification risk. What that costs in practice.</description>
      <content:encoded><![CDATA[<p>A pH electrode running continuously in an activated sludge bioreactor rarely fails cleanly. It drifts. The glass membrane picks up biological film. The reference junction (the small porous frit that completes the electrical circuit) clogs with mixed liquor solids. By the time the weekly calibration reveals a 0.4-unit offset, the process controller has spent several days adding chemistry to correct a reading that was already wrong.</p>
<p>That is the problem this post addresses. Not sensor failure, but the quiet, accumulating cost of biased pH measurement in a biological nutrient removal process.</p>
<p><b>Quotable insight: A pH sensor reading 6.8 when the bioreactor is actually at 7.3 is not a measurement problem: it is a chemical dosing problem and a compliance risk, compounded over every day between calibrations.</b></p>
<hr />
<h2 id="why-ph-electrodes-drift-in-activated-sludge">Why pH electrodes drift in activated sludge</h2>
<p>Activated sludge is an unusually hostile medium for a glass pH electrode. Mixed liquor suspended solids (MLSS) in a typical activated sludge basin run between 1,500 and 4,000 mg/L. That biological mass contains filamentous organisms, extracellular polymeric substances, oils from influent, and carbonate and phosphate compounds that scale surfaces.</p>
<p>Three mechanisms drive drift in this environment:</p>
<p><b>Liquid junction fouling.</b> The reference electrode inside a pH probe makes contact with the sample through a porous junction (typically a ceramic frit, a PTFE pin, or a fiber wick). In activated sludge, biological solids and scale progressively block that junction. As flow across the junction slows, a junction potential builds up that shifts the measured voltage in a direction that looks like a pH decrease. A frit junction that was calibrated in a clean buffer at 7.00 and then immersed in mixed liquor can read 0.2 to 0.5 units low within four to seven days, without any change in actual sample pH. Ion-selective electrode (ISE) ammonium sensors use the same liquid junction design and face the same fouling mechanism in activated sludge; how that accumulates into an aeration control error is covered in <a href="/blog/ammonia-sensor-drift-wastewater-aeration-control.html">ammonia sensor drift in wastewater aeration control</a>.</p>
<p><b>Glass membrane fouling.</b> The glass membrane at the tip of the electrode is sensitive to surface charge. A thin biofilm or scale layer on the membrane blunts the response, increasing response time and flattening the measurement near the setpoint. This is less common than junction fouling but produces a different pattern: sluggish response to rapid pH swings rather than a steady offset.</p>
<p><b>Temperature compensation errors.</b> Most in-line pH transmitters apply automatic temperature compensation (ATC) using the Nernst equation (59.16 mV/pH unit at 25 degrees C, adjusting with temperature). A fouled temperature sensor, or one positioned poorly relative to the pH sensor in a mixed liquor flow, can add a further 0.1 to 0.2 unit error under typical bioreactor conditions (16 to 28 degrees C).</p>
<p>The <a href="https://www.epa.gov/sites/default/files/2019-04/documents/nutrient-control-design-manual-epa-600-r-10-100.pdf" rel="noopener">EPA Nutrient Control Design Manual</a> notes that pH is one of the most important parameters to maintain in biological nutrient removal (BNR) processes, with nitrification strongly affected by pH values below 7.0 and above 8.5. The manual identifies optimal nitrification pH as approximately 7.5 to 8.0 (Section 3). The cost of getting that range wrong starts with chemicals and ends with biology.</p>
<hr />
<h2 id="what-a-biased-reading-costs-in-chemical-dosing">What a biased reading costs in chemical dosing</h2>
<p>Nitrification consumes alkalinity. For every milligram of ammonium-nitrogen (NH4-N) oxidized, approximately 7.1 mg of alkalinity is consumed as calcium carbonate equivalent. Most activated sludge plants operating for biological nitrogen removal must supplement influent alkalinity to keep pH from dropping below 7.0 and inhibiting nitrifiers. The <a href="https://www.epa.gov/sites/default/files/2019-04/documents/nutrient-control-design-manual-epa-600-r-10-100.pdf" rel="noopener">EPA Nutrient Control Design Manual</a> recommends maintaining residual alkalinity above 50 to 80 mg/L as CaCO3 through the aerobic zone.</p>
<p>The control system for alkalinity addition (typically sodium bicarbonate, sodium hydroxide, or carbon dioxide) uses the pH reading as its primary feedback signal.</p>
<p>When that signal is biased low, the controller adds chemical it does not need.</p>
<p><b>Modelled example: sodium bicarbonate addition (do not use as a cost estimate for your plant; these are illustrative figures):</b></p>
<p>A 5 MGD secondary treatment plant with biological nitrification adds approximately 500 to 800 lbs per day of sodium bicarbonate (soda ash) to maintain alkalinity through the aerobic zone. At a bulk price of roughly $0.10 to $0.15 per pound (commodity sodium bicarbonate, industry trade range), that is $18,000 to $44,000 per year in chemical cost.</p>
<p>A pH electrode reading 0.4 units low (the controller sees pH 7.1 when the basin is actually at 7.5) causes the controller to add alkalinity it does not need. If the result is a 15 percent chemical overdose, that is roughly $2,700 to $6,600 per year in unnecessary chemical spend at this modelled plant size.</p>
<p>At larger plants (20 to 50 MGD) or those using sodium hydroxide for pH control, the proportional waste is larger in absolute dollars. CO2 injection systems for pH reduction (in high-alkalinity influent situations) can run $15,000 to $60,000 per year in gas costs; a systematic overdose from a biased high-reading pH probe proportionally compounds that cost.</p>
<p>The chemical cost alone rarely justifies a sensor upgrade. The process stability risk usually does.</p>
<hr />
<h2 id="the-nitrification-stability-risk">The nitrification stability risk</h2>
<p>Nitrifying bacteria (primarily Nitrosomonas and Nitrobacter) are the slowest-growing organisms in a conventional activated sludge process. Specific growth rates for Nitrosomonas are typically 0.5 to 0.9 per day at 20 degrees C, compared to 5 to 15 per day for heterotrophic organisms. That slow growth means that if nitrifiers are inhibited or washed out, recovery takes two to four weeks under favorable conditions.</p>
<p>pH below 6.5 is a reliable nitrification inhibitor. Below that threshold, Nitrosomonas activity drops sharply and can cease within hours for a sustained low-pH event.</p>
<p>The <a href="https://www.epa.gov/sites/default/files/2019-04/documents/nutrient-control-design-manual-epa-600-r-10-100.pdf" rel="noopener">EPA Nutrient Control Design Manual</a> documents the sensitivity: a drop from pH 7.2 to 6.5 can reduce nitrification rates by 50 percent or more. A sustained drop below 6.0 can cause near-complete nitrification failure in a conventional activated sludge process.</p>
<p>Here is the drift scenario that matters operationally: the pH sensor is reading 0.4 units high (fouled membrane, positive offset). The basin is actually at 6.7. The controller, seeing 7.1, adds no alkalinity. The pH continues to drop. The operator, seeing a trending decline on a screen that shows 6.8 when the basin is at 6.4, adds alkalinity, but perhaps not enough and perhaps too late. Three days later, the effluent NH4-N spikes. The root cause is a biased electrode.</p>
<p>This scenario is not rare. The <a href="https://www.epa.gov/sites/default/files/2015-09/documents/primer.pdf" rel="noopener">EPA Primer for Municipal Wastewater Treatment Systems</a> cites nitrification upsets from inadequate alkalinity as a recurring operational challenge in activated sludge. In most cases, those upsets were detectable earlier in the historian data, if the right parameters were being measured accurately.</p>
<hr />
<h2 id="the-gap-between-calibration-events">The gap between calibration events</h2>
<p>NPDES permits typically require pH monitoring of final effluent daily or continuously, under the monitoring provisions required as permit conditions by <a href="https://www.ecfr.gov/current/title-40/chapter-I/subchapter-D/part-122/section-122.48" rel="noopener">40 CFR 122.48</a>. In-process pH (the bioreactor reading that drives chemical dosing) is measured continuously by the in-line electrode but calibrated on a weekly or bi-weekly schedule at many facilities.</p>
<p>That interval is the window in which drift accumulates uncorrected.</p>
<p>A junction that fouls at a rate of 0.05 to 0.08 pH units per day (a plausible rate in a high-MLSS basin) produces a 0.35 to 0.56 unit error by day seven. For six of those seven days, the chemical dosing and stability assessment were based on data that was off by a meaningful fraction.</p>
<p>The standard correction is to calibrate more frequently. That is a valid approach and appropriate for critical processes. But it adds labor and requires taking the sensor offline or using a submersible probe that can be withdrawn without process interruption.</p>
<p>A complementary approach is to detect drift earlier without additional calibration cycles. In an activated sludge process, pH is not independent of dissolved oxygen, oxidation-reduction potential (ORP), and conductivity. When nitrification is proceeding normally, these parameters have predictable correlations on a day-over-day basis. A pH reading that diverges from what the correlated signals suggest, without a corresponding change in influent chemistry or loading, is a candidate for drift, not process change.</p>
<p>That distinction is what continuous, multi-parameter monitoring with drift detection logic enables: identifying the difference between a real pH excursion and a sensor that needs cleaning, before the calibration event makes it obvious.</p>
<hr />
<h2 id="what-to-look-for-in-your-data">What to look for in your data</h2>
<p>If your plant has historian data with pH, DO, ORP, and conductivity logged at one-minute or five-minute intervals, the following patterns are consistent with pH electrode fouling rather than process change:</p>
<ul class="post-list"><li>A slow, unidirectional drift in pH over three to five days, without a corresponding change in influent flow, temperature, or DO setpoint behavior</li><li>pH response that lags sharply after a process change (blower speed change, recycle flow adjustment) when DO responds in the expected direction within seconds</li><li>A pH reading that is inconsistent with the alkalinity addition rate the controller is commanding (for example, the controller adding significant caustic but pH not trending upward)</li></ul>
<p>None of these is definitive without a calibration check. But any of them is a prompt to calibrate early and to look at the post-calibration offset. A plant that consistently finds 0.3 to 0.5 unit offsets at calibration is running on biased data between calibrations, and the cost of that is a combination of chemical waste and process stability margin that is not there when the biology actually needs it.</p>
<hr />
<h2 id="a-next-step">A next step</h2>
<p>If you want to understand how continuous sensor monitoring applies to your activated sludge process, specifically how correlated parameter analysis can flag pH electrode fouling between calibration events, <a href="https://aquamesh.ai/#contact">request a demo</a> and we can walk through an example with your historian data.</p>
<p>For more on the measurement gap in wastewater treatment, see <a href="/blog/what-25-days-of-historian-data-shows.html">what 25 days of historian data actually shows about a treatment plant</a>, <a href="/blog/do-sensor-calibration-interval-activated-sludge.html">DO sensor calibration intervals in activated sludge</a>, and <a href="/blog/intermittent-aeration-nitrogen-removal-do-control.html">intermittent aeration and DO sensor precision</a>. For NPDES permit compliance implications of in-process measurement gaps, see <a href="/blog/nitrogen-permit-compliance-sampling-gap.html">NPDES nitrogen permit compliance: the DMR sampling gap</a>.</p>
<p><em>Alex Towfigh is a founder at AquaMesh.</em></p>]]></content:encoded>
      <category>pH sensor</category>
      <category>activated sludge</category>
      <category>chemical dosing</category>
      <category>nitrification</category>
      <category>sensor drift</category>
      <category>wastewater treatment</category>
      <category>NPDES</category>
    </item>
    <item>
      <title>NPDES nitrogen permit compliance: the DMR sampling gap</title>
      <link>https://aquamesh.ai/blog/nitrogen-permit-compliance-sampling-gap.html</link>
      <guid isPermaLink="true">https://aquamesh.ai/blog/nitrogen-permit-compliance-sampling-gap.html</guid>
      <pubDate>Thu, 05 Nov 2026 09:00:00 GMT</pubDate>
      <description>NPDES nitrogen permits assess compliance from a handful of sampled days. The unsampled days are invisible to the DMR. Here is why that gap matters.</description>
      <content:encoded><![CDATA[<p>A wastewater plant producing average effluent total nitrogen of 6 mg/L against a monthly average permit limit of 8 mg/L looks comfortable on paper. The treatment is working. The monthly discharge monitoring reports (DMRs) are clean. But if that plant is sampling twice a week, its monthly compliance record is built from eight or nine data points. The 20 or 21 days that produce no sample are invisible to the permit record.</p>
<p>On those unsampled days, a process event -- a cold snap reducing nitrifier activity, a storm event diluting the activated sludge, a biosolids return load arriving at the wrong time in the cycle -- could push effluent nitrogen well above the daily maximum limit. It will not appear in the DMR. And the plant has no data to show it did not happen.</p>
<p>This is the compliance exposure that most discussions about nitrogen permit compliance skip. Getting to a low average is a treatment problem. Knowing what your process was doing between sample events is a measurement problem.</p>
<p><b>A wastewater plant sampling twice a week builds its monthly compliance record from 8 or 9 data points. The other 21 days are invisible to the DMR.</b></p>
<h2 id="how-npdes-nitrogen-permits-are-structured">How NPDES nitrogen permits are structured</h2>
<p>Under <a href="https://www.ecfr.gov/current/title-40/chapter-I/subchapter-D/part-122/section-122.44" rel="noopener">40 CFR 122.44(d)(1)</a>, permit writers must set water quality-based effluent limits for a pollutant whenever technology-based limits alone are insufficient to protect receiving water quality standards. For nitrogen, this means permit limits are typically derived from the nutrient criteria applicable to the downstream waterbody -- a phosphorus-impaired estuary, a Chesapeake Bay tributary under nutrient trading regulations, or a nitrogen-sensitive groundwater recharge zone.</p>
<p>A typical nitrogen permit for a 5 to 50 MGD municipal wastewater treatment plant includes:</p>
<ul class="post-list"><li>A <b>monthly average limit (MAL)</b> -- the arithmetic mean of all valid samples collected in the calendar month</li><li>A <b>weekly average limit</b> -- the mean of samples in any calendar week, typically set 1.5x the MAL</li><li>In permits for sensitive receiving waters, a <b>daily maximum</b> -- a hard ceiling that no single composite sample should exceed</li></ul>
<p>Compliance is assessed against the values the permittee reports in the monthly DMR, submitted under <a href="https://www.ecfr.gov/current/title-40/chapter-I/subchapter-D/part-122/section-122.41" rel="noopener">40 CFR 122.41(l)</a>. The DMR calculation uses only the samples the permit requires the facility to collect. If the permit specifies twice-weekly grab samples for total nitrogen, the monthly compliance determination rests on those eight or nine values and nothing else.</p>
<p>A plant with a TN MAL of 8 mg/L, a weekly average limit of 12 mg/L, and a daily maximum of 20 mg/L is not at risk from exceedances its sampling program cannot detect -- unless it has a daily maximum limit and a process that can spike above 20 mg/L, or unless a state regulator or EPA inspector collects an independent sample on an unmonitored day.</p>
<h2 id="the-math-behind-sampling-gaps">The math behind sampling gaps</h2>
<p>The <a href="https://www.epa.gov/sites/default/files/2015-06/documents/nutrient-control-design-manual-full.pdf" rel="noopener">EPA Nutrient Control Design Manual (2010)</a> documents the process variability that biological nitrogen removal systems encounter in normal operation: influent load variation, diurnal flow patterns, seasonal temperature effects on nitrifier kinetics, return activated sludge rate fluctuations, and dissolved oxygen control performance all affect day-to-day effluent nitrogen. Plants designed to achieve 6 mg/L TN on an average day do not produce exactly 6 mg/L every day.</p>
<p>As a modelled example: consider a 10 MGD activated sludge plant with a TN monthly average limit of 8 mg/L, operating at a design-day average of 6 mg/L. Its permit requires twice-weekly composite samples. In a 30-day month, it submits nine valid TN values to the DMR. The remaining 21 days produce no sample.</p>
<p>If on two of those unsampled days a process event -- a sudden drop in aerobic zone dissolved oxygen during an intermittent aeration cycle, as described in <a href="/blog/intermittent-aeration-nitrogen-removal-do-control.html">Intermittent aeration for nitrogen removal: what DO sensor precision actually controls</a> -- causes effluent TN to spike to 13 mg/L, the monthly average across all actual daily discharges would be approximately 6.4 mg/L. The DMR monthly average, calculated from the nine sampled days, might show 6.1 mg/L. Both are compliant. One is more accurate.</p>
<p>The gap between the two is not a records violation. It is simply what scheduled sampling is designed to produce: a statistically valid estimate of compliance, not a continuous record of actual discharge quality.</p>
<h2 id="where-the-regulatory-exposure-lies">Where the regulatory exposure lies</h2>
<p>NPDES compliance is not self-reported in isolation. Under <a href="https://www.ecfr.gov/current/title-40/chapter-I/subchapter-D/part-122/section-122.41" rel="noopener">40 CFR 122.41(h)</a>, permittees must allow EPA and state personnel to enter the facility and collect independent samples at any reasonable time. States and EPA Regions conduct Compliance Evaluation Inspections (CEIs) to verify that reported data is accurate and that effluent quality matches the DMR pattern.</p>
<p>An inspector who collects an independent grab sample on a day the facility did not sample, and finds a result significantly above the permit limit, has found something the DMR does not show. This does not automatically constitute a permit violation -- a single sample is not an MAL exceedance. But it can prompt a more intensive review, an accelerated inspection schedule, or a compliance schedule discussion.</p>
<p>The exposure is higher for plants in nutrient-sensitive watersheds where regulators are actively looking at effluent quality data. Chesapeake Bay basin states have operated nutrient trading programs since the 2000s; Virginia and Maryland permit writers have access to both DMR data and periodic independent sampling results. A plant with a persistent gap between its reported DMR quality and what independent sampling finds is in a difficult position to explain.</p>
<h2 id="what-continuous-process-monitoring-changes">What continuous process monitoring changes</h2>
<p>A plant with continuous in-process monitoring -- real-time DO sensors, online nitrogen analyzers, or flow-proportional sampling systems calibrated to known performance curves -- does not eliminate the scheduled sampling requirement. It adds something the DMR cannot provide: a process record for every day of the month.</p>
<p>When an operator can see the aerobic zone DO dropping at 6 PM on a Tuesday, and the anoxic zone nitrate residual climbing the following morning, there is an opportunity to intervene before the effluent crosses a threshold. For plants with in-process ammonium analyzers running ammonia-based aeration control (ABAC), that ammonium signal provides a direct real-time view of aerobic zone nitrification performance -- but only to the degree the sensor is tracking actual ammonium concentration between service events: <a href="/blog/ammonia-sensor-drift-wastewater-aeration-control.html">ammonia sensor drift in wastewater aeration control</a>. The data from the <a href="/blog/what-25-days-of-historian-data-shows.html">What 25 days of historian data actually shows about a treatment plant</a> post illustrates how process variability that is invisible in scheduled compliance samples becomes apparent in continuous SCADA records.</p>
<p>For audit purposes, a continuous process record also provides something that no DMR can: a basis for demonstrating what the process was doing on days when no effluent sample was collected. That is a different compliance posture than a facility that can only show its nine sampled days were clean.</p>
<h2 id="a-practical-starting-point">A practical starting point</h2>
<p>Most wastewater utilities do not have a clear picture of how much variability exists between their scheduled sample days. The facility knows its design-basis performance and trusts that scheduled samples are representative.</p>
<p>Whether they are representative depends on what the process was actually doing on the other 21 days of the month. That answer lives in the SCADA historian -- if it is being used. The <a href="/blog/aeration-energy-cost-wastewater-plants.html">Aeration energy at wastewater plants: the cost, the waste, and the measurement gap</a> post covers what plant historian data typically shows about aeration performance variability across operating days. The <a href="/blog/wastewater-blower-vfd-energy-savings-erode.html">wastewater blower VFD energy savings erode</a> post shows the same dynamic from the energy side: a DO sensor drifting between calibration events causes the control loop to optimize against the wrong input, with no alarm raised and no indication in the daily operational record until someone looks at the historian across a longer window. <a href="/blog/do-sensor-calibration-interval-activated-sludge.html">DO sensor calibration intervals in activated sludge</a> covers how calibration frequency directly affects the measurement precision that determines when nitrogen compliance exposure builds.</p>
<p>Understanding your actual variability is the first step. It tells you whether your current monitoring program is adequate for the permit limits you are operating under -- and whether the unsampled days in your DMR represent a real exposure or a comfortable margin. For industrial dischargers to the municipal system, a parallel sampling gap exists at the pretreatment level: monthly or quarterly grab samples taken during production runs do not capture cleaning cycles or batch upsets: <a href="/blog/industrial-pretreatment-self-monitoring-gaps.html">industrial pretreatment: the self-monitoring blind spot</a>. Where continuous online sensors fit into the compliance picture alongside lab sampling requirements is covered in <a href="/blog/online-sensors-vs-lab-sampling-wastewater-discharge.html">online sensors vs. lab sampling for discharge compliance</a>.</p>
<p>AquaMesh works with treatment plants and utilities to analyze existing plant data and identify where measurement programs could be doing more to support both operational control and compliance confidence. If you want to see what your process historian shows about the days between your compliance samples, <a href="https://aquamesh.ai/#contact">request a demo</a>.</p>
<p><em>Alex Towfigh is a founder at AquaMesh.</em></p>]]></content:encoded>
      <category>nitrogen</category>
      <category>NPDES</category>
      <category>permit compliance</category>
      <category>wastewater</category>
      <category>monitoring</category>
      <category>DMR</category>
      <category>effluent limits</category>
    </item>
    <item>
      <title>Wastewater blower VFD: why energy savings erode</title>
      <link>https://aquamesh.ai/blog/wastewater-blower-vfd-energy-savings-erode.html</link>
      <guid isPermaLink="true">https://aquamesh.ai/blog/wastewater-blower-vfd-energy-savings-erode.html</guid>
      <pubDate>Sat, 31 Oct 2026 09:00:00 GMT</pubDate>
      <description>DO sensor drift silently reverses VFD energy savings at wastewater plants. Here are three historian patterns that expose the cause and how to fix it.</description>
      <content:encoded><![CDATA[<p>Variable frequency drives on aeration blowers are among the most reliable energy investments a wastewater plant can make. The engineering is sound: slow the blower when oxygen demand drops, and power consumption falls by the cube of the speed reduction. Many plants documented 15 to 25 percent reductions in blower energy in the months after commissioning. Not all of them still see those numbers.</p>
<p>The most common reason is not a failing drive, a degraded motor, or the wrong control strategy. It is the dissolved oxygen sensor feeding the control loop.</p>
<p><b>Quotable insight:</b> A blower VFD can only be as efficient as the measurement telling it when to slow down. If that sensor reads low, the controller drives blowers harder, and the savings disappear into electricity you cannot account for.</p>
<h2 id="how-vfd-do-control-is-supposed-to-work">How VFD DO control is supposed to work</h2>
<p>In a properly functioning aeration control loop, a dissolved oxygen sensor reports basin concentration to a controller. The controller compares that reading to a setpoint, typically 1.5 to 2.5 mg/L for activated sludge, and adjusts blower output through one or more variable frequency drives. When influent load drops overnight, demand-based control backs the blowers off. When load rises in the morning, blowers ramp back up.</p>
<p>The energy arithmetic behind this is centrifugal fan physics. For a multi-stage centrifugal or turbo blower, power consumption scales with the cube of rotational speed. A blower running at 80 percent of design speed uses approximately 51 percent of design power. At 90 percent of design speed, it uses approximately 73 percent. The difference between those two operating points, 22 percentage points of rated power, is what demand-based control captures when it works as designed.</p>
<p>The Florida Water and Pollution Control Operators Association published analysis in April 2026 showing that demand-based DO control typically reduces blower energy by 10 to 25 percent at secondary treatment plants, and that PLC-optimized control can reach 20 to 40 percent. (<a href="https://www.fwpcoa.org/content.aspx?page_id=5&amp;amp;club_id=859275&amp;amp;item_id=134074" rel="noopener">FWPCOA, "Demand-Based DO Control And Energy Optimization," April 2026</a>) Those ranges describe what the control system saves when the sensor is providing an accurate signal.</p>
<h2 id="the-sensor-drift-that-quietly-reverses-it">The sensor drift that quietly reverses it</h2>
<p>Dissolved oxygen sensors in mixed liquor face a relentless fouling environment. Biofilm builds on membrane-covered electrochemical sensors and on the optical surfaces of luminescent sensors. Suspended solids accumulate on probe housings. The most common consequence of fouling is a reading that drifts low: the deposit impedes diffusion or light transmission, and the sensor reports less oxygen than is actually present.</p>
<p>A sensor reading 1.4 mg/L when the basin contains 2.1 mg/L presents the controller with a problem that does not exist. The basin is already 0.6 mg/L above setpoint. The controller sees it 0.1 mg/L below setpoint and commands higher blower output. The VFD speeds up. Over-aeration begins, and it is completely silent: no alarms fire, the historian records the sensor value as if it were accurate, and the control loop looks like it is working.</p>
<p>The result is a system where the VFD is still installed, the control loop is still active, and the blowers are running harder than the process needs. The savings documented at commissioning, when the sensors were freshly calibrated, erode as sensor bias accumulates between calibration events.</p>
<p>The same pattern appears in other process control loops where a single sensor drives an automated response. In cooling tower systems, conductivity sensor drift causes blowdown cycles to run incorrectly: purging more treated water than water chemistry requires, or leaving scale-forming chemistry in place when it should be removed, as described in <a href="/blog/cooling-tower-blowdown-conductivity-sensor-drift.html">cooling tower blowdown: conductivity control and sensor drift</a>.</p>
<h2 id="what-the-cost-of-eroded-savings-looks-like">What the cost of eroded savings looks like</h2>
<p>The U.S. EPA benchmarks secondary treatment activated sludge plants at approximately 1,500 kWh per million gallons for electricity consumption, with aeration systems representing 50 to 75 percent of that total. (<a href="https://www.epa.gov/sites/default/files/2016-12/documents/guide_energymanagement.pdf" rel="noopener">U.S. EPA, Ensuring a Sustainable Future: An Energy Management Guidebook for Wastewater and Water Utilities</a>) The U.S. Energy Information Administration reported the national average industrial electricity rate at $0.0977 per kWh in July 2026. (<a href="https://www.eia.gov/electricity/monthly/epm_table_grapher.php?t=epmt_5_6_a" rel="noopener">EIA Electric Power Monthly, September 2026</a>)</p>
<p><b>Modelled example for a 10 MGD secondary treatment plant:</b></p>
<div class="table-wrap"><table class="cs-table"><thead><tr><th></th><th></th></tr></thead><tbody><tr><td>Annual aeration energy at benchmark</td><td>3,285,000 kWh/yr</td></tr><tr><td>Aeration electricity cost at $0.098/kWh</td><td>~$322,000/yr</td></tr><tr><td>Savings at commissioning (20%, demand-based control)</td><td>~$64,000/yr</td></tr><tr><td>Savings after 18 months of sensor drift (reduced to ~8%)</td><td>~$26,000/yr</td></tr><tr><td>Annual value of eroded savings</td><td>~$38,000/yr</td></tr></tbody></table></div>
<p>This is a modelled example using published benchmarks and illustrative drift-induced savings erosion. Actual figures depend on load profile, treatment configuration, calibration schedule, and utility tariff. A plant running at higher flow or with older blowers will see larger absolute numbers. A plant with weekly calibration will see less erosion.</p>
<p>The point of the table is the ratio: the cost of sensor drift that is invisible on the dashboard can exceed the annual maintenance budget for the sensors causing it.</p>
<h2 id="three-patterns-that-appear-in-the-historian">Three patterns that appear in the historian</h2>
<p>When VFD savings have eroded due to sensor drift, three patterns tend to appear in historical process data. None of them requires new instrumentation to find.</p>
<p><b>Blowers at sustained high output during low-demand periods.</b> A functioning demand-based control system should show measurable blower turndown overnight and on weekends, when influent flow and biological oxygen demand are lower. If the historian shows blowers running at 85 percent or higher during those periods, either the setpoint is configured too aggressively or the sensors are overreporting the oxygen deficit.</p>
<p><b>Step changes in blower output immediately after calibration events.</b> When a sensor has drifted low by 0.5 to 1.0 mg/L and is then recalibrated to the correct value, the control loop sees an immediate improvement and commands lower blower output. That step down shows up in the historian. Its size reveals how far the sensor had drifted, and how long the plant was paying for that drift since the last calibration. A plant running monthly calibration on a sensor that drifts 0.5 mg/L in three weeks has paid for that drift for three weeks before the correction.</p>
<p><b>Divergent blower output across parallel basins.</b> At plants with multiple aeration basins drawing from a common air header, sensors in different basins drift at different rates depending on local solids loading and maintenance consistency. When two basins receive similar influent loads but their blower outputs differ by 20 percent or more, the most common cause is a measurement offset in one of the sensors. The control loop is faithfully responding to what the sensors report. What the sensors report has diverged from reality.</p>
<p>We documented this exact pattern in <a href="/blog/what-25-days-of-historian-data-shows.html">25 days of historian data from an operating plant</a>: one basin was running its blower at more than three times the DO level of its parallel twin, with the blower pinned at maximum output for the full duration of the record. Both basins drew from the same air header. The control loop was functioning. The deviation was worth $3,316 per year in direct blower power, and it was invisible without comparing basins side by side at hourly resolution.</p>
<p>The mechanics of parallel equipment drift in aeration systems, and why the gap tends to grow rather than self-correct, are covered in <a href="/blog/why-parallel-equipment-drifts.html">why parallel equipment drifts apart and what the gap costs</a>.</p>
<h2 id="the-calibration-interval-question">The calibration interval question</h2>
<p>There is no universal answer for how often a DO sensor in activated sludge service should be calibrated. Manufacturer recommendations typically range from weekly to monthly. Actual calibration frequency at operating plants varies considerably, and some plants rely on grab samples and visual inspection rather than a fixed interval.</p>
<p>What determines the right interval is the sensor's actual drift rate in your mixed liquor at your current solids loading. A sensor that drifts 0.1 mg/L per week and holds that drift until the membrane is replaced needs a different schedule than one that degrades faster. The only way to know which situation you are in is to compare sensor readings against a calibrated handheld reference at intervals shorter than your current calibration cycle, not just immediately after calibration.</p>
<p>A calibration-versus-drift log, built from a few weeks of parallel readings, makes the calibration interval decision quantitative rather than a guess based on manufacturer defaults. We examine this approach in detail in <a href="/blog/do-sensor-calibration-interval-activated-sludge.html">DO sensor calibration intervals in activated sludge</a>, including how to build a drift profile and what a typical profile looks like in activated sludge service.</p>
<p>For plants running <a href="/blog/intermittent-aeration-nitrogen-removal-do-control.html">intermittent aeration for simultaneous nitrification and denitrification</a>, the stakes around calibration intervals are higher. That control strategy uses precise DO swings through the nitrification setpoint to cycle between aerobic and anoxic conditions. A sensor with a standing offset of 0.5 mg/L can shift the timing of those swings in ways that reduce nitrogen removal efficiency and create the compliance exposure described in <a href="/blog/nitrogen-permit-compliance-sampling-gap.html">NPDES nitrogen permit compliance: the DMR sampling gap</a>, entirely apart from the energy cost. In BNR processes, pH electrode drift adds a compounding risk alongside DO offset: biased alkalinity dosing and masked nitrification inhibition are covered in <a href="/blog/ph-sensor-drift-wastewater-activated-sludge.html">pH sensor drift in activated sludge: the cost of 0.5 units</a>.</p>
<h2 id="three-questions-before-adjusting-the-control-strategy">Three questions before adjusting the control strategy</h2>
<p>When blower energy costs are higher than expected, or when early savings have faded, the instinct is to revisit the control logic: adjust setpoints, tune PID parameters, upgrade the control software. That work sometimes helps. It often does not, because the problem is not the strategy but the measurement feeding it.</p>
<p>Before changing anything in the control loop, three questions are worth answering directly from historian and calibration records.</p>
<p><b>How large were the step changes in blower output after the last three calibration events?</b> If each calibration produces a visible step down in blower output, the sensor has been reading low between calibration cycles. The step size tells you how much the drift cost.</p>
<p><b>Do parallel basins show divergent blower outputs during similar load periods?</b> Persistent divergence with no other explanation almost always traces to a sensor offset. It will not resolve with setpoint adjustments.</p>
<p><b>When was the last time sensor readings were verified against a calibrated handheld at midday, not just immediately after calibration?</b> Sensors that look fine at calibration can drift substantially before the next service event. A midday check, two weeks after calibration, shows whether the reading is holding.</p>
<p>None of this analysis requires hardware beyond what most plants already have. The data is usually in the historian. The constraint is having a way to examine it at the right resolution, across basins, over a window long enough to see the patterns clearly. The steps for recovering blower energy once the measurement baseline is confirmed, including DO setpoint reduction, feedback loop verification, pressure tuning, and blower staging, are in <a href="/blog/aeration-energy-cost-wastewater-plants.html">wastewater blower energy: cost and control</a>.</p>
<hr />
<p>If your plant installed blower VFDs with projected savings that have not materialized, or savings that were real initially and have since faded, the measurement quality question is the right place to start. An initial review of historian data from your aeration system can show whether the control loop is working from accurate inputs and where the largest gaps are.</p>
<p>See how AquaMesh approached a related analysis at a membrane bioreactor reuse plant: <a href="https://aquamesh.ai/case-studies/water-reuse-plant.html">a membrane-bioreactor water reuse plant case study</a>. Or <a href="https://aquamesh.ai/#contact">reach out to schedule a working session with the team</a>.</p>
<p><em>Alex Towfigh is a founder at AquaMesh.</em></p>]]></content:encoded>
      <category>blower energy</category>
      <category>VFD</category>
      <category>dissolved oxygen control</category>
      <category>wastewater aeration</category>
      <category>energy optimization</category>
    </item>
    <item>
      <title>DO sensor calibration intervals in activated sludge</title>
      <link>https://aquamesh.ai/blog/do-sensor-calibration-interval-activated-sludge.html</link>
      <guid isPermaLink="true">https://aquamesh.ai/blog/do-sensor-calibration-interval-activated-sludge.html</guid>
      <pubDate>Fri, 30 Oct 2026 09:00:00 GMT</pubDate>
      <description>DO sensor calibration intervals in activated sludge should match your process fouling rate, not the manufacturer default. Here is the historian method.</description>
      <content:encoded><![CDATA[<p>Most wastewater plant SOPs specify dissolved oxygen sensor calibration intervals based on the manufacturer's recommendation. For polarographic membrane sensors in aeration basins, that is typically every two to four weeks. For optical (luminescent) sensors, manufacturers often suggest every one to three months.</p>
<p>These intervals are reasonable starting points but not optimized schedules. Manufacturer recommendations are set to cover the demanding end of the range of applications, which in wastewater means high-MLSS activated sludge with aggressive fouling conditions. If your specific process fouls sensors faster, a monthly schedule runs the risk that your aeration control is operating on significantly drifted data for the last two weeks of each cycle. If your process fouls sensors more slowly, a two-week schedule means extra calibration events with no benefit to process control and real costs in technician time and recalibration-induced step changes.</p>
<p>The relevant question for a wastewater plant operator is not "what does the manufacturer recommend?" but "what is the drift rate of this sensor in this basin, and how long can I run before that drift materially affects my aeration control?"</p>
<h2 id="what-do-calibration-is-actually-doing">What DO calibration is actually doing</h2>
<p>A DO calibration does one thing: it resets the working baseline that your aeration control loop uses to make decisions. During the calibration event, the sensor output is adjusted to match a known reference, typically a two-point calibration against air-saturated water and a zero-oxygen reference. After calibration, the sensor is reading accurately.</p>
<p>Between calibration events, the sensor output drifts. Drift has two main sources: membrane fouling and electrolyte depletion (polarographic), or optical cap fouling and fluorescent dye degradation (luminescent). In both cases, the sensor reads lower than the actual dissolved oxygen concentration as fouling accumulates.</p>
<p>For aeration control, this matters because the blower or diffuser system is responding to the sensor output, not to the actual dissolved oxygen in the basin. A sensor reading 0.8 mg/L when the actual concentration is 1.4 mg/L will hold the blowers on long past the point where the setpoint was met. The basin gets over-aerated. Energy is wasted: the full cost structure of blower energy at wastewater plants is covered in <a href="https://aquamesh.ai/blog/aeration-energy-cost-wastewater-plants.html">wastewater aeration energy: cost and the measurement gap</a>, and the compounding erosion of blower VFD savings from sustained DO drift is documented in <a href="https://aquamesh.ai/blog/wastewater-blower-vfd-energy-savings-erode.html">wastewater blower VFD: why energy savings erode</a>. In an intermittent aeration system used for nitrogen removal, the aerobic phase extends, the anoxic fraction shrinks, and denitrification efficiency drops -- the quantified cost of that shift is covered in <a href="https://aquamesh.ai/blog/intermittent-aeration-nitrogen-removal-do-control.html">intermittent aeration for nitrogen removal: what DO sensor precision actually controls</a>. That efficiency drop on unmonitored days creates the permit-record gap described in <a href="/blog/nitrogen-permit-compliance-sampling-gap.html">NPDES nitrogen permit compliance: the DMR sampling gap</a>. (<a href="https://www.epa.gov/sites/default/files/2019-08/documents/nutrient-control-design-manual.pdf" rel="noopener">EPA Nutrient Control Design Manual, 2010, Section 5</a>) How continuous in-process sensors compare with the official lab-based sampling program for DMR purposes is examined in <a href="/blog/online-sensors-vs-lab-sampling-wastewater-discharge.html">online sensors vs. lab sampling for discharge compliance</a>.</p>
<h2 id="fouling-rates-vary-by-process">Fouling rates vary by process</h2>
<p>Fouling in activated sludge aeration basins is driven primarily by biofilm accumulation on sensor surfaces and entrapment of suspended solids. Both are functions of mixed liquor suspended solids concentration, hydraulic conditions at the sensor location, and aeration intensity.</p>
<p>A sensor positioned directly in a high-turbulence zone of a high-MLSS basin will foul faster than one in a lower-flow zone of a more lightly loaded tank. Optical sensors tend to accumulate a thin biofilm on the luminescent cap that polarographic sensors do not, but they are also not subject to membrane rupture or electrolyte exhaustion. Neither technology eliminates drift; both require calibration at intervals matched to the process.</p>
<p>Standard Methods 4500-O G, which covers the membrane electrode method for dissolved oxygen measurement, recommends calibration whenever "the readings are questioned, when samples of unusual character are analyzed, when temperature changes are sudden, or when the probe is relocated." (<a href="https://www.standardmethods.org/" rel="noopener">Standard Methods for the Examination of Water and Wastewater, 23rd edition, 4500-O G</a>) This is a performance-based standard: calibrate when there is reason to believe the reading has drifted, not on a fixed schedule alone. For a continuous in-line sensor in an aeration basin, that translates to understanding when your process typically produces readings you should question.</p>
<h2 id="using-historian-data-to-calibrate-your-calibration-schedule">Using historian data to calibrate your calibration schedule</h2>
<p>Every time a sensor is calibrated in the field, the technician records the pre-calibration reading and the reference value. If your SCADA historian preserves this data, you have a direct record of how far the sensor drifted from each post-calibration baseline to the next pre-calibration check.</p>
<p>Pulling this data over 12 months and plotting cumulative drift against days since last calibration produces a process-specific drift curve for each sensor location. The curve tells you two things: how fast your sensors typically drift, and whether drift is linear (consistent fouling) or step-function (fouling event, usually a process upset or loading change).</p>
<p>For a plant targeting a DO setpoint of 2.0 mg/L with a control band of 0.5 mg/L, the calibration interval should be set so that expected drift at the end of the cycle stays within a fraction of that tolerance. If your drift curve shows 0.3 mg/L per week, a four-week interval means you could be running with a 1.2 mg/L offset by the end of the cycle. That is enough to hold blowers at full speed when the basin is already adequately aerated, or to trigger alarms that reflect sensor state rather than process state.</p>
<p>At an industrial rate of $0.0977/kWh (<a href="https://www.eia.gov/electricity/monthly/epm_table_grapher.php?t=epmt_5_6_a" rel="noopener">EIA Electric Power Monthly, September 2026</a>) and a typical blower draw of 150 kW, one extra hour of unnecessary aeration per day costs approximately $5.34, or roughly $1,950 per year. Across multiple basins with independent sensors, the cumulative cost of operating with drifted DO data is not trivial.</p>
<p><b>Modelled example: drift cost at two calibration intervals</b></p>
<div class="table-wrap"><table class="cs-table"><thead><tr><th></th><th>2-week calibration cycle</th><th>6-week calibration cycle</th></tr></thead><tbody><tr><td>Assumed drift rate</td><td>0.2 mg/L per week</td><td>0.2 mg/L per week</td></tr><tr><td>Max offset at end of cycle</td><td>0.4 mg/L</td><td>1.2 mg/L</td></tr><tr><td>Estimated excess blower runtime</td><td>Low</td><td>Substantial</td></tr><tr><td>Annual cost (modelled, 150 kW blower)</td><td>Approx. $500</td><td>Approx. $3,200</td></tr></tbody></table></div>
<p>These are modelled figures using a uniform drift assumption. Actual costs depend on blower size, number of sensors, DO setpoints, and site-specific fouling rates. The direction is consistent: intervals that let drift accumulate past your control tolerance cost more than the calibration events they deferred.</p>
<h2 id="what-the-historian-also-shows">What the historian also shows</h2>
<p>When you pull the calibration records from the historian, you are likely to see a second pattern: step changes in sensor output immediately after calibration. Some of these reflect real pre-calibration drift being corrected. Others reflect calibration error: a sensor re-zeroed with a small bubble on the membrane, or calibrated in a bucket of tap water at a different temperature than the basin. Frequent calibration events introduce more opportunities for calibration error, particularly in plants with high technician turnover.</p>
<p>The right interval is one that keeps drift within your control tolerance while keeping calibration frequency low enough that each calibration event can be done carefully. For most activated sludge applications with polarographic sensors, that is typically in the two-to-three-week range when fouling is measured and the drift curve is actually checked, not simply assumed to match the manufacturer default.</p>
<h2 id="the-underlying-condition">The underlying condition</h2>
<p>Whether the calibration schedule is two weeks or six, the aeration control loop operates on sensor output, not on the actual dissolved oxygen concentration. When the interval is appropriate and calibrations are executed correctly, the gap between what the sensor says and what the process holds is small. When the interval is too long, or when calibrations are missed due to staffing constraints or maintenance backlog, the control loop can drift significantly from its design intent.</p>
<p>The analysis that reveals this is in the historian. The drift records are already there: pre-calibration versus post-calibration readings, blower runtimes, DO trends. The pattern of drift accumulation and its effect on blower behavior is visible at cycle resolution, which typically means five-minute data over a 30-to-90-day window.</p>
<p>This is the same kind of pattern identified in the <a href="https://aquamesh.ai/blog/why-parallel-equipment-drifts.html">parallel-basin drift analysis</a> and expanded in <a href="https://aquamesh.ai/blog/what-25-days-of-historian-data-shows.html">what 25 days of historian data actually shows about a treatment plant</a>. The cooling tower equivalent -- conductivity sensor drift triggering premature blowdown -- is documented in <a href="https://aquamesh.ai/blog/cooling-tower-blowdown-conductivity-sensor-drift.html">cooling tower blowdown control: conductivity drift cost</a>. The DO calibration interval problem is not a different issue; it is the same measurement gap operating on a longer time scale. The same drift accumulation applies to pH electrodes in the same bioreactors, where junction fouling biases alkalinity dosing and masks early nitrification inhibition: <a href="/blog/ph-sensor-drift-wastewater-activated-sludge.html">pH sensor drift in activated sludge: the cost of 0.5 units</a>. Ammonium ISEs in plants running ABAC control accumulate the same reference junction fouling on similar timelines, with the added consequence that drift distorts the DO setpoint calculation upstream of the blower: <a href="/blog/ammonia-sensor-drift-wastewater-aeration-control.html">ammonia sensor drift in wastewater aeration control</a>.</p>
<hr />
<p>If you want to understand whether your DO calibration intervals are matched to your actual process fouling rates, the starting point is the calibration and historian data you already have. See what that kind of review looks like at an operating plant: <a href="https://aquamesh.ai/case-studies/water-reuse-plant.html">a membrane-bioreactor water reuse plant case study</a>. Or <a href="https://aquamesh.ai/#contact">reach out to schedule a working session with the team</a>.</p>
<p><em>Alex Towfigh is a founder at AquaMesh.</em></p>]]></content:encoded>
      <category>dissolved oxygen</category>
      <category>calibration</category>
      <category>activated sludge</category>
      <category>aeration control</category>
      <category>process control</category>
    </item>
    <item>
      <title>Cooling tower blowdown: conductivity control</title>
      <link>https://aquamesh.ai/blog/cooling-tower-blowdown-conductivity-sensor-drift.html</link>
      <guid isPermaLink="true">https://aquamesh.ai/blog/cooling-tower-blowdown-conductivity-sensor-drift.html</guid>
      <pubDate>Wed, 28 Oct 2026 09:00:00 GMT</pubDate>
      <description>Conductivity sensor drift triggers premature blowdown, wasting makeup water and chemicals. Here is what it costs and how to detect it in your data.</description>
      <content:encoded><![CDATA[<p>Conductivity blowdown control keeps a cooling tower's cycles of concentration at the designed value by measuring dissolved solids in the circulating water and opening the blowdown valve when they exceed the setpoint. The control loop has three parts that operators and engineers should understand and verify separately: the setpoint, the sensor accuracy, and the method for detecting when the sensor drifts.</p>
<p><b>How the control works:</b> A conductivity sensor in the basin or blowdown line reads dissolved solids continuously. A controller compares that reading to a programmed setpoint corresponding to the design cycles of concentration (COC) -- the ratio of dissolved solids in circulating water to incoming makeup. At six COC, the system extracts six evaporative passes from each gallon before discharging; at three COC, it discharges three times as often for the same cooling output. The DOE Federal Energy Management Program documents a 20 percent reduction in makeup water and a 50 percent reduction in blowdown when moving from three to six COC. That difference is controlled entirely by where the conductivity sensor's reading lands relative to actual basin concentration. (<a href="https://www.energy.gov/femp/best-management-practice-10-cooling-tower-management" rel="noopener">DOE FEMP Best Management Practice #10: Cooling Tower Management</a>)</p>
<p><b>How to detect a drifting sensor:</b></p>
<ol class="post-list"><li>Cross-check the in-line sensor against a calibrated portable meter at each maintenance visit. A divergence greater than 5 percent means the control loop should not be trusted until the sensor is cleaned and recalibrated.</li><li>Pull blowdown valve timestamps and conductivity readings from the historian for a rolling 60-day window. Rising blowdown frequency with stable makeup water quality is a drift signature, not a process change.</li><li>Compare blowdown frequency normalized to cooling load. A sensor fouled high appears as more blowdown events per unit of heat rejection over time, with no ambient or makeup-water explanation.</li><li>Check makeup water conductivity readings over the same period. If makeup quality is stable and blowdown frequency is rising, the sensor or controller setpoint has changed, not the incoming water.</li></ol>
<p><b>What happens when drift goes undetected:</b> A sensor reading high triggers blowdown when basin water is still within its design operating range, discharging water that had remaining evaporative capacity. A sensor reading low allows dissolved solids to concentrate past the setpoint, depositing scale on heat exchange surfaces and potentially pushing discharge chemistry past permit limits. Neither failure triggers an alarm: the control loop operates normally, and the cost accumulates in the historian before it appears anywhere else.</p>
<h2 id="why-cycles-of-concentration-are-the-performance-metric">Why cycles of concentration are the performance metric</h2>
<p>The conductivity setpoint corresponds directly to cycles of concentration: the ratio of dissolved solids in the circulating water compared to the makeup water. At three cycles of concentration, the circulating water is three times as concentrated as the incoming makeup. At six cycles, it is six times as concentrated, meaning the system has extracted six times the evaporative cooling work from each unit of water before discharging it.</p>
<p>The Department of Energy's Federal Energy Management Program, in its Best Management Practice for Cooling Tower Management, documents the consequence of that difference: increasing cycles from three to six reduces cooling tower makeup water consumption by 20 percent and cooling tower blowdown by 50 percent. (<a href="https://www.energy.gov/femp/best-management-practice-10-cooling-tower-management" rel="noopener">DOE FEMP Best Management Practice #10: Cooling Tower Management</a>)</p>
<p>Most cooling towers run at two to four cycles of concentration. The practical ceiling is set by water chemistry: calcium, magnesium, silica, and other dissolved minerals will eventually scale heat exchange surfaces or corrode metal components if concentration runs too high. Water treatment programs add scale inhibitors and corrosion inhibitors specifically to allow higher cycles of concentration without those chemistry problems. But there is a second constraint that receives less attention: the accuracy of the conductivity measurement that tells the system what concentration it is actually running at.</p>
<h2 id="what-sensor-drift-does-to-blowdown-frequency">What sensor drift does to blowdown frequency</h2>
<p>Conductivity sensors in cooling tower service face a corrosive environment. Basin water contains biological growth, suspended solids, scale-forming minerals, and treatment chemicals. Sensor fouling from biofilm and mineral deposits is the most common failure mode, and the most common direction of failure is a reading that drifts high: a deposit on the sensor surface creates an impedance that the controller interprets as elevated dissolved solids in the water.</p>
<p>A sensor reading high produces the same control response as actual high concentration: it triggers blowdown. If the setpoint is 1,500 microsiemens per centimeter and the sensor reads 1,600 when the actual basin conductivity is 1,200, the controller sees an out-of-spec condition and opens the valve. The water being discharged was at 1,200 microsiemens, well within the setpoint, and capable of running through additional evaporative cycles before requiring blowdown.</p>
<p>This is not a hypothetical edge case. Sensor fouling accumulates gradually, and in many plants the cleaning interval for conductivity sensors is set by a calendar schedule rather than by measurement-to-measurement agreement. A sensor cleaned monthly in a high-load summer period is operating for weeks with accumulating offset. During that interval, every blowdown event the offset triggers is water the system did not need to discharge.</p>
<p>The reverse failure also occurs: a sensor that reads low because of a calibration shift or a failed element delays blowdown past the intended setpoint. Scale begins depositing on heat transfer surfaces, thermal performance drops, and in severe cases chemical treatment effectiveness falls as the system runs beyond its design envelope. Scale deposits of even 1/16 inch on heat exchanger surfaces can reduce thermal efficiency by roughly 10 percent. (<a href="https://www.sciencedirect.com/science/article/abs/pii/S1359431116341849" rel="noopener">Rahmani et al., Applied Thermal Engineering, 2017, doi:10.1016/j.applthermaleng.2016.10.191</a>)</p>
<h2 id="quantifying-the-cost-of-over-blowdown">Quantifying the cost of over-blowdown</h2>
<p>The DOE documents a 50 percent reduction in blowdown when moving from three to six cycles of concentration, and that figure provides a basis for estimating what premature blowdown costs. The calculation depends on facility size, local water costs, and sewer discharge fees, but the structure is consistent:</p>
<p><b>At a facility with 1 million gallons per day of makeup water:</b></p>
<div class="table-wrap"><table class="cs-table"><thead><tr><th>Operating scenario</th><th>Approximate daily makeup water</th><th>Approximate daily blowdown</th></tr></thead><tbody><tr><td>6 cycles of concentration (design intent)</td><td>1,000,000 gallons</td><td>~200,000 gallons</td></tr><tr><td>3 cycles of concentration (sensor reads high)</td><td>1,200,000+ gallons</td><td>~400,000 gallons</td></tr><tr><td>Difference from premature blowdown</td><td>200,000+ gallons/day</td><td>~200,000 gallons/day</td></tr></tbody></table></div>
<p><em>Modelled example using DOE FEMP documented percentage differences; actual values depend on evaporation rate, local water chemistry, and the degree of sensor offset.</em></p>
<p>That makeup water difference carries three layers of cost. First, water purchase cost: industrial water rates vary by region, but an industrial facility paying in the range of $1 to $5 per thousand gallons for purchased water faces a material cost when hundreds of thousands of gallons per day are discharged unnecessarily. Second, sewer and wastewater discharge costs: many jurisdictions charge for cooling tower blowdown discharge separately from water purchase, and in areas with sewer surcharges for high-conductivity industrial discharge, this cost can approach or exceed the water purchase cost. Third, chemical treatment cost: every gallon of blowdown replaced with fresh makeup water requires treatment chemicals to protect the system from the new makeup water's chemistry.</p>
<p>For facilities that discharge cooling tower blowdown directly to surface waters, there is a fourth cost layer: regulatory exposure. Cooling tower blowdown discharged to waters of the United States requires authorization under the EPA's National Pollutant Discharge Elimination System (NPDES) program, and industrial discharge permits for many categories include limits on specific conductance or total dissolved solids as a proxy for concentrated minerals and treatment chemicals. A sensor reading low (allowing cycles of concentration to run above design) may produce a discharge that exceeds permit limits even when the control system appears to be operating normally, because the system registers a conductivity that is within setpoint while actual basin concentration has climbed past it. (<a href="https://www.ecfr.gov/current/title-40/chapter-I/subchapter-D/part-122" rel="noopener">40 CFR Part 122, EPA NPDES Permit Regulations</a>)</p>
<p>None of this is visible at the level of a single blowdown event. It becomes visible in trends: makeup water consumption trending upward over a season, chemical feed rates increasing without a corresponding change in makeup water quality, blowdown frequency in the historian rising faster than ambient temperature or load would predict.</p>
<h2 id="the-diagnostic-that-is-already-in-the-historian">The diagnostic that is already in the historian</h2>
<p>The same diagnostic approach that reveals <a href="/blog/do-sensor-calibration-interval-activated-sludge.html">DO sensor drift in aeration systems</a> applies to conductivity sensor drift in cooling tower control. The data is in the historian: blowdown valve timestamps, conductivity sensor readings, makeup water flow, and any makeup water quality monitoring already being logged.</p>
<p>The signature of a sensor reading high is blowdown events that cluster at a conductivity reading just above the setpoint, while makeup water conductivity readings stay constant and blowdown event frequency increases over time on a consistent load. The signature of a sensor reading low is the opposite: infrequent blowdown with conductivity readings that stay below the setpoint even as the tower accumulates load, followed eventually by scale chemistry indicators or pressure-drop increases in the heat exchangers.</p>
<p>This is a temporal version of the spatial drift problem described in a parallel-equipment analysis. In that case, two identical aeration basins showed diverging blower runtime and energy use that reflected sensor offset rather than actual process differences. Here, the same principle operates over time in a single system: the measurement feeding the control loop has an offset, the loop acts on the offset rather than the process state, and the cost accumulates in the historian before it appears anywhere else.</p>
<p>The full parallel-basin methodology is in <a href="/blog/why-parallel-equipment-drifts.html">why parallel equipment drifts and what it costs</a>. For data centers where cooling tower blowdown directly affects the WUE metric, the connection between sensor drift and reported efficiency is covered in <a href="/blog/data-center-wue-measurement-gap.html">data center WUE: what it measures and what it misses</a>.</p>
<h2 id="what-accurate-conductivity-measurement-enables">What accurate conductivity measurement enables</h2>
<p>When the conductivity sensor controlling blowdown is reading accurately and corrected between calibration intervals, the control system can run the tower at the cycles of concentration the water treatment program was designed for. The 20 to 50 percent water savings the DOE documents from moving from three to six cycles are achievable without chemistry changes or capital investment -- they require that the measurement driving the control loop reflects actual basin conditions.</p>
<p>For industrial facilities with large cooling loads -- power generation, petrochemical, food and beverage processing, large manufacturing plants -- the difference between running at three cycles and six cycles is not a marginal optimization. At the scale of a 2,000-ton cooling tower system, it is a question of hundreds of thousands of gallons per day.</p>
<p>The diagnostic questions are straightforward. Is the conductivity sensor reading the same at the start and end of a calibration interval? Does blowdown frequency track makeup water conductivity the way it should, or does it drift independently? Are blowdown events triggering at the expected conductivity setpoint, or consistently below it? The answers are in the historian. The constraint is having a way to pull the data at the resolution that shows blowdown-event-level behavior over a long enough window to separate sensor drift from process variation.</p>
<hr />
<p>If your cooling tower water use has trended upward or your blowdown frequency looks inconsistent with your system load, the control data is already in your historian.</p>
<p>See what a water system measurement review looks like at an operating industrial facility: <a href="https://aquamesh.ai/case-studies/water-reuse-plant.html">a membrane-bioreactor water reuse plant case study</a>. Or <a href="https://aquamesh.ai/#contact">reach out to schedule a working session with the team</a>.</p>
<p><em>Alex Towfigh is a founder at AquaMesh.</em></p>]]></content:encoded>
      <category>cooling tower</category>
      <category>blowdown</category>
      <category>conductivity</category>
      <category>water efficiency</category>
      <category>industrial water management</category>
    </item>
    <item>
      <title>The $1.5M sewer repair: what discharge monitoring missed</title>
      <link>https://aquamesh.ai/blog/water-online-sewer-repair-evidence.html</link>
      <guid isPermaLink="true">https://aquamesh.ai/blog/water-online-sewer-repair-evidence.html</guid>
      <pubDate>Mon, 26 Oct 2026 09:00:00 GMT</pubDate>
      <description>Guest column by Alex Towfigh in Water Online on a $1.5M restitution case and what upstream monitoring could have shown about the discharge.</description>
      <content:encoded><![CDATA[<hr />
<p>title: The $1.5M sewer repair: what discharge monitoring missed slug: water-online-sewer-repair-evidence date: 2026-10-26 section: In the press description: Guest column by Alex Towfigh in Water Online on a $1.5M restitution case and what upstream monitoring could have shown about the discharge. tags: industrial wastewater, pretreatment, discharge compliance, sewer surcharges, spectral sensing, Water Online</p>
<hr />
<p>Water Online published a guest column by Alex Towfigh, AquaMesh's founder and CEO, on 26 October 2026. The full piece is on the Water Online site: <a href="https://www.wateronline.com/doc/the-million-sewer-repair-and-the-water-quality-evidence-missing-upstream-0001" rel="noopener">The $1.5 Million Sewer Repair -- And The Water-Quality Evidence Missing Upstream</a>. What follows is an expanded summary with additional context on the pretreatment regulatory framework.</p>
<h2 id="the-argument-in-brief">The argument in brief</h2>
<p><b>The case.</b> A California wire manufacturer was ordered to pay $1.5 million in restitution after acidic wastewater damaged a public trunk sewer. EPA's 2010 announcement described 14 illegal discharges over four months of 2008, with pH as low as 1 against a permit minimum of 6. The column is careful about what this shows: the restitution covered cumulative damage, and a new analyzer alone would not necessarily have stopped the discharges. Its question is whether the plant could have recognized a repeated pattern early enough to investigate the treatment system behind it.</p>
<p><b>An endpoint is not an explanation.</b> A pH probe at the discharge point tells you whether neutralization worked. It does not tell you whether the cause was a transfer, a dosing fault, a bypass, a reagent shortage or maintenance. Better detection cannot replace working treatment equipment, but it can shorten the path from a suspect reading to its cause.</p>
<p><b>Two bills for the same discharge.</b> Lost product and treatment surcharges are separate ledgers, and the column warns against counting the same saving twice. It works through a dairy scenario using a published loss range of 1% to 3% of milk processed, and a COD example using San Diego's published January 2026 rate of $0.27 per pound: at 1 million gallons a day with 1,000 mg/L of added COD, about $2,252 a day. Both are sensitivity calculations, not measured losses or billing forecasts, and the column says so.</p>
<p><b>The blind spot between systems.</b> Read-only access to PLC, SCADA and historian data reconstructs what production was doing and when. A multiwavelength UV/visible sensor can show changes in the stream that pH, flow or conductivity miss, provided it is calibrated on site. A rinse that lines up with an optical change is a lead to check against a lab sample, not a finding.</p>
<p><b>The clock decides usefulness.</b> Detection only helps if it arrives in time to act. EPA event-detection research found that many unexpected water-quality changes came from operational changes or sensor problems rather than contamination. So the measures that matter are lead time, confirmed causes, false alerts per shift and cost, not anomaly accuracy on its own.</p>
<h2 id="the-pretreatment-gap-why-violations-repeat">The pretreatment gap: why violations repeat</h2>
<p>The EPA pretreatment program, codified at <a href="https://www.ecfr.gov/current/title-40/chapter-I/subchapter-N/part-403" rel="noopener">40 CFR Part 403</a>, requires significant industrial users discharging to publicly owned treatment works (POTWs) to monitor their effluent and submit self-monitoring reports against local and categorical limits. The California case is notable for documenting 14 violations over four months rather than a single incident.</p>
<p>That pattern matters. A one-time equipment failure is an operations problem. Repeated violations over multiple months suggest the monitoring and investigation loop was too slow to interrupt the cycle: a suspect reading, an investigation that finds no root cause, another event. The column's core argument is that an endpoint measurement at the discharge cannot tell you which upstream condition caused it. The same gap operates at the municipal discharge level: <a href="/blog/nitrogen-permit-compliance-sampling-gap.html">NPDES nitrogen permit compliance: the DMR sampling gap</a> covers how periodic DMR sampling misses what happens between sampling dates at the POTW receiving the industrial discharge.</p>
<p>Under the pretreatment program, a significant industrial user found in violation may face formal Notices of Violation, civil penalties under <a href="https://www.epa.gov/enforcement/water-enforcement" rel="noopener">CWA Section 309</a>, and in cases involving physical POTW damage, restitution claims from the municipal utility. The $1.5 million figure in this case reflected sewer repair costs, not standard permit penalties.</p>
<p><a href="/blog/spectral-change-analysis-vs-calibrated-parameters.html">Spectral sensing can reveal discharge chemistry that endpoint pH alone misses</a>. But identifying that a problem occurred is different from tracing its cause upstream. That is where process historian data plays a role. <a href="/blog/what-25-days-of-historian-data-shows.html">When 25 days of process historian data were analyzed at one industrial facility</a>, unexpected deviations mapped to specific shifts and equipment states rather than random failures, providing the upstream context the column argues is missing from most monitoring architectures.</p>
<h2 id="three-questions-for-the-next-operations-meeting">Three questions for the next operations meeting</h2>
<p>For each recurring discharge, the column asks:</p>
<ol class="post-list"><li>What is the earliest validated water-quality measurement?</li><li>Which upstream records would explain it?</li><li>What authorized operating change would prevent it from happening again?</li></ol>
<p>It closes with a test of its own: find the single missing measurement that would most improve the conclusion. If there is not one, more sensors may not be the right investment. And AquaMesh's value has to be shown the same way, through lower confirmed losses, earlier diagnosis or less investigation effort.</p>
<p>The column's framing applies broadly to any facility managing discharges under a POTW pretreatment permit. The practical question is whether the current monitoring architecture would allow operations staff to trace the source of a problem within the same shift it occurs, rather than after the next scheduled compliance sample. If the answer is no, that is the gap worth measuring first. What continuous online sensors add, and what they cannot replace, in a facility's monitoring architecture is covered in <a href="/blog/online-sensors-vs-lab-sampling-wastewater-discharge.html">online sensors vs. lab sampling for discharge compliance</a>.</p>
<p><a href="https://www.wateronline.com/doc/the-million-sewer-repair-and-the-water-quality-evidence-missing-upstream-0001" rel="noopener">Read the full column in Water Online</a>. For facilities working through a similar monitoring gap, <a href="https://aquamesh.ai/#contact">request a conversation with the AquaMesh team</a>.</p>
<p><em>Alex Towfigh is a founder at AquaMesh.</em></p>]]></content:encoded>
      <category>industrial wastewater</category>
      <category>pretreatment</category>
      <category>discharge compliance</category>
      <category>sewer surcharges</category>
      <category>spectral sensing</category>
      <category>Water Online</category>
    </item>
    <item>
      <title>Intermittent aeration: why DO sensor precision matters</title>
      <link>https://aquamesh.ai/blog/intermittent-aeration-nitrogen-removal-do-control.html</link>
      <guid isPermaLink="true">https://aquamesh.ai/blog/intermittent-aeration-nitrogen-removal-do-control.html</guid>
      <pubDate>Fri, 23 Oct 2026 09:00:00 GMT</pubDate>
      <description>DO sensor precision controls whether intermittent aeration for nitrogen removal saves energy or misses permit limits. Here is what happens when it drifts.</description>
      <content:encoded><![CDATA[<p>Biological nitrogen removal is one of the more demanding process control challenges in activated sludge treatment. Plants with tight effluent nitrogen limits need to run two competing biological processes: nitrification under aerobic conditions and denitrification under anoxic ones, efficiently, in the same biological system. For plants using intermittent aeration to accomplish this in a single-basin configuration, the precision of dissolved oxygen control during aeration cycles is not an instrumentation detail. It is the variable that determines whether the strategy works.</p>
<h2 id="what-intermittent-aeration-is-trying-to-do">What intermittent aeration is trying to do</h2>
<p>Conventional biological nitrogen removal stages aerobic and anoxic zones in sequence: ammonia is nitrified to nitrate in an aerated zone, then denitrified to nitrogen gas in a downstream anoxic zone. The system is reliable but capital-intensive and requires careful hydraulic design.</p>
<p>Intermittent aeration takes a different path: the same basin alternates between aerobic and anoxic phases by cycling blowers on and off. During the aerated phase, nitrifiers convert ammonia to nitrate. During the unaerated phase, denitrifiers reduce nitrate to nitrogen gas, using carbon already present in the mixed liquor. Some organisms can carry out both reactions near-simultaneously at low dissolved oxygen concentrations, a phenomenon called simultaneous nitrification-denitrification (SND). The EPA Nutrient Control Design Manual identifies low-DO operation, typically below 1.0 mg/L, as a condition that promotes SND in conventional activated sludge by creating anoxic microsites within biological flocs. (<a href="https://www.epa.gov/sites/default/files/2019-08/documents/nutrient-control-design-manual.pdf" rel="noopener">EPA, Nutrient Control Design Manual, 2010, Section 5</a>)</p>
<p>A 2022 review in Chemical Engineering Journal analyzed over 50 intermittent aeration studies and found that precise DO control during aeration phases was the most consistent predictor of nitrogen removal efficiency, ahead of cycle length or influent carbon-to-nitrogen ratio. (<a href="https://www.sciencedirect.com/science/article/abs/pii/S1385894721047604" rel="noopener">Miao et al., 2022, Chemical Engineering Journal, doi:10.1016/j.cej.2021.133516</a>)</p>
<p>The practical appeal is real: intermittent aeration can achieve substantial nitrogen removal without a separate anoxic zone, with lower aeration energy than continuous-aeration systems, and with flexible response to variable influent loads. These advantages depend on cycle timing, and cycle timing depends on DO measurement.</p>
<h2 id="where-do-control-precision-comes-in">Where DO control precision comes in</h2>
<p>The cycle timing in intermittent aeration is almost always governed by a dissolved oxygen setpoint rather than a fixed timer. Aeration turns on when DO falls below a lower bound; it turns off when DO rises above an upper bound. The accuracy of that DO band controls the effectiveness of the strategy:</p>
<p><b>Sensor reading low (common fouling failure):</b> A sensor that reads 0.4 mg/L when actual DO is 1.8 mg/L keeps blowers running long past the point where nitrification has completed. The aerobic phase extends beyond the design intent. The anoxic fraction shrinks. Denitrification suffers, nitrogen removal drops, and the plant uses more electricity than the intermittent strategy was supposed to require.</p>
<p><b>Sensor reading high:</b> Aeration shuts off early. The basin does not complete nitrification in one cycle. Ammonia carries over. In a plant with a permit nitrogen limit, this is an effluent compliance risk, not just an energy problem.</p>
<p><b>Fouling that accumulates between cycles:</b> Mixed liquor accumulates on membrane or optical sensor surfaces during the unaerated phase. When aeration restarts, the sensor enters the aerobic phase with an offset that compounds over successive cycles if cleaning intervals are not calibrated to match actual fouling rates. What those intervals should be for activated sludge service is covered in <a href="/blog/do-sensor-calibration-interval-activated-sludge.html">DO sensor calibration intervals in activated sludge</a>.</p>
<p>The cumulative effect is that the DO setpoints operators specify on the control screen may not correspond to the dissolved oxygen concentration the organisms experience. The blower control loop optimizes against a sensor output, not against the process state.</p>
<h2 id="what-this-looks-like-in-practice">What this looks like in practice</h2>
<p>The same phenomenon described in the parallel-basin aeration analysis applies here in a temporal rather than spatial dimension. In that case, two basins running simultaneously showed a DO divergence that reflected sensor drift rather than process differences: one blower was pinned at maximum output, consuming over $3,000 per year more than its twin, for the full 25-day record. The spatial version was visible by comparing basins side by side. The parallel equipment drift analysis is in <a href="https://aquamesh.ai/blog/why-parallel-equipment-drifts.html">why parallel equipment drifts apart</a>, and the full historian data methodology is in <a href="https://aquamesh.ai/blog/what-25-days-of-historian-data-shows.html">what 25 days of historian data actually shows about a treatment plant</a>.</p>
<p>The temporal version in intermittent aeration requires a different analysis: charting DO sensor output against blower on/off timestamps at five-minute or finer resolution, and comparing aerobic phase duration over time. A sensor reading accurately produces aerobic phases of consistent duration for a given influent load. A sensor that is drifting produces phases that systematically lengthen (reading low, blowers run long) or shorten (reading high, blowers cut off early) in ways that do not track influent load changes.</p>
<p>Both versions share the same root cause: the measurement feeding the control loop has an offset that the loop cannot self-detect, because there is no independent reference running alongside it.</p>
<h2 id="the-energy-calculation-for-intermittent-aeration">The energy calculation for intermittent aeration</h2>
<p>The aeration energy baseline for secondary treatment, around 1,500 kWh per million gallons for activated sludge, assumes reasonably continuous aeration. Intermittent aeration in a nitrogen-removal configuration typically aims to reduce aeration runtime to 50 to 70 percent of total time, depending on the carbon-to-nitrogen ratio of the influent and the cycle design. That would imply aeration energy in the range of 750 to 1,050 kWh per million gallons when timing is accurate.</p>
<p>At a 10 MGD plant and the current EIA industrial rate of $0.0977 per kWh (<a href="https://www.eia.gov/electricity/monthly/epm_table_grapher.php?t=epmt_5_6_a" rel="noopener">EIA Electric Power Monthly, September 2026</a>), the difference between accurate intermittent aeration (approximately 900 kWh/MG as a midpoint) and sensor-extended aeration (1,500 kWh/MG because the sensor is holding blowers on too long) is approximately $215,000 per year at that scale.</p>
<p><b>Modelled example for a 10 MGD plant with intermittent aeration:</b></p>
<div class="table-wrap"><table class="cs-table"><thead><tr><th></th><th>Accurate DO control</th><th>DO sensor drifting low</th></tr></thead><tbody><tr><td>Aeration runtime share</td><td>60%</td><td>100% (effectively continuous)</td></tr><tr><td>Estimated kWh/MG</td><td>900</td><td>1,500</td></tr><tr><td>Annual energy cost at $0.098/kWh</td><td>$193,100</td><td>$321,900</td></tr><tr><td>Annual cost difference</td><td></td><td>$128,800</td></tr></tbody></table></div>
<p>This is a modelled example using published benchmarks; actual values depend on plant-specific loading, cycle design, and the degree of sensor offset. The direction is consistent: sensor drift in an intermittent aeration system does not only affect nitrogen removal performance. It erodes the energy savings that justified the intermittent strategy in the first place. VFD-controlled blowers in continuous aeration systems show the same erosion pattern through a different mechanism: <a href="/blog/wastewater-blower-vfd-energy-savings-erode.html">why wastewater blower VFD energy savings erode</a>.</p>
<h2 id="what-accurate-do-measurement-enables">What accurate DO measurement enables</h2>
<p>When DO measurement in an intermittent aeration system is continuous, accurate, and corrected against drift between calibration intervals, the control loop can do what it was designed for: maintain the aerobic-to-anoxic balance that both nitrification and denitrification require.</p>
<p>For plants approaching permit limits on total nitrogen, the margin for process error is narrow. The compliance exposure those limits create -- specifically what the scheduled DMR sampling program can and cannot see on unmonitored days -- is covered in <a href="/blog/nitrogen-permit-compliance-sampling-gap.html">NPDES nitrogen permit compliance: the DMR sampling gap</a>. A sensor that introduces a systematic offset into the timing of aeration cycles is a measurement problem that manifests as an effluent compliance risk. The same monitoring gap -- scheduled samples capturing a narrow window of actual operating conditions -- applies to industrial dischargers under the National Pretreatment Program: <a href="/blog/industrial-pretreatment-self-monitoring-gaps.html">industrial pretreatment: the self-monitoring blind spot</a>. The issue is usually visible in the historian data before it appears in the permit reporting: aerobic phase durations extending over time, DO baseline at cycle start gradually increasing, nitrogen removal efficiency declining in ways that do not correlate with influent load.</p>
<p>The diagnostic questions are the same ones that apply to continuous aeration systems. Are the DO sensors reading accurately between calibrations? Does blower runtime at a given influent load match what the setpoints predict? Are cycle times stable, or are they drifting? The methodology for that analysis is in <a href="https://aquamesh.ai/blog/aeration-energy-cost-wastewater-plants.html">aeration energy at wastewater plants: the cost, the waste, and the measurement gap</a>.</p>
<p>For plants running ammonia-based aeration control (ABAC), where an in-process ammonium analyzer trims blower output based on effluent ammonium concentration, that control loop is only as reliable as the ammonia sensor feeding it: <a href="/blog/ammonia-sensor-drift-wastewater-aeration-control.html">ammonia sensor drift in wastewater aeration control</a>.</p>
<p>In the same biological process, pH electrodes face the same drift accumulation. A biased pH reading keeps alkalinity dosing off-target and can mask developing nitrification inhibition before the effluent shows it: <a href="/blog/ph-sensor-drift-wastewater-activated-sludge.html">pH sensor drift in activated sludge: the cost of 0.5 units</a>.</p>
<p>None of this analysis requires new hardware. The data is in the historian. The constraint is having a way to extract it at the resolution that shows cycle-level behavior, over a long enough window, against a reference that was not subject to the same drift.</p>
<hr />
<p>If your plant uses intermittent aeration for nitrogen removal and you want to understand whether your DO control is performing as designed, the place to start is the cycle data already in your historian.</p>
<p>See what that review looked like at an operating plant: <a href="https://aquamesh.ai/case-studies/water-reuse-plant.html">a membrane-bioreactor water reuse plant case study</a>. Or <a href="https://aquamesh.ai/#contact">reach out to schedule a working session with the team</a>.</p>
<p><em>Alex Towfigh is a founder at AquaMesh.</em></p>]]></content:encoded>
      <category>nitrogen removal</category>
      <category>intermittent aeration</category>
      <category>dissolved oxygen</category>
      <category>SND</category>
      <category>wastewater treatment</category>
    </item>
    <item>
      <title>Data center WUE: what it measures and misses</title>
      <link>https://aquamesh.ai/blog/data-center-wue-measurement-gap.html</link>
      <guid isPermaLink="true">https://aquamesh.ai/blog/data-center-wue-measurement-gap.html</guid>
      <pubDate>Thu, 22 Oct 2026 09:00:00 GMT</pubDate>
      <description>WUE measures liters per kWh of IT load. It does not show where water goes, verify cooling controls, or prove tower savings for reporting.</description>
      <content:encoded><![CDATA[<p>Water Usage Effectiveness (WUE) is a data center water metric introduced by <a href="https://www.thegreengrid.org/resources/library-and-tools" rel="noopener">The Green Grid</a>. The formula: liters of water consumed at the facility divided by kilowatt-hours delivered to IT equipment, where lower is better and zero is the theoretical ideal.</p>
<p>What WUE does not answer:</p>
<ul class="post-list"><li>Where the water is going (evaporation versus blowdown)</li><li>Whether the sensors driving cooling tower control are accurate</li><li>Whether blowdown volume can be reduced on demand under reporting or drought rules</li><li>What specific operating changes move the number versus what is fixed by physics</li></ul>
<p>Each of those gaps affects whether a WUE target in a sustainability report is operationally achievable.</p>
<h2 id="what-wue-actually-measures">What WUE actually measures</h2>
<p>The metric was introduced by The Green Grid as a companion to PUE (Power Usage Effectiveness). Like PUE, it is a ratio: lower is better and zero is the theoretical ideal. A site with a WUE of 2.0 consumes two liters of water for every kilowatt-hour its servers use.</p>
<p>Research on actual facility performance puts the range between 1 and 9 liters per kilowatt-hour depending on cooling approach, climate zone, and facility age, with a central tendency around 1.5 to 2 L/kWh for older evaporative cooling installations. (<a href="https://www.nature.com/articles/s41545-021-00101-w" rel="noopener">Mytton, D. "Data centre water consumption." npj Clean Water 4, 52 (2021)</a>) Newer hyperscale builds using advanced chemistry management or closed-loop cooling report considerably lower values.</p>
<p>WUE is typically calculated annually. Some operators report quarterly. It is a useful benchmark for comparing facilities against each other and tracking long-run progress. It is a poor guide for what to do on Tuesday.</p>
<h2 id="where-the-water-actually-goes">Where the water actually goes</h2>
<p>For facilities using evaporative cooling towers, which covers the majority of installed capacity, the water balance has two main terms: evaporation and blowdown.</p>
<p>Evaporation is the productive part. The cooling process works by letting water absorb heat and leave the tower as vapor. Roughly 70 to 80 percent of the water that enters the tower in this way is not recoverable; it has done its job and left the site. (<a href="https://www.nature.com/articles/s41545-021-00101-w" rel="noopener">Mytton 2021</a>) This fraction tracks the heat load on the facility and varies with ambient wet-bulb temperature. A site cannot significantly reduce evaporative losses without changing the physics of how it cools its equipment.</p>
<p>Blowdown is different. As water circulates through the tower and a portion evaporates, the dissolved minerals left behind concentrate. Calcium, silica, chlorides, and biological material accumulate. If concentration rises too high, scale deposits on heat exchange surfaces and corrosion accelerates. To prevent this, some fraction of the circulating water is periodically discharged and replaced with fresh makeup water. That discharge is blowdown.</p>
<p>The ratio of dissolved mineral concentration in the circulating water to the concentration in the makeup supply is called the cycles of concentration. At three cycles, for every gallon that evaporates, roughly half a gallon is discharged as blowdown, and one and a half gallons are drawn in as makeup. At six cycles, blowdown drops to roughly one fifth of the evaporation volume. At ten cycles, it drops to one ninth.</p>
<p><b>Modelled example:</b> For a 10-megawatt IT load facility with a moderate WUE of 2.0 L/kWh operating at 85 percent utilization, annual water consumption is roughly 149 million liters. If evaporation represents 75 percent of that total, the remaining 37 million liters per year is blowdown. Moving from 3 to 6 cycles of concentration reduces that blowdown component by approximately 60 percent, or 22 million liters per year. This is a modelled estimate based on the ratios above; actual results depend on makeup water chemistry, heat load profile, and seasonal variation.</p>
<h2 id="the-measurement-problem">The measurement problem</h2>
<p>The control system that manages blowdown uses a conductivity sensor as its primary input. Conductivity in the circulating water rises as dissolved minerals concentrate, and the controller uses this as a proxy for cycles of concentration. When conductivity reaches a set threshold, the blowdown valve opens and makeup water flows in.</p>
<p>This means the conductivity sensor, sitting in a basin or on a sidestream line, is the direct operational driver of water efficiency performance. A sensor that reads accurately and holds its calibration enables tight control at higher cycles. A sensor that has fouled, drifted, or been calibrated infrequently has the opposite effect.</p>
<p>A sensor reading low when concentration is actually high causes the controller to keep the blowdown valve closed longer than intended. Dissolved solids accumulate beyond design limits. Scale risk increases and the operator may not know it until a heat exchanger inspection. A sensor reading high when concentration is actually normal causes the controller to open the blowdown valve prematurely and repeatedly, discharging good water and drawing in fresh makeup water unnecessarily. The facility's WUE is higher than it needs to be, and the cause does not appear in the annual report.</p>
<p>Neither failure mode triggers an alarm in a standard building management system. The control loop is running. The conductivity reading is within the operating range. The controller is doing exactly what it was programmed to do with the signal it receives.</p>
<p>This is the same silent failure pattern that affects dissolved oxygen control in wastewater aeration basins, where a fouled sensor drives blowers toward wasted energy rather than toward needed oxygen. The structure of the problem is identical: a lagging indicator (WUE for water, energy per gallon treated for wastewater) does not point to the sensor drift that is the actual cause. The gap shows up only when you look at the control inputs directly, not at the outcomes they produce. The parallel is covered in more detail in the analysis of <a href="/blog/why-parallel-equipment-drifts.html">why parallel equipment drifts apart</a>. It also applies to intermittent aeration cycles for nitrogen removal, where DO sensor drift affects both effluent compliance and the energy savings the intermittent strategy was meant to deliver: <a href="/blog/intermittent-aeration-nitrogen-removal-do-control.html">intermittent aeration for nitrogen removal: what DO sensor precision actually controls</a>.</p>
<p>For a detailed breakdown of what conductivity sensor drift costs in cooling tower water volume and chemical spend, including a worked example using DOE benchmark data, see <a href="/blog/cooling-tower-blowdown-conductivity-sensor-drift.html">cooling tower blowdown control: what conductivity sensor drift costs in water and chemicals</a>.</p>
<h2 id="proving-cooling-tower-water-savings-for-reporting">Proving cooling tower water savings for reporting</h2>
<p>A WUE target in a sustainability disclosure or a California Water Scarcity Plan is defensible only if three things are in place: a metered baseline, an auditable record of the variables that drive consumption, and a method to demonstrate the change was real rather than a calculation artifact.</p>
<p>For cooling towers, that means measuring at the points where water actually enters and leaves the system.</p>
<p><b>Metering checklist</b></p>
<ul class="post-list"><li><b>Makeup water flow</b>: A flow meter on the makeup supply line, totalizing by day. This is the primary input: it is what the utility invoices and what regulators count as consumption. Without a direct makeup meter, total consumption is estimated from conductivity readings and design assumptions, and that estimate cannot catch sensor-driven overconsumption.</li><li><b>Blowdown volume</b>: A flow meter or timed valve record on the blowdown line. Blowdown is the discretionary water loss, the portion a facility can actually reduce through chemistry management. Without a direct measurement, blowdown is typically back-calculated from conductivity set points, which creates circular logic if the conductivity sensor is the variable under audit.</li><li><b>Basin conductivity with historian logging</b>: A continuous conductivity reading linked to the historian, not just a set-point indicator. The logged record shows what cycles of concentration the facility actually operated at over time, versus what the design target was. This is the evidence base for any claim that COC improved.</li><li><b>Conductivity sensor verification schedule</b>: A periodic grab-sample check using an independent meter compared against the installed sensor. The <a href="https://www.energy.gov/eere/femp/best-management-practice-10-cooling-tower-management" rel="noopener">DOE Federal Energy Management Program Best Management Practice 10</a> identifies conductivity monitoring and cycles-of-concentration tracking as core elements of water efficiency for cooling towers. A documented record of sensor verification during both the baseline and the improvement period is the evidence that the signal driving blowdown control was accurate.</li></ul>
<p><b>Establishing an auditable baseline</b></p>
<p>A baseline for a WUE improvement claim needs at least 12 months of data before the change, covering seasonal variation in wet-bulb temperature and heat load. Cooling water consumption tracks ambient conditions, and a baseline drawn only from summer months overstates annual average performance.</p>
<p>From that record, calculate:</p>
<ol class="post-list"><li>Average cycles of concentration by month</li><li>Makeup volume relative to evaporative load (calculable from heat load data and ambient conditions)</li><li>Blowdown as a percentage of total makeup, by month</li><li>Unexplained blowdown events: volume spikes not correlated with changes in heat load</li></ol>
<p>Unexplained events, whether from conductivity sensor drift, a valve held open during maintenance, or a chemical feed miscalculation, show up as isolated increases in blowdown volume not matched by heat load increases. In a 25-day historian review at an industrial process facility, three anomalies of this type were identified, adding an estimated 12 to 18 percent to water consumption for that period. (<a href="/blog/what-25-days-of-historian-data-shows.html">What a 25-day historian audit finds at a treatment plant</a>) The same review methodology applies directly to cooling tower makeup and blowdown logs.</p>
<p><b>When direct blowdown metering is not installed</b></p>
<p>Many operating data centers lack a flow meter on the blowdown line. Blowdown volume is inferred from the conductivity set point and the makeup meter, which means the calculated efficiency number cannot detect the cases where the sensor controlling blowdown is drifting. A dedicated blowdown meter converts an estimated variable into a verified one. For a site preparing a California Water Scarcity Plan, that difference is material: a commitment to reduce blowdown by a specific percentage is auditable only if blowdown is measured independently of the control signal.</p>
<h2 id="the-california-compliance-context">The California compliance context</h2>
<p>California's AB 2619, signed in September 2026, requires data center operators to report total and direct annual water use by source at every business license renewal, under penalty of perjury. AB 2469 requires a Water Scarcity Plan by 2028 for facilities requiring construction permits. (<a href="/blog/california-data-center-water-reporting.html">California data center water reporting requirements</a>)</p>
<p>A Water Scarcity Plan is a commitment to reduce water consumption under specified drought conditions, including staged curtailment and targeted reductions by use type. The commitment is credible only if the site can demonstrate that it knows where its water goes and can reduce specific uses on demand.</p>
<p>A site operating at three cycles of concentration because its conductivity sensors drift between maintenance intervals cannot reliably commit to operating at six cycles under drought conditions. The chemistry that makes six cycles safe to operate requires measurement quality that makes four and a half cycles the actual operational average. The gap between the plan and the operational reality is a sensor reliability problem, not a strategy problem.</p>
<h2 id="three-questions-worth-checking-before-setting-a-wue-target">Three questions worth checking before setting a WUE target</h2>
<p>Before writing a WUE improvement target into a sustainability report or a Water Scarcity Plan, three questions are worth answering from existing historian data.</p>
<p><b>What are your actual cycles of concentration, by tower?</b> This is calculable from makeup flow, blowdown flow, and conductivity readings. If the number is consistently below design, the control logic or the sensor driving it is worth examining before adjusting targets.</p>
<p><b>How does blowdown volume correlate with wet-bulb temperature?</b> Evaporation tracks ambient conditions predictably. Blowdown should not spike independently of heat load. Unexplained blowdown events in the historian are almost always sensor artifacts or valve events, not chemistry responses.</p>
<p><b>What share of total water use does blowdown represent?</b> If blowdown is running above 30 percent of total consumption on a facility that should be operating at moderate cycles, the conductivity control system is not performing at design. This shows up in 12 months of metered data and does not require a new study to identify.</p>
<p>None of this analysis requires capital investment. The data exists in most historian systems. The constraint is usually having a way to look at it with enough resolution and the right reference points to distinguish signal from noise.</p>
<hr />
<p>If your site is preparing for California's 2027 renewal requirements or developing a scarcity plan for 2028, the water balance question is worth addressing before the commitment is made, not after. AquaMesh can review the historian data that is already being collected and identify where the measurement gaps are.</p>
<p>See how this kind of analysis surfaced a $3,316 blower anomaly at a wastewater treatment plant: <a href="/blog/what-25-days-of-historian-data-shows.html">what 25 days of historian data actually shows</a>. Or <a href="https://aquamesh.ai/#contact">reach out to schedule a working session</a>.</p>
<p><em>Alex Towfigh is a founder at AquaMesh.</em></p>]]></content:encoded>
      <category>data centers</category>
      <category>WUE</category>
      <category>water efficiency</category>
      <category>cooling tower</category>
      <category>water monitoring</category>
      <category>water usage effectiveness</category>
    </item>
    <item>
      <title>Wastewater blower energy: cost and control</title>
      <link>https://aquamesh.ai/blog/aeration-energy-cost-wastewater-plants.html</link>
      <guid isPermaLink="true">https://aquamesh.ai/blog/aeration-energy-cost-wastewater-plants.html</guid>
      <pubDate>Wed, 21 Oct 2026 09:00:00 GMT</pubDate>
      <description>Aeration is 50 to 75% of a wastewater plant's electricity bill. What it costs per MGD, why savings erode, and how to cut blower energy without new equipment.</description>
      <content:encoded><![CDATA[<p>Aeration is the single largest line on the electricity bill for most activated sludge wastewater plants. Aeration systems typically represent 50 to 75 percent of total electricity consumption at activated sludge facilities. The logical response is to install dissolved oxygen sensors, add variable frequency drives to blowers, and let a control loop do the rest. Many plants have done exactly that. Not all of them kept the savings.</p>
<p>Here is what the cost actually looks like, why the standard fix often falls short, and how to recover blower energy without adding hardware.</p>
<h2 id="how-large-the-prize-is">How large the prize is</h2>
<p>The U.S. Energy Information Administration reported the national average industrial electricity rate at $0.0977 per kWh in July 2026, up from $0.0933 the prior year. (<a href="https://www.eia.gov/electricity/monthly/epm_table_grapher.php?t=epmt_5_6_a" rel="noopener">EIA Electric Power Monthly, September 2026</a>)</p>
<p>Secondary treatment activated sludge plants typically use between 1,200 and 1,800 kWh per million gallons treated. The EPA's water and wastewater energy management documentation cites a common benchmark near 1,500 kWh per million gallons for secondary treatment. (<a href="https://www.epa.gov/sites/default/files/2016-12/documents/guide_energymanagement.pdf" rel="noopener">U.S. EPA, Ensuring a Sustainable Future: An Energy Management Guidebook for Wastewater and Water Utilities</a>)</p>
<p><b>Modelled example for a 10 MGD secondary treatment plant:</b></p>
<div class="table-wrap"><table class="cs-table"><thead><tr><th></th><th></th></tr></thead><tbody><tr><td>Annual flow treated</td><td>3,650 million gallons</td></tr><tr><td>Electricity at 1,500 kWh/MG</td><td>5,475,000 kWh</td></tr><tr><td>Aeration share at 60%</td><td>3,285,000 kWh</td></tr><tr><td>At $0.098/kWh industrial rate</td><td>$321,900/year</td></tr></tbody></table></div>
<p>This is a modelled example based on published benchmarks. Actual costs vary with load patterns, treatment configuration, equipment age, and the utility tariff in effect. The kilowatt-hours are the right order of magnitude; the dollar figure resolves once you pull an actual utility bill.</p>
<p>At this scale, a 20 percent reduction in blower energy is worth roughly $64,000 per year. The Florida Water and Pollution Control Operators Association published analysis in April 2026 showing that basic demand-based DO control typically reduces blower energy by 10 to 25 percent, and that PLC-optimized control can reach 20 to 40 percent. (<a href="https://www.fwpcoa.org/content.aspx?page_id=5&amp;amp;club_id=859275&amp;amp;item_id=134074" rel="noopener">FWPCOA, "Demand-Based DO Control And Energy Optimization," April 2026</a>)</p>
<p>Those are compelling numbers. The problem is that they describe what the system saves when it is working as designed. The gap between the theoretical savings and what a plant actually captures and keeps is usually a measurement problem.</p>
<h2 id="where-the-savings-go">Where the savings go</h2>
<p>A DO control loop has three components: a sensor reporting the current dissolved oxygen level, a controller comparing that reading to a setpoint, and a variable frequency drive adjusting blower output. The logic is straightforward.</p>
<p>The sensor is the weak link.</p>
<p>Electrochemical membranes foul and require regular replacement. Optical sensors are more stable but still accumulate biofilm in mixed liquor. A sensor that reads 1.8 mg/L when the actual basin concentration is 3.2 mg/L signals the controller that the basin is under-oxygenated. The controller responds by running blowers harder. The plant pays for electricity that achieves nothing useful, and nothing in the system indicates anything is wrong.</p>
<p>The inverse failure is less common but more consequential: a sensor fouled in a way that reads artificially high causes the controller to back off blowers when DO is genuinely borderline, with the process instability that follows.</p>
<p>Both failure modes are silent. The control loop is running. No alarms fire. The historian records the sensor output without knowing it has drifted. Operators checking the dashboard see normal-looking DO values.</p>
<p>This is why many plants install DO control hardware and see strong early results followed by gradual erosion. The hardware is still there. The control loop is still active. But the measurement feeding it has drifted, and the loop is now optimizing against a wrong input. The broader picture of how VFD blower savings erode, including mechanical wear and pressure creep alongside measurement drift, is in <a href="/blog/wastewater-blower-vfd-energy-savings-erode.html">wastewater blower VFD: why energy savings erode</a>.</p>
<h2 id="the-parallel-basin-case">The parallel basin case</h2>
<p>Plants with multiple aeration basins running in parallel face a compounding version of this problem. Sensors in different basins drift at different rates depending on local solids loading, position relative to diffusers, and how consistently each one gets cleaned. The basins may share a common air header, but if their DO sensors are reading different offsets, the blower outputs serving them will diverge in ways that reflect sensor drift rather than actual process differences.</p>
<p>We covered the mechanics and cost of this kind of hidden divergence in <a href="https://aquamesh.ai/blog/why-parallel-equipment-drifts.html">why parallel equipment drifts apart</a>. The direct energy cost of an over-aerated basin shows up in blower runtime. The downstream cost in nitrogen removal performance, when an anoxic zone receives excess dissolved oxygen, is harder to price but real.</p>
<h2 id="what-the-historian-data-showed">What the historian data showed</h2>
<p>When we analyzed 25 days of historian data from an operating plant earlier this year, one of the first findings was a DO-driven blower anomaly: one basin was running at more than three times the DO level of its parallel twin, with its blower command pinned at maximum output, for the full duration of the record.</p>
<p>Both basins drew on the same shared air header. The control loop was functioning. The deviation was visible only by comparing the two basins side by side at hourly resolution. A daily manual grab from either basin, taken at a representative time, could easily have missed it.</p>
<p>The finding was worth $3,316 per year in direct blower power, plus the process risk from carrying excess oxygen into a downstream anoxic zone. It was already in the data. It just had not been looked at in this way. The full analysis is in <a href="https://aquamesh.ai/blog/what-25-days-of-historian-data-shows.html">what 25 days of historian data actually shows about a treatment plant</a>.</p>
<h2 id="what-accurate-measurement-changes">What accurate measurement changes</h2>
<p>When DO measurement is continuous, accurate, and does not accumulate drift between calibration cycles, the control loop has what it was designed to work from. Blower output tracks actual oxygen demand rather than sensor artifacts. Setpoints can be tightened because operators are not building in a buffer to account for measurement uncertainty.</p>
<p>For plants running <a href="https://aquamesh.ai/blog/intermittent-aeration-nitrogen-removal-do-control.html">intermittent aeration for simultaneous nitrification/denitrification</a>, the DO swing required is fast and narrow. The economics of that strategy depend entirely on measurement that responds quickly and without offset.</p>
<p>Some plants extend this further with online analyzers for ammonium or nitrate to feed an <a href="/blog/ammonia-sensor-drift-wastewater-aeration-control.html">ammonia-based aeration control (ABAC) loop</a>. Those instruments use a different measurement principle than a DO probe, and what they actually detect versus what they are marketed as measuring matters before you specify one. <a href="https://aquamesh.ai/blog/spectral-change-analysis-vs-calibrated-parameters.html">Spectral change analysis versus calibrated parameters</a> covers the distinction.</p>
<p>Accurate measurement also creates a defensible record of what happened. When a plant documents savings for a utility incentive program or a regulatory report, both the baseline and current performance need measurement that holds up to scrutiny. For plants on nitrogen-limited discharge permits, the exposure created by DO control gaps and sensor drift is covered in <a href="/blog/nitrogen-permit-compliance-sampling-gap.html">NPDES nitrogen permit compliance: the DMR sampling gap</a>.</p>
<p>The same measurement-loop dynamic drives water waste in cooling towers, where a conductivity sensor reading high causes the system to discharge water still within operating range. The cost structure for that failure mode is in <a href="https://aquamesh.ai/blog/cooling-tower-blowdown-conductivity-sensor-drift.html">cooling tower blowdown control: what conductivity sensor drift costs in water and chemicals</a>. In activated sludge, the same drift accumulation affects pH electrodes, where a biased reading creates alkalinity overdosing and can mask early nitrification inhibition: <a href="/blog/ph-sensor-drift-wastewater-activated-sludge.html">pH sensor drift in activated sludge: the cost of 0.5 units</a>.</p>
<h2 id="how-to-cut-blower-energy-without-new-equipment">How to cut blower energy without new equipment</h2>
<p>The FWPCOA analysis cited above documents 10 to 40 percent reductions in blower energy from demand-based DO control, compared to running blowers without feedback. Plants already running VFDs in manual mode, or with feedback loops that have not been recommissioned since original installation, often have most of that gap still available. The five steps below work in sequence; later steps depend on the earlier ones being in place.</p>
<p><b>1. Verify DO sensor accuracy before adjusting any setpoint.</b> A setpoint change based on a drifted sensor moves the control loop to a different wrong reference. The first step is confirming what the sensor is actually reading, using a calibrated portable meter at different times of day and across the full diurnal load range, not just immediately after cleaning or calibration. This establishes whether subsequent steps are starting from an accurate baseline. How often DO sensors should be recalibrated in activated sludge mixed liquor, and what calibration intervals are appropriate for different sensor types, is covered in <a href="/blog/do-sensor-calibration-interval-activated-sludge.html">DO sensor calibration intervals in activated sludge</a>.</p>
<p><b>2. Reduce the DO setpoint to the process minimum your permit allows.</b> Most activated sludge systems designed for BOD removal can sustain full nitrification at setpoints between 1.5 and 2.0 mg/L. The EPA Nutrient Control Design Manual (2010, Section 5) documents that nitrification in suspended-growth systems can proceed at dissolved oxygen concentrations as low as 0.5 mg/L, with typical operating setpoints higher to buffer for load variation. If your plant is running a legacy setpoint of 2.5 or 3.0 mg/L without a process reason for it, the difference is directly recoverable blower energy. (<a href="https://www.epa.gov/sites/default/files/2015-06/documents/nutrient-control-design-manual-full.pdf" rel="noopener">EPA Nutrient Control Design Manual, 2010</a>)</p>
<p><b>3. Close the feedback loop if it is running open.</b> VFDs installed without an active PID loop holding them to a DO setpoint are often switched to manual by operators who found manual mode more predictable. An open-loop VFD captures some savings compared to a fixed-speed motor but none of the demand-following reduction that justifies the installation. Closing the loop requires that step 1 is true: the feedback signal has to be accurate, or the closed loop will chase measurement noise.</p>
<p><b>4. Set blower discharge pressure to the minimum the diffuser system requires.</b> Discharge pressure setpoints are frequently established at commissioning and not revisited. Reducing header pressure in measured increments, while monitoring airflow distribution and DO response across the basin, identifies how much design margin was left in place. The installed blower manufacturer's pressure-power curve quantifies the energy difference for the specific unit; it should be in the plant files or is available from the manufacturer directly.</p>
<p><b>5. Stage blower capacity to follow actual oxygen demand.</b> Running extra standby capacity continuously, or cycling blowers on a fixed time schedule regardless of load, keeps equipment running in inefficient operating regions. A staging sequence that adds or drops blower capacity based on measured demand keeps each running unit nearer its design point. For plants with multiple basins and a shared air header, <a href="https://aquamesh.ai/blog/why-parallel-equipment-drifts.html">the parallel-basin divergence problem</a> can mask whether staging is working as intended; blower runtime comparisons across basins are a useful check.</p>
<p>None of these changes requires new equipment. Steps 2 through 5 all depend on step 1 being correct, because a setpoint, a feedback loop, a pressure target, or a staging sequence that starts from a drifted sensor reading optimizes against the wrong input.</p>
<h2 id="three-questions-worth-asking-before-tuning-the-control-loop">Three questions worth asking before tuning the control loop</h2>
<p>Before commissioning a new aeration control strategy or investigating why existing savings have eroded, three questions are worth answering directly from available data.</p>
<p><b>Are the DO sensors reading accurately?</b> A comparison between sensor output and a calibrated handheld grab at different times of day, not just right after calibration, shows whether a sensor is holding its reading or drifting between service intervals.</p>
<p><b>What does the blower run-time history show?</b> If blowers are running at sustained high output during periods when influent flow and temperature are moderate, the measurement feeding the control loop deserves examination before changing the control logic itself.</p>
<p><b>Are parallel basins loading evenly?</b> Persistent divergence in blower output between basins handling equivalent loads is almost always a measurement issue. It rarely responds to setpoint adjustments alone.</p>
<p>None of these analyses requires new hardware. The data is usually in the historian. The constraint is having a way to look at it at the right resolution, over a long enough window, against the right reference.</p>
<hr />
<p>If your plant is spending in the range described above, the measurement question is worth examining directly. An initial review of your historian data can show whether the control loop is working from accurate inputs and where the largest gaps are.</p>
<p>See how AquaMesh approached this analysis at a membrane bioreactor reuse plant: <a href="https://aquamesh.ai/case-studies/water-reuse-plant.html">a membrane-bioreactor water reuse plant case study</a>. Or <a href="https://aquamesh.ai/#contact">reach out to schedule a working session with the team</a>.</p>
<p><em>Alex Towfigh is a founder at AquaMesh.</em></p>]]></content:encoded>
      <category>aeration energy</category>
      <category>wastewater treatment</category>
      <category>dissolved oxygen</category>
      <category>blower control</category>
      <category>energy optimization</category>
    </item>
    <item>
      <title>Why dairy plants miss their most costly process losses</title>
      <link>https://aquamesh.ai/blog/dairy-foods-process-losses.html</link>
      <guid isPermaLink="true">https://aquamesh.ai/blog/dairy-foods-process-losses.html</guid>
      <pubDate>Mon, 12 Oct 2026 09:00:00 GMT</pubDate>
      <description>Milk solids lost during transitions often go unmeasured. A Dairy Foods contributed article on spectral sensing and context for finding them.</description>
      <content:encoded><![CDATA[<p>Dairy Foods published a contributed article by Alex Towfigh, AquaMesh's founder and CEO, on 12 October 2026. The full piece is on the Dairy Foods site: <a href="https://www.dairyfoods.com/articles/99452-why-dairy-plants-still-miss-their-most-expensive-process-losses" rel="noopener">Why dairy plants still miss their most expensive process losses</a>.</p>
<p>What follows is a short summary with additional context. The article itself is exclusive to Dairy Foods.</p>
<h2 id="the-scale-of-the-problem">The scale of the problem</h2>
<p>A dairy processing facility moves large volumes of product through shared piping continuously. Industry estimates put routine milk-to-sewer losses at one to three percent of the milk a plant processes per year. At a mid-size fluid milk plant processing 500,000 pounds per day, that range translates to roughly 1,500 to 4,500 pounds of milk solids entering the wastewater stream daily, before a single equipment failure or operator error occurs.</p>
<p>The losses do not show up as a single large event. They accumulate across hundreds of small decisions made every shift: when to divert, when to release the line, when to call a CIP complete. A quarterly accounting of milk-to-drain volume often looks like normal shrinkage until individual transitions are examined one at a time.</p>
<p>Federal effluent guidelines for dairy products processing set BOD and TSS limits on wastewater discharge (<a href="https://www.ecfr.gov/current/title-40/chapter-I/subchapter-N/part-405" rel="noopener">40 CFR Part 405</a>; see also the <a href="https://www.epa.gov/eg/dairy-products-processing-effluent-guidelines" rel="noopener">EPA Dairy Products Processing Effluent Guidelines</a>). A plant's permit reflects those limits, but a permit only tells you whether the total load passed the threshold. It says nothing about which transitions produced the loss or what tighter endpoint detection would have recovered.</p>
<h2 id="the-argument-in-brief">The argument in brief</h2>
<p><b>The losses happen during transitions.</b> Published research puts routine losses of milk to sewer at 1% to 3% of the milk a plant processes. Most of it leaves during startups, interruptions, shutdowns and rinses: short windows where the pipe contents change and someone has to decide where product ends and water begins.</p>
<p><b>More tags do not fix a missing measurement.</b> Conductivity tells you something changed, but not how much product is in the line. A flowmeter tells you how much is moving, but not what it is. Lab results come back after the decision has been made. If the plant never measured the chemistry, software running on the historian has nothing to analyze.</p>
<p><b>Spectral sensing helps, if it can be trusted.</b> An optical spectrum carries far more information than a single-variable sensor. But temperature, bubbles, solids and a fouled window all move the reading. Without local calibration, a dirty window can look like a product loss, so every reading needs a confidence level attached. How the distinction between calibrated parameter outputs and spectral change detection applies in practice is covered in <a href="/blog/spectral-change-analysis-vs-calibrated-parameters.html">spectral change analysis versus calibrated parameters</a>.</p>
<p><b>Context decides what a signal means.</b> Rising conductivity is expected during caustic circulation and alarming during a product transfer. The article describes starting from data the plant already has (CIP steps, valve positions, lab results) to establish operating state, adding sensing only where a real blind spot exists, and producing an explanation of what changed and why rather than another generic alarm.</p>
<p><b>Prove it one event at a time.</b> Start read-only. Existing controls and sanitation checks stay in charge. Pick one expensive event, measure its baseline, catch it earlier, change the decision, and verify the result. One peer-reviewed UHT cleaning study cited in the piece used phase-endpoint measurement to cut a cleaning sequence from 142 to 66 minutes and roughly halve its effluent.</p>
<h2 id="why-cip-cycles-are-worth-measuring-closely">Why CIP cycles are worth measuring closely</h2>
<p>CIP is where most of the volume decision happens. A caustic wash, rinse, acid wash and final rinse can each carry residual product if the endpoint is missed. Many plants time their steps rather than measure them: the rinse runs for three minutes because that has worked in the past. Three minutes is a proxy for conductivity dropping below a threshold, which is itself a proxy for product concentration falling low enough to release the line.</p>
<p>Measuring directly, using a sensor that distinguishes caustic from milk fat from clean water, collapses that chain of proxies. It also creates an audit trail: was this CIP event as clean as the last one? If conductivity at the drain started high and ended high, that is a different failure than a rinse that ran long but ended clean. The sensor data makes the diagnosis available in real time rather than weeks later when a customer complaint or effluent exceedance arrives.</p>
<p>The UHT study cited in the Dairy Foods article is a useful benchmark. Cutting cycle time from 142 to 66 minutes is not a marginal gain; it changes the economics of the line. The same logic applies to the product-to-water transition at the start of a run: an earlier, more accurate determination of when to divert lets the plant recover product that would otherwise go to drain.</p>
<p>This connects to the same measurement challenge described in <a href="/blog/why-parallel-equipment-drifts.html">why parallel equipment drifts and what it costs</a> and the broader case in <a href="/blog/what-25-days-of-historian-data-shows.html">what 25 days of historian data reveals</a>: the historian typically has flow and conductivity records; what it lacks is component-level identification that turns a conductivity trend into a recoverable, accountable event.</p>
<h2 id="next-step">Next step</h2>
<p>If your facility has a CIP loop or product-to-water transition that runs on timers rather than direct measurement, AquaMesh can show what an instrumented baseline looks like at one point in that system. <a href="https://aquamesh.ai">Request a demo</a> or read the full article on the Dairy Foods site.</p>
<p><a href="https://www.dairyfoods.com/articles/99452-why-dairy-plants-still-miss-their-most-expensive-process-losses" rel="noopener">Read the full article in Dairy Foods</a>.</p>]]></content:encoded>
      <category>dairy</category>
      <category>product loss</category>
      <category>CIP</category>
      <category>spectral sensing</category>
      <category>process water</category>
      <category>Dairy Foods</category>
    </item>
    <item>
      <title>California data center water reporting</title>
      <link>https://aquamesh.ai/blog/california-data-center-water-reporting.html</link>
      <guid isPermaLink="true">https://aquamesh.ai/blog/california-data-center-water-reporting.html</guid>
      <pubDate>Thu, 01 Oct 2026 09:00:00 GMT</pubDate>
      <description>AB 2619 requires California data centers to report direct water use by source at renewal, under penalty of perjury. AB 2469 adds a Water Scarcity Plan by 2028.</description>
      <content:encoded><![CDATA[<p>On 21 September 2026 the Governor of California signed <a href="https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=202520260AB2619" rel="noopener">AB 2619</a>, now Chapter 437 of the 2026 Statutes, and <a href="https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=202520260AB2469" rel="noopener">AB 2469</a> alongside it. Between them they change what a data center operator in California has to be able to say about water, and when.</p>
<p>The reporting obligation is the part worth reading closely, because it is more specific than the coverage suggests.</p>
<h2 id="what-ab-2619-actually-asks-for">What AB 2619 actually asks for</h2>
<p>AB 2619 works through the business license process rather than through a new permit. It sorts data centers into three types: Type I, or hyperscale, at more than 10,000 servers or more than 25 megawatts; Type II between 2 and 25 megawatts; and Type III below 2 megawatts.</p>
<p>At initial application, the operator gives the water supplier, and repeats on the license application form, a good faith estimate of expected water use, the anticipated source, and projected volume for the maximum day, the maximum month, and the average year.</p>
<p>At every renewal, the operator reports the preceding calendar year's annual water use (and here the statute is precise) "including total water use and direct water use," along with the cooling system type.</p>
<p>Direct annual water use is defined in the bill as:</p>
<blockquote><p>The volume of water withdrawn, delivered, or otherwise used onsite for data center operations, including cooling, sanitation, irrigation, and any other operational use, identified by source, including potable water, nonpotable water, or recycled water.</p></blockquote>
<p>All of it under penalty of perjury.</p>
<p>Two things in that definition do the work. The first is <em>identified by source</em>: potable, nonpotable, recycled, broken out, not aggregated. The second is the split between total and direct, which means a single number off the utility bill does not satisfy the renewal requirement.</p>
<p>One correction worth making, because it has circulated widely: several summaries describe AB 2619 as requiring disclosure of <em>indirect</em> water use, meaning the water consumed generating the site's electricity. The enacted text does not define indirect use and does not require it. If you are scoping a compliance programme, scope it to what the statute says.</p>
<h2 id="what-ab-2469-adds-and-the-2028-date">What AB 2469 adds, and the 2028 date</h2>
<p>AB 2469 operates on the other side of the counter. It applies to cities and counties approving discretionary or ministerial permits for data center construction or expansion that increases maximum peak water use, and it requires a Water Supply Assessment under <a href="https://leginfo.legislature.ca.gov/faces/codes_displayText.xhtml?lawCode=WAT&amp;amp;division=6.&amp;amp;title=&amp;amp;part=2.10.&amp;amp;chapter=&amp;amp;article=" rel="noopener">Water Code section 10910 et seq.</a>, projected water use and efficiency measures, workforce disclosures, and an undertaking that the applicant carries the full cost of any water conveyance, treatment, storage or distribution infrastructure the project needs.</p>
<p>Effective 1 January 2028, it also requires a Water Scarcity Plan. That plan has to set out measures for five named conditions: an abnormally dry year, and moderate, severe, extreme and exceptional drought years, including staged reductions, curtailment, recycling and reuse, thermal load reductions, and load shedding.</p>
<p>AB 2469 does not mandate closed-loop cooling. Some coverage says it does. It does not.</p>
<h2 id="the-gap-this-opens">The gap this opens</h2>
<p>A Water Scarcity Plan is a commitment to cut consumption by a stated amount under stated conditions, written down in advance, in front of a permitting authority. Writing a credible one requires knowing two things that many sites do not currently know with any precision.</p>
<p>The first is where the water is going now, by loop and by source. Evaporative cooling, blowdown, makeup, sanitation and irrigation are different lines with different elasticities, and a site that only knows its total has no basis for saying which it would cut first.</p>
<p>The second is how far each one can be cut before something else breaks. Curtailment on a cooling tower usually means running at higher cycles of concentration: less blowdown, less makeup, and a circuit that is now closer to the scaling and corrosion limits of its own chemistry. How conductivity sensor drift interacts with that control strategy, and what it costs when blowdown runs at the wrong frequency, is covered in <a href="/blog/cooling-tower-blowdown-conductivity-sensor-drift.html">cooling tower blowdown: conductivity control and sensor drift</a>. Recycling and reuse, both named in the statute, move in the same direction. The water you keep in the loop is water whose composition is drifting.</p>
<p>That is the real operational content of a scarcity plan. It is not a reporting exercise, it is a commitment to run the plant closer to its limits on demand, and the plan is only as good as the site's ability to see what happens when it does.</p>
<h2 id="where-measurement-earns-its-place">Where measurement earns its place</h2>
<p>Volumes are a flow metering problem, and the renewal report in AB 2619 is largely a question of whether a site is instrumented well enough to break its own consumption apart by source. Many are not, and 2027's renewals will be the first time that shows up with a perjury declaration attached.</p>
<p>Composition is a different problem, and it is the one that decides whether a curtailment commitment is safe to make. A plant running higher cycles, or taking a recycled source that was not in the original design basis, is operating a water chemistry that moves. <a href="/blog/spectral-change-analysis-vs-calibrated-parameters.html">Spectral change analysis</a> is a reasonable fit here precisely because it does not depend on having a calibrated panel for every constituent that might matter: it watches the spectrum for what is correlating with something the operator cares about, and flags when that relationship moves. BOD is the one parameter AquaMesh calibrates; the rest is change detection.</p>
<p>For sites that already have a few months of history in a historian, the first question is usually answerable from data that exists. <a href="/blog/what-25-days-of-historian-data-shows.html">What 25 days of one plant's data showed</a> is a reasonable picture of what that kind of look turns up, and it was not a water reporting exercise: it was a cost exercise that happened to require knowing where things actually go.</p>
<p>Nothing in either bill requires a sensor. Both require a site to be able to answer questions about its own water that a surprising number of operators currently answer by estimate. The first renewals under AB 2619 will establish how much tolerance there is for that, and the 2028 scarcity plans will establish how much tolerance there is for a commitment that nobody has tested.</p>
<p>Operators working through what this means for an existing site can see how we approach <a href="/industries/data-centers.html">data center cooling water</a>, or start with an audit of the data already being collected. Understanding what the site's own WUE figure actually measures, and what it omits, is useful context before committing to a curtailment plan; <a href="/blog/data-center-wue-measurement-gap.html">the measurement gap in data center WUE</a> covers that ground.</p>
<p><em>Alex Towfigh is a founder at AquaMesh.</em></p>]]></content:encoded>
      <category>data centers</category>
      <category>water reporting</category>
      <category>AB 2619</category>
      <category>AB 2469</category>
      <category>cooling water</category>
      <category>water reuse</category>
      <category>California</category>
    </item>
    <item>
      <title>What 25 days of historian data reveals</title>
      <link>https://aquamesh.ai/blog/what-25-days-of-historian-data-shows.html</link>
      <guid isPermaLink="true">https://aquamesh.ai/blog/what-25-days-of-historian-data-shows.html</guid>
      <pubDate>Wed, 30 Sep 2026 09:00:00 GMT</pubDate>
      <description>A 25-day SCADA export found $42,000/yr in losses at an MBR plant with no site visit. Method, normalized energy benchmarks, and a six-step replication guide.</description>
      <content:encoded><![CDATA[<hr />
<p>title: What 25 days of historian data reveals slug: what-25-days-of-historian-data-shows date: 2026-09-30 section: Field notes description: A 25-day SCADA export found $42,000/yr in losses at an MBR plant with no site visit. Method, normalized energy benchmarks, and a six-step replication guide. image: assets/audit/fig1.jpg tags: historian data, wastewater energy, plant audit, SCADA, membrane bioreactor</p>
<hr />
<p>A 25-day export of a water reuse plant's historian located roughly $42,000 a year in recoverable losses without a site visit, new instruments, or a capital request. The main items were aeration energy running above the plant's own control targets, membrane scour airflow above the manufacturer's specific guidance, and pump cycling at more than twenty times the design rate.</p>
<p>The method is repeatable on any plant with a SCADA or DCS historian: pull an export, locate the parallel equipment pairs, compare each one's performance against the plant's own posted operating ranges, and price every gap that falls outside them. The rest of this post covers that method in detail.</p>
<p>That figure came entirely from the plant's existing SCADA archive: 1,001 tags, 7,172,532 recorded values, covering 28 June to 23 July.</p>
<figure class="post-fig"><img src="../assets/audit/fig1.jpg" alt="Audit summary page: no-capital savings, fully instrumented savings, design flow percentage, and effluent turbidity. Values tagged M are measured from SCADA; E means calculated with stated assumptions." loading="lazy" decoding="async" /><figcaption>The front page of the delivered audit. Values tagged <b>M</b> are measured straight from the plant's SCADA; <b>E</b> means we calculated them, and the assumptions are written down.</figcaption></figure>
<h2 id="why-a-historian-export-is-enough-to-start">Why a historian export is enough to start</h2>
<p>Almost every plant already records far more than anyone reads. A SCADA historian captures every tag on a schedule, forever, and the overwhelming majority of those tags are never looked at unless something breaks. The data is not the constraint. Attention is.</p>
<p>Three properties make that archive useful:</p>
<ul class="post-list"><li><b>It is continuous.</b> A lab result describes one moment. A historian describes every moment, which is the only way to see a slope rather than a point.</li><li><b>It contains pairs.</b> Plants run parallel equipment, twin basins, duty and standby pumps, two membrane trains. Equipment built to do the same job under the same conditions should behave the same way. When it stops, that divergence is a measurement in itself.</li><li><b>It is already paid for.</b> No capital request, no procurement cycle, no installation window.</li></ul>
<h2 id="the-method">The method</h2>
<p>For each signal we model what normal looks like for that signal at that plant, rather than against a textbook range. Where the plant publishes its own operating ranges, most do on the weekly sampling schedule, those become the reference. A finding only counts as a finding if it sits outside the range the plant itself posts.</p>
<p>Then each item is priced, which is the part that decides whether anyone acts:</p>
<ul class="post-list"><li>Blower power from the isentropic relation, using measured airflow and measured discharge pressure</li><li>Pump power from measured flow and head</li><li>Runtime from historian totaliser deltas</li><li>Everything converted to money using the plant's own tariff</li></ul>
<p>Each line is labelled <b>measured</b> or <b>estimated</b>. Where it is an estimate, the assumptions are written down so the plant's own engineer can check the arithmetic rather than take a vendor's word for it.</p>
<h2 id="normalizing-findings-by-output">Normalizing findings by output</h2>
<p>Absolute dollar figures tell you what an opportunity is worth at one plant at one tariff. Normalizing by output tells you whether a plant is already close to its floor or has structural headroom left.</p>
<p>Energy intensity is the standard metric. The <a href="https://www.epa.gov/sites/default/files/2015-08/documents/wastewater-energymanagement.pdf" rel="noopener">EPA's Energy Management Guidebook for Wastewater and Water Utilities</a> benchmarks conventional activated sludge with secondary treatment and UV disinfection at roughly 400 to 1,000 kWh per million gallons treated, depending on nutrient removal requirements. Plants with membrane bioreactors typically run higher, in the range of 2,000 to 4,000 kWh per MG, because continuous air scouring to keep membranes permeable adds a large fixed aeration load that conventional basins do not carry. Plotting a plant's measured intensity on that distribution answers a different question than the dollar figure alone: whether an opportunity is ordinary or exceptional for a plant of that type.</p>
<p>The same logic extends to chemical dosing. Chlorine, polymer, and coagulant dose all should track the flow or load passing through the relevant process step. Dose per unit volume, expressed as mg/L or lb per MG in US customary units, is the trackable metric. A flat chemical dose through wide seasonal flow variation shows up clearly in normalized form. Raw dose data measured in gallons of solution pumped per day hides both overdosing and load-proportional drift.</p>
<p><b>Modelled chemicals example:</b> A plant dosing sodium hypochlorite for effluent disinfection at a constant pump rate, regardless of treated flow, will show a normalized dose that rises as flow falls even though the raw pump-rate trend looks flat. At 1.5 MGD, a constant 2 gallons per hour of 12.5 percent sodium hypochlorite solution delivers roughly 4 mg/L as Cl2. At 0.5 MGD, the same rate delivers roughly 12 mg/L, three times the residual target without any change to chemical spend per day. The overdosing does not appear in the raw historian pump-rate trend. It becomes visible the moment dose is divided by flow. In the historian, this shows as a flat pump-rate line through variable flow, where a correctly controlled system would show a line that tracks flow proportionally. Pump rates that do not vary with load are candidates for proportional control, and the historian already contains the data to confirm or rule it out.</p>
<p>For the MBR plant in this audit, the actual flow rate is withheld to preserve anonymity. The calculation is the same at any plant: total kWh divided by total volume treated for energy intensity; total chemical mass or volume divided by total flow for normalized dose rate. One month of historian data plus a flow totaliser read is enough to place a plant on the EPA benchmark curve.</p>
<h2 id="what-came-out">What came out</h2>
<h3 id="one-basin-running-3-6x-above-its-own-posted-range">One basin running 3.6x above its own posted range</h3>
<p>The plant's sampling schedule posts pre-air dissolved oxygen at 0.5 to 1.5 mg/L. Basin 2 sat right in that band, median 0.49 mg/L. Basin 1 held a median of 5.35 mg/L on every day of the record, roughly eleven times its twin, with its blower command pinned at 99.99%.</p>
<p>Both basins draw on the same shared air header. That detail is what turns this from a mystery into a maintenance task: if the air supply is common and the outcomes are not, the problem is in the split, a valve or a damper, not in the control strategy. Worth $3,316 a year in blower power, plus the nitrogen-removal risk: the <a href="https://www.epa.gov/sites/default/files/2019-08/documents/nutrient-control-design-manual.pdf" rel="noopener">EPA Nutrient Control Design Manual, 2010</a> identifies dissolved oxygen intrusion into anoxic zones as a primary cause of denitrification inefficiency, since aerobic organisms outcompete denitrifiers at elevated DO concentrations. For plants using intermittent aeration to achieve biological nitrogen removal in a single basin, excess DO also disrupts aerobic and anoxic cycle timing, compounding nitrogen removal losses and increasing blower runtime. The mechanism is documented in <a href="/blog/intermittent-aeration-nitrogen-removal-do-control.html">intermittent aeration for nitrogen removal: why DO sensor precision matters</a>. For plants subject to nitrogen permit limits, that same process inconsistency maps to permit risk: <a href="/blog/nitrogen-permit-compliance-sampling-gap.html">NPDES nitrogen permit compliance: the DMR sampling gap</a> covers how monthly grab-sample schedules can miss real exceedances that continuous monitoring would catch.</p>
<p>One caveat worth repeating: a fouled DO probe can sit on an offset and still respond to airflow changes, making it hard to distinguish a measurement artifact from a real process difference. The reading gets verified against a calibrated handheld before anyone opens a valve. How quickly DO sensors drift in activated sludge mixed liquor, and what calibration intervals are appropriate for that environment, is covered in <a href="/blog/do-sensor-calibration-interval-activated-sludge.html">DO sensor calibration intervals in activated sludge</a>.</p>
<h3 id="membrane-scour-air-above-what-the-flux-needed">Membrane scour air above what the flux needed</h3>
<p>Airflow tracked the permeate rate almost perfectly, R-squared of 1.00, at a ratio well above the manufacturer's specific-scour guidance. Trimming toward the guideline at the measured flux, using the controls already installed, came to $5,391 a year. That was the largest no-capital line on the list.</p>
<h3 id="pumps-starting-every-twenty-five-seconds">Pumps starting every twenty-five seconds</h3>
<p>Reclaimed water pumps were logging 13 to 28 starts an hour against a manufacturer guideline near six for pumps that size. One asset logged 143 in an hour, a start every twenty-five seconds.</p>
<p>Every start draws locked-rotor current and heats the windings far more than running does. The <a href="https://www.energy.gov/sites/prod/files/2014/05/f16/pump.pdf" rel="noopener">DOE's "Improving Pumping System Performance: A Sourcebook for Industry"</a> documents the relationship between start frequency and insulation aging: permissible starts per hour depend on motor frame size and thermal class, and the typical industrial limit for motors in the fractional to ten-horsepower range is well below what this historian was recording. This is a control deadband adjustment, not a purchase, and it was quietly writing a bill in motors, starters, and contactors: about $3,500 a year.</p>
<h3 id="sixteen-of-sixteen-reuse-totalisers-frozen">Sixteen of sixteen reuse totalisers frozen</h3>
<p>Every irrigation zone has a volume totaliser. Not one of them changed in twenty-five days.</p>
<p>Until they record, nobody can demonstrate how much recycled water went to beneficial reuse rather than straight to disposal, which is exactly the number a reuse programme is judged on. A configuration fix rather than a purchase, and it unlocks roughly $7,300 a year.</p>
<h2 id="what-it-added-up-to">What it added up to</h2>
<div class="table-wrap"><table class="cs-table"><thead><tr><th></th><th></th></tr></thead><tbody><tr><td>Direct savings, no capital at all</td><td><b>$42,050 / yr</b></td></tr><tr><td>Including about 7 h/week of reporting time returned</td><td><b>$58,428 / yr</b></td></tr><tr><td>Fully instrumented</td><td><b>$92,200 / yr</b></td></tr></tbody></table></div>
<p>Two subtotals are kept deliberately apart: money that comes off a bill, and operator hours given back. The hours are real, but they are not money off a bill, so they never get folded into the first number.</p>
<figure class="post-fig"><img src="../assets/audit/fig2.jpg" alt="Opportunity table with every line flagged measured or estimated, including blower energy, scour air, pump cycling, totaliser restoration, RAS flow, predictive maintenance, and reporting automation." loading="lazy" decoding="async" /><figcaption>Every line labelled. The predictive-maintenance figure is built asset by asset from the plant's own run hours and start counts, not taken as a percentage of a maintenance budget.</figcaption></figure>
<h2 id="how-to-run-this-at-your-plant">How to run this at your plant</h2>
<p>The same analysis works on any SCADA or DCS historian export. Six steps:</p>
<ol class="post-list"><li><b>Pull the export.</b> Three to six weeks covers most seasonal operating modes. IT can usually produce a CSV or Parquet pull from OSIsoft PI, Ignition, or a standard historian without a change request.</li><li><b>Inventory the tags.</b> Group them by equipment and then by process zone. The goal is not to read all 1,000 tags. It is to find the parallel equipment pairs and the flow totalizers.</li><li><b>Find the paired equipment.</b> Twin basins, duty and standby pumps, multiple filters or membrane trains all generate internal comparisons. A performance difference between units drawing on the same feed is a finding with the evidence already in the data. The mechanics of why parallel equipment diverges are covered in <a href="/blog/why-parallel-equipment-drifts.html">why parallel equipment drifts apart</a>.</li><li><b>Check against the plant's own ranges.</b> The sampling schedule or operations manual usually publishes target ranges. A historian signal sitting outside those ranges is a priority flag, because the plant itself told you what normal looks like.</li><li><b>Normalize by output.</b> Divide energy use by treated volume in kWh/MG and chemical dose by treated volume in lb/MG or mg/L. Plot each as a time series. A rising normalized trend over weeks or months indicates process drift that periodic grab sampling will not catch. The same measurement gap shapes compliance exposure for industrial dischargers under the National Pretreatment Program: <a href="/blog/industrial-pretreatment-self-monitoring-gaps.html">industrial pretreatment: the self-monitoring blind spot</a>.</li><li><b>Price each finding.</b> Blower and pump power from the isentropic relation or motor-rated amperage; chemical cost from unit price and dose rate; pump starts from the contactor log. Use the plant's own utility tariff. Label measured from estimated.</li></ol>
<p>The sequence takes longer the first time. The second time, most of the work is steps five and six.</p>
<h2 id="the-caveat-that-matters-most">The caveat that matters most</h2>
<p>Everything above is priced on a residential time-of-use schedule blending to $0.362/kWh. At 406,610 kWh a year this plant is well past residential scale and would normally sit on a commercial schedule at roughly half that blended rate.</p>
<p>On a commercial tariff, the same physical savings are worth about half what is shown.</p>
<p>The kilowatt-hours are measured either way; only their price is in question, and one utility bill settles it. We put that on the front page of the report rather than in a footnote, because showing the range is worth more than picking the flattering end of it.</p>
<h2 id="what-this-does-not-tell-you">What this does not tell you</h2>
<p>A twenty-five day window in summer will not capture seasonal variation. Capital figures other than instrumentation are order-of-magnitude and need firming with quotes. And every number here is a ceiling that assumes the recommendation is actually implemented and then maintained: an identified opportunity, not a realised saving.</p>
<p>For context on the energy benchmarking side, the <a href="/blog/aeration-energy-cost-wastewater-plants.html">wastewater blower energy cost post</a> covers how to compare blower costs against facility-type benchmarks and what continuous DO monitoring changes about that math. For how blower VFD energy savings erode over time and what the measurement failures look like in historian data, see <a href="/blog/wastewater-blower-vfd-energy-savings-erode.html">wastewater blower VFD: why energy savings erode</a>.</p>
<figure class="post-fig"><img src="../assets/audit.jpg" alt="The AquaMesh recommendations view: findings with severity, the action to take, the evidence tags and the annual value, with accept, resolve and dismiss on each card." loading="lazy" decoding="async" /><figcaption>The same findings in the platform, where they keep recomputing on a rolling window rather than ageing in a PDF. Plant name removed; everything else is the product as it ships.</figcaption></figure>
<p>If your plant keeps a historian, the same exercise is available on your data. It needs an export and nothing else. For how that analysis runs on a rolling window rather than as a one-time audit: <a href="/modeling.html">AquaMesh modeling</a>.</p>
<p>Read the full anonymised audit: <a href="/case-studies/water-reuse-plant.html">a membrane-bioreactor water reuse plant</a>.</p>
<p><em>Alex Towfigh is a founder at AquaMesh.</em></p>]]></content:encoded>
      <category>historian data</category>
      <category>wastewater energy</category>
      <category>plant audit</category>
      <category>SCADA</category>
      <category>membrane bioreactor</category>
    </item>
    <item>
      <title>AquaMesh physical AI: early detection for industrial plants</title>
      <link>https://aquamesh.ai/blog/connect-press-release-physical-ai.html</link>
      <guid isPermaLink="true">https://aquamesh.ai/blog/connect-press-release-physical-ai.html</guid>
      <pubDate>Mon, 28 Sep 2026 09:00:00 GMT</pubDate>
      <description>Connect published AquaMesh's 2026 physical AI announcement. This post explains what the platform does and why industrial plant operators are paying attention.</description>
      <content:encoded><![CDATA[<p>The gap between what an industrial facility measures and what is actually happening inside its processes is where operating problems form. A sensor drifts. A chemical feed runs long. A blower draws more power than the control system expects. The signal often reaches the operator hours later in a lab result, the next morning in a shift log, or weeks later in a utility bill.</p>
<p>On September 28, 2026, <a href="https://connect.org/" rel="noopener">Connect</a>, the San Diego innovation organization, published an announcement about how AquaMesh addresses this gap. The announcement describes what the company calls physical AI: software intelligence that connects directly to physical process data and acts on what it finds.</p>
<p>The key distinction is coverage. Industrial plants generate large volumes of historian records, meter readings and lab results. What they often lack is the measurement density to catch a problem while it is still forming rather than after it has accumulated cost.</p>
<p><a href="https://connect.org/aquamesh-builds-physical-ai-to-help-industrial-facilities-catch-costly-problems-early/" rel="noopener">Read the full announcement on Connect</a>.</p>
<h2 id="what-physical-ai-means-in-practice">What "physical AI" means in practice</h2>
<p>Most enterprise AI applications operate on business documents: invoices, tickets, contracts, email. Physical AI, as the term is used in this context, means software that is connected directly to the instruments embedded in an operating process, reads what those instruments report in real time, and reasons about what the numbers mean for that process right now.</p>
<p>For a wastewater plant, that includes dissolved oxygen analyzers, ammonia sensors, flow meters and blower control signals. For a cooling water system it includes conductivity, temperature, blowdown flow and makeup water consumption. For a food processing facility it includes pH, conductivity, flow through CIP circuits and effluent parameters.</p>
<p>The AquaMesh platform connects to the data infrastructure a facility already has, identifies where measurement coverage is incomplete, adds sensing where necessary, and runs a continuous loop: detect an anomaly, locate the likely cause, estimate what it is costing, recommend a corrective action, and confirm the action worked.</p>
<h2 id="the-measurement-gap-problem">The measurement gap problem</h2>
<p>Industrial facilities operate under a structural tension. Continuous analyzers give real-time visibility into a process variable, but they require calibration, maintenance and interpretation. Laboratory samples give accurate reference measurements, but they are taken infrequently and results arrive after the fact. Manual readings fill in some gaps, but depend on shift timing and individual judgment.</p>
<p>In practice, this means that small deviations in a process variable often compound undetected between monitoring events. A dissolved oxygen sensor that has drifted 0.3 mg/L affects aeration control decisions for every cycle between calibration visits. A conductivity sensor reading high causes blowdown to run longer than the water balance requires. A chemical feed that is slightly over set point adds up across thousands of batches or cycles.</p>
<p>The U.S. Environmental Protection Agency's <a href="https://www.epa.gov/npdes/npdes-permit-limits" rel="noopener">NPDES permit program</a> requires industrial dischargers to monitor and report effluent quality at defined intervals. Those intervals set a floor for compliance reporting, but what happens between required sampling events is often tracked only indirectly. The gap between required monitoring and continuous process visibility is where many compliance exceedances originate and where the largest operating cost opportunities tend to sit.</p>
<p>This pattern is consistent across industrial water treatment, wastewater discharge compliance, food and beverage process efficiency, and data center cooling water management. The specific parameters differ, but the structure is the same: a measurement gap, a slow drift, a delayed signal.</p>
<h2 id="what-aquamesh-announced">What AquaMesh announced</h2>
<p>The September 2026 release describes a platform that connects to existing plant historian and control infrastructure and adds coverage where the measurement picture is incomplete. The core workflow operates in five steps: detect a process deviation, find the probable cause, estimate the operating cost of the deviation, recommend a corrective action, and verify that the correction held.</p>
<p>The target operating categories span water consumption, chemical use, energy, labor and discharge compliance. Alex Towfigh, AquaMesh CEO, described the central problem in the release: "We are closing the blind spot about what is physically happening inside a process." The stated target for operating cost improvement is a range of 10 to 25 percent across those categories. AquaMesh presents this as a target range based on design intent, not a guaranteed outcome for any facility.</p>
<p>The longer-term direction is facilities that identify and correct operating problems continuously, without waiting for the next lab sample or maintenance visit. The platform is designed to work with what a facility already has rather than requiring replacement of existing instrumentation or control systems.</p>
<p>The announcement was also distributed through <a href="https://www.prlog.org/13173077-aquamesh-builds-physical-ai-to-help-industrial-facilities-catch-costly-problems-early/" rel="noopener">PRLog</a>, which reaches a broad industrial and technology audience.</p>
<h2 id="who-this-applies-to">Who this applies to</h2>
<p>The platform is designed for facilities where water, chemical or energy costs are tracked against a permit, a tariff or an operating budget and where the measurement infrastructure exists to connect to.</p>
<p>In practice, this includes municipal and industrial wastewater treatment plants operating under NPDES discharge permits, industrial water treatment systems managing chemical dosing and energy for pumping and aeration, food and beverage processing plants with wastewater pretreatment obligations under EPA categorical pretreatment standards, and data centers managing cooling water systems subject to state or municipal water use reporting requirements.</p>
<p>The decision makers who typically carry responsibility for these costs are plant operations managers, directors of utilities, and VPs of operations. Environmental compliance officers at facilities with complex permit conditions often have a parallel interest in monitoring gap analysis.</p>
<p>For a closer look at the specific measurement problems the platform addresses, related posts on this site cover <a href="/blog/ammonia-sensor-drift-wastewater-aeration-control.html">why ammonia sensor drift affects aeration control</a>, <a href="/blog/nitrogen-permit-compliance-sampling-gap.html">the DMR sampling gap in nitrogen permit compliance</a>, and <a href="/blog/wastewater-blower-vfd-energy-savings-erode.html">how blower VFD energy savings erode between calibration events</a>.</p>
<h2 id="next-step">Next step</h2>
<p>If you manage a facility with water, energy or compliance operating costs and want to see what the platform finds in your data, <a href="https://aquamesh.ai">request a demo at aquamesh.ai</a>.</p>
<p><em>Alex Towfigh, AquaMesh</em></p>]]></content:encoded>
      <category>physical AI</category>
      <category>industrial water treatment</category>
      <category>plant operations</category>
      <category>monitoring</category>
      <category>wastewater</category>
    </item>
    <item>
      <title>Why parallel equipment drifts and what it costs</title>
      <link>https://aquamesh.ai/blog/why-parallel-equipment-drifts.html</link>
      <guid isPermaLink="true">https://aquamesh.ai/blog/why-parallel-equipment-drifts.html</guid>
      <pubDate>Sat, 26 Sep 2026 09:00:00 GMT</pubDate>
      <description>Parallel blowers, pumps and process trains drift apart from filter fouling, wear and sensor mismatch. Here is what the gap costs and how to find it.</description>
      <content:encoded><![CDATA[<p>Parallel blowers and pumps in treatment plants drift apart for three reasons: inlet filter fouling, mechanical wear, and sensor or setpoint mismatch on one unit. Each is fixable. The gap between a 150 hp blower pair drifting 12 percentage points apart from filter loading alone can cost $5,000 to $7,000 per year in excess electricity -- and the fix is a filter service, not a capital project. The problem is that the gap produces no control-system alarm. Two basins can sit inside every operating limit while running at completely different airflow rates, and the divergence compounds for months or years undetected.</p>
<p>The cost shows up in three ways: wasted blower electricity, accelerated wear on whichever unit is running harder, and measurement errors that propagate into process and compliance decisions. A basin pair drifting 20 percentage points apart in aeration demand adds hundreds to over a thousand dollars a year in energy alone, before accounting for the wear asymmetry that follows.</p>
<p>The most reliable finding in plant data is also the simplest: two things built to do the same job, under the same conditions, that have stopped behaving the same way.</p>
<p>It is reliable because it needs no baseline, no model of correct operation, and no agreement about what good looks like. The two units are each other's reference. If one has drifted from the other, something changed, and that something usually has a cost attached.</p>
<h2 id="why-the-control-system-will-not-tell-you">Why the control system will not tell you</h2>
<p>A modern control system is very good at holding a value and raising an alarm when a limit is crossed. Divergence between two units is neither of those things.</p>
<p>Both units can sit inside every alarm limit while being nowhere near each other. A pump at 84% duty and its twin at 39% are both running. A basin at 5.35 mg/L and its twin at 0.49 mg/L are both aerated. Nothing is out of range in the sense the alarm system understands, so nothing fires, and the gap persists for years because it never presents as an event.</p>
<p>That is the gap worth watching, and it is visible in data a plant already records.</p>
<figure class="post-fig post-dia"><svg class="dia" viewBox="0 0 960 300" role="img" aria-label="Two basins on shared airflow over the same period. Basin 1 holds a median of 5.35 mg/L of dissolved oxygen with its blower pinned at 99.99%; Basin 2 sits inside the posted range of 0.5 to 1.5 mg/L at 0.49 mg/L.">
  <g font-family="Plus Jakarta Sans, sans-serif">
    <rect x="70" y="150" width="820" height="46" fill="#e9effd"/>
    <text x="80" y="214" font-size="13" font-weight="600" fill="#2449e8">POSTED RANGE 0.5–1.5 mg/L</text>
    <path d="M70 60 C 150 66, 230 54, 310 62 S 470 56, 550 64 S 710 58, 890 62" fill="none" stroke="#c2410c" stroke-width="3"/>
    <path d="M70 176 C 150 170, 230 182, 310 174 S 470 180, 550 172 S 710 178, 890 174" fill="none" stroke="#2449e8" stroke-width="3"/>
    <text x="70" y="44" font-size="15" font-weight="700" fill="#c2410c">Basin 1 — median 5.35 mg/L, blower pinned at 99.99%</text>
    <text x="70" y="248" font-size="15" font-weight="700" fill="#2449e8">Basin 2 — median 0.49 mg/L, same shared airflow</text>
    <path d="M905 62 V174" stroke="#66728a" stroke-width="1.5" stroke-dasharray="4 4"/>
    <text x="898" y="120" text-anchor="end" font-size="14" font-weight="700" fill="#0b1026">3.6× the posted limit</text>
  </g>
</svg></figure>
<p>Neither trace is outside an alarm limit. Both basins are aerated, both are running, and nothing fires.</p>
<h2 id="what-drives-the-gap-in-blowers-and-pumps">What drives the gap in blowers and pumps</h2>
<p><b>Blowers</b> drift apart for three main reasons. Inlet filter loading is the most common: as filters accumulate dust, mist and biogenic material, differential pressure rises and actual airflow falls even as the speed command stays the same. Two blowers in the same building foul their filters at different rates depending on position and local airflow patterns. Impeller deposits, particularly from fine aerosols and greasy compounds in municipal treatment plants, shift the speed-to-output curve over years of operation. And control-side drift compounds both effects: a dissolved oxygen sensor reading above actual suppresses the blower command on one basin while its twin continues aerating at a higher level, creating a gap that looks like a mechanical problem until the sensors are checked.</p>
<p>For plants with VFD-equipped blowers, there is a fourth degradation path independent of sensors: belt and coupling losses, impeller wear, and pressure setpoint creep each reduce the airflow a blower delivers at a given speed command. <a href="/blog/wastewater-blower-vfd-energy-savings-erode.html">Wastewater blower VFD: why energy savings erode</a> covers how to identify whether a divergence between parallel blowers has a mechanical cause, a sensor cause, or both.</p>
<p><b>Pumps</b> drift from wear. Return sludge and mixed liquor carry grit that erodes impeller vanes; wear ring clearance grows, and the same speed command delivers progressively less flow. Bearing friction increases as lubrication degrades. Mechanical seals add drag as they age. None of these changes triggers an alarm, but each shifts the relationship between command and output.</p>
<p><b>Sensor and setpoint mismatch</b> drives divergence on either type of equipment. A dissolved oxygen sensor reading 0.3 mg/L above actual holds one basin's aeration command lower than its twin. A conductivity sensor reading high on a blowdown line, as described in <a href="/blog/cooling-tower-blowdown-conductivity-sensor-drift.html">cooling tower blowdown and conductivity drift</a>, opens the blowdown valve beyond what is needed. The control logic on each unit is executing correctly; the gap is upstream of it, in the measurement. How frequently DO sensors drift and what that calibration interval costs in blower energy is covered in <a href="/blog/do-sensor-calibration-interval-activated-sludge.html">DO sensor calibration intervals for activated sludge</a>.</p>
<p>Identifying which cause applies determines the fix: filter cleaning, impeller inspection, or sensor recalibration. The first step is seeing that the gap exists.</p>
<h2 id="the-three-shapes-it-takes">The three shapes it takes</h2>
<h3 id="duty-that-never-alternates">Duty that never alternates</h3>
<p>Lead and lag assets are supposed to swap. When the alternation is disabled, or was never configured, or was turned off during a commissioning problem and never turned back on, the lead unit accumulates all the wear.</p>
<p>Lead-lag alternation (a standard reliability practice described in the <a href="https://www.epa.gov/sites/default/files/2015-09/documents/primer.pdf" rel="noopener">EPA Primer for Municipal Wastewater Treatment Systems</a>) is designed to equalize wear across redundant assets and ensure the standby unit is exercised regularly. The standby that has not run in months is the one the plant will depend on when the lead finally fails.</p>
<p>In one plant's record, return-activated-sludge pumps showed 504 run hours against 232 in the same twenty-five days, and the lifetime counters showed the gap compounding: 44,299 hours against 30,953. The lead unit reaches overhaul years earlier than it needs to while the standby sits idle.</p>
<figure class="post-fig post-dia"><svg class="dia" viewBox="0 0 960 300" role="img" aria-label="Run hours for two RAS pumps. Over 25 days the lead pump ran 504 hours against its twin's 232. Lifetime counters show 44,299 hours against 30,953.">
  <g font-family="Plus Jakarta Sans, sans-serif">
    <text x="60" y="38" font-size="13" font-weight="700" letter-spacing="2" fill="#66728a">RUN HOURS, SAME 25 DAYS</text>
    <rect x="60" y="58" width="620" height="44" rx="8" fill="#c2410c"/>
    <rect x="60" y="116" width="285" height="44" rx="8" fill="#2449e8" opacity="0.35"/>
    <text x="76" y="86" font-size="17" font-weight="700" fill="#fff">Pump A — 504 h</text>
    <text x="76" y="144" font-size="17" font-weight="700" fill="#0b1026">Pump B — 232 h</text>
    <text x="700" y="86" font-size="14" font-weight="700" fill="#c2410c">84% duty</text>
    <text x="365" y="144" font-size="14" font-weight="700" fill="#66728a">39% duty</text>
    <path d="M60 196 H900" stroke="#dfe6f3" stroke-width="2"/>
    <text x="60" y="228" font-size="13" font-weight="700" letter-spacing="2" fill="#66728a">LIFETIME COUNTERS — THE GAP COMPOUNDING</text>
    <text x="60" y="262" font-size="21" font-weight="800" fill="#c2410c">44,299 h</text>
    <text x="210" y="262" font-size="21" font-weight="800" fill="#66728a">30,953 h</text>
    <text x="370" y="262" font-size="15" fill="#66728a">— the standby is the one you will need</text>
  </g>
</svg></figure>
<p>Twenty-five days of run hours, and the lifetime counters behind them.</p>
<p>Lead-lag alternation costs nothing to enable. What makes this finding persuasive rather than theoretical is that the same plant's MBR blowers and permeate pumps <em>were</em> balanced, which marks it as an oversight on one asset rather than a design choice.</p>
<h3 id="a-shared-supply-unequally-split">A shared supply, unequally split</h3>
<p>Twin basins on a common air header should see similar dissolved oxygen at similar load. When one sits far above the other with its blower command pinned wide open, the control strategy is not the problem: the air is not arriving where it is supposed to.</p>
<p>This one is cheap to investigate and cheap to correct, which is what makes it worth finding. A valve position or a damper, not a capital project.</p>
<p>A modelled scenario for scale: two 150 hp centrifugal blowers serving twin aeration basins, each averaging 60% load. A filter-loading imbalance drives one blower 12 percentage points above its twin. At that differential, the harder-running blower consumes roughly 8 to 10 kW more than needed. At $0.085 per kWh over an 8,000-hour operating year, that is $5,400 to $6,800 in excess annual electricity on one pair. A single filter service or damper adjustment that costs a few hundred dollars eliminates the cause; the recovery is measured in weeks, not quarters.</p>
<p>Aeration typically accounts for 55 to 70 percent of a treatment plant's electricity consumption (<a href="https://www.epa.gov/sites/default/files/2016-12/documents/guide_energymanagement.pdf" rel="noopener">EPA, Ensuring a Sustainable Future: An Energy Management Guidebook for Wastewater and Water Utilities</a>). An air distribution imbalance between twin basins does not show up in the daily energy report: it shows up in blower run hours, and the cause is upstream of both. How that translates to operating cost is described in <a href="/blog/aeration-energy-cost-wastewater-plants.html">aeration energy at wastewater plants</a>.</p>
<h3 id="instruments-that-no-longer-agree">Instruments that no longer agree</h3>
<p>Two turbidimeters on the same permeate stream, one reading a median of 0.11 NTU with a 90th percentile of 0.26, the other reading 0.14 and 2.12: three and a half times the plant's posted upper limit.</p>
<p>One of them is drifting. Until you know which, compliance decisions are resting on an instrument that cannot be fully trusted. An approach that can identify which instrument shifted, without waiting for a scheduled calibration, is described in <a href="/blog/spectral-change-analysis-vs-calibrated-parameters.html">spectral change analysis versus calibrated parameters</a>. The encouraging part, in that case, was that the true value looked excellent.</p>
<p>Dissolved oxygen instruments follow the same pattern. In activated sludge, a DO sensor reading above actual holds the blower on past the setpoint; one reading below actual cuts aeration short before the biology is satisfied. The process consequence of that divergence, particularly during intermittent aeration cycles for nitrogen removal, is covered in <a href="/blog/intermittent-aeration-nitrogen-removal-do-control.html">intermittent aeration for nitrogen removal</a>.</p>
<p>When that DO sensor divergence coincides with permit limits on total nitrogen or ammonia, the compliance exposure extends beyond process efficiency. <a href="/blog/nitrogen-permit-compliance-sampling-gap.html">NPDES nitrogen permit compliance: the DMR sampling gap</a> covers how plants relying on monthly grab samples can be reporting compliant while actual effluent conditions periodically exceed the permit limit.</p>
<h2 id="why-frequent-starts-cost-more-than-running">Why frequent starts cost more than running</h2>
<p>A related pattern worth naming on its own, because it is consistently underestimated.</p>
<p>Pumps are often judged on run hours. The more expensive variable is usually starts. Every start draws locked-rotor current several times the running current, and the heat goes into the windings. Motor thermal protection guidelines, as documented in the <a href="https://www.energy.gov/sites/prod/files/2014/05/f16/pump.pdf" rel="noopener">DOE "Improving Pumping System Performance: A Sourcebook for Industry"</a>, establish typical starting limits for mid-size motors because inrush heat cannot dissipate faster than winding insulation allows. Manufacturer specifications typically put that limit at a handful of starts per hour.</p>
<p>In one record, reclaimed water pumps were logging 13 to 28 starts an hour. One asset logged 143 in a single hour: a start every twenty-five seconds. No alarm fired, because at no point was any pump outside its operating limits. It is a control deadband, and the bill arrives years later in motors, starters and contactors.</p>
<p>A modelled scenario puts numbers on that exposure. A 25 hp pump motor rated for 4 starts per hour at full load accumulates locked-rotor heat on each start equivalent to roughly 1 to 2 minutes of continuous running. At 20 starts per hour, the thermal budget that would extend winding life to 8 to 12 years is consumed in 2 to 4 years. A motor rewind at that frame size typically runs $3,000 to $6,000. Annualized across the accelerated interval: $500 to $1,000 per pump per year in a failure mode no process alarm predicted. A facility running four pumps in this pattern carries $2,000 to $4,000 per year in accelerated motor wear, before accounting for starter and contactor replacement at elevated cycle counts.</p>
<figure class="post-fig post-dia"><svg class="dia" viewBox="0 0 960 260" role="img" aria-label="Pump starts per hour. Manufacturer guidance is near six an hour. Observed range was 13 to 28, with one asset logging 143 in an hour — a start every 25 seconds.">
  <g font-family="Plus Jakarta Sans, sans-serif">
    <path d="M60 150 H900" stroke="#dfe6f3" stroke-width="3"/>
    <g>
      <path d="M100 136 V164" stroke="#137a4b" stroke-width="3"/>
      <text x="100" y="124" text-anchor="middle" font-size="14" font-weight="700" fill="#137a4b">~6/h</text>
      <text x="100" y="192" text-anchor="middle" font-size="13" fill="#66728a">guidance</text>
    </g>
    <rect x="196" y="140" width="190" height="20" rx="10" fill="#c2410c" opacity="0.25"/>
    <text x="291" y="124" text-anchor="middle" font-size="14" font-weight="700" fill="#c2410c">13–28/h</text>
    <text x="291" y="192" text-anchor="middle" font-size="13" fill="#66728a">observed range</text>
    <g>
      <circle cx="840" cy="150" r="11" fill="#c2410c"/>
      <text x="840" y="120" text-anchor="middle" font-size="19" font-weight="800" fill="#c2410c">143/h</text>
      <text x="840" y="192" text-anchor="middle" font-size="13" fill="#66728a">one asset</text>
      <text x="840" y="214" text-anchor="middle" font-size="13" font-weight="700" fill="#0b1026">a start every 25 seconds</text>
    </g>
    <text x="60" y="40" font-size="15" fill="#66728a">Every start draws locked-rotor current. Run hours are not the expensive variable.</text>
  </g>
</svg></figure>
<h2 id="how-to-look-for-it-in-your-own-data">How to look for it in your own data</h2>
<p>You do not need a platform to start. With a historian export and a spreadsheet:</p>
<ol class="post-list"><li><b>List your parallel assets.</b> Anything with a twin: pumps, blowers, basins, trains, filters.</li><li><b>Compare run hours over the same window</b>, and compare lifetime counters if you have them. A persistent ratio far from 1:1 is the signal.</li><li><b>Count starts per hour</b>, not just runtime, and compare against the manufacturer guidance for that size of machine.</li><li><b>For process variables, plot the pair on the same axis.</b> Not each against its limit: against each other. Divergence that holds for weeks is not noise.</li><li><b>Check duplicates against each other.</b> Two instruments on one stream that disagree are telling you something regardless of which is right.</li></ol>
<p>What you are looking for in every case is persistence. A transient divergence is operations. A divergence that holds for the whole record is a setting, a valve, or an instrument, and all three are fixable.</p>
<h2 id="why-this-is-where-we-start">Why this is where we start</h2>
<p>When we read a plant's history, parallel equipment is the first thing modelled, because it produces findings that are concrete, cheap to verify, and hard to argue with. Nobody has to accept a model's opinion about what dissolved oxygen should be. They only have to accept that two basins on the same air header ought to look alike.</p>
<p>More on what a historian export produces: <a href="/blog/what-25-days-of-historian-data-shows.html">what 25 days of data actually shows</a>.</p>
<p><em>Author: Alex Towfigh, AquaMesh</em></p>]]></content:encoded>
      <category>parallel equipment</category>
      <category>pump duty</category>
      <category>lead lag alternation</category>
      <category>aeration</category>
      <category>predictive maintenance</category>
    </item>
    <item>
      <title>Spectral change analysis versus calibrated parameters</title>
      <link>https://aquamesh.ai/blog/spectral-change-analysis-vs-calibrated-parameters.html</link>
      <guid isPermaLink="true">https://aquamesh.ai/blog/spectral-change-analysis-vs-calibrated-parameters.html</guid>
      <pubDate>Tue, 22 Sep 2026 09:00:00 GMT</pubDate>
      <description>UV-Vis probes measure absorbance spectra, not parameters. What that means for BOD, why calibration models drift, and when change detection is the better tool.</description>
      <content:encoded><![CDATA[<hr />
<p>title: Spectral change analysis versus calibrated parameters slug: spectral-change-analysis-vs-calibrated-parameters date: 2026-09-22 section: Measurement description: UV-Vis probes measure absorbance spectra, not parameters. What that means for BOD, why calibration models drift, and when change detection is the better tool. image: assets/starch-spectra.jpg tags: UV-Vis, spectral analysis, BOD, COD, online water quality monitoring, surrogate parameters</p>
<hr />
<p>An online water quality analyser usually arrives with a list: COD, BOD, TSS, nitrate, TOC. The list is the product. The implication is that the instrument measures each of those things the way a lab measures them, only continuously.</p>
<p>That is not how spectral measurement works, and the gap between how it is sold and how it behaves is where a lot of disappointment comes from.</p>
<h2 id="what-the-instrument-actually-produces">What the instrument actually produces</h2>
<p>A UV/Vis probe does one thing: it measures how much light the water absorbs across a range of wavelengths, many times a second. What comes back is a curve, an absorbance spectrum. That curve is the raw measurement. Everything else is interpretation.</p>
<figure class="post-fig"><img src="../assets/starch-spectra.jpg" alt="UV/Vis and NIR absorbance spectra at clean, rising, high and very high loss, a reference fingerprint across 200-1700 nm, and spectral deviation from that reference." loading="lazy" decoding="async" /><figcaption>The raw measurement: absorbance across wavelength, many times a second. Everything downstream is interpretation of this curve.</figcaption></figure>
<p>Converting that curve into "COD = 212 mg/L" requires a model that maps spectral shape to a lab value. Such models exist, they can be built, and for some parameters on some streams they are good. <a href="https://www.iso.org/standard/28624.html" rel="noopener">ISO 15839</a>, which sets performance specifications for online water quality analysers, addresses this directly: models must be validated for the specific application, and performance characteristics may vary by site. In practice, this means the conditions that worked at one plant do not transfer automatically:</p>
<ul class="post-list"><li>They are often <b>site-specific</b>. Transferability must be demonstrated; a model validated on one stream should not be assumed valid on another.</li><li>They <b>drift with composition</b>. Change the recipe, the season, the raw material or the upstream process, and the relationship between spectrum and lab value moves.</li><li>They <b>degrade quietly</b>. A calibration that has stopped being right does not announce it. It keeps producing plausible numbers.</li></ul>
<p>Without reference checks and maintenance, a parameter model can keep producing plausible numbers after its error has grown. The same silent-drift behavior applies to simpler sensors driving process control decisions. What calibration intervals look like in practice, and what deferred recalibration costs in energy and nitrogen removal performance, is covered for three sensor types: <a href="/blog/do-sensor-calibration-interval-activated-sludge.html">DO sensor calibration intervals in activated sludge</a>, <a href="/blog/ph-sensor-drift-wastewater-activated-sludge.html">pH sensor drift in activated sludge</a>, and <a href="/blog/ammonia-sensor-drift-wastewater-aeration-control.html">ammonia sensor drift in wastewater aeration control</a>.</p>
<h2 id="the-alternative-watch-the-spectrum-move">The alternative: watch the spectrum move</h2>
<p>There is a different use of the same hardware, and in a plant it is usually the more valuable one.</p>
<p>Instead of asking <em>what is the concentration</em>, ask <em>has the fingerprint changed, and what does that change correlate with</em>. Build a one-time model of what the plant's own stream normally looks like. Then watch for deviation from that reference, and tie the deviation to what the plant was doing at the time: feed rate, wash flow, machine state, cleaning records.</p>
<p>This is a materially easier question to answer well, and it is insensitive to the things that break calibrations:</p>
<ul class="post-list"><li>It does not need an absolute value, so it does not need to hold an absolute calibration</li><li>It can handle defined operating modes more transparently when each recipe or season has an appropriate reference; new conditions still need review</li><li>A drifting baseline shows up as drift rather than as a wrong number presented confidently</li></ul>
<p>{{diagram:calibration-vs-change}} Same hardware, same spectrum, two different questions. Two very different maintenance burdens.</p>
<p>What you lose is the ability to report a concentration to a regulator. What you gain is a clear operating signal whose performance can be checked against current references over time.</p>
<h2 id="so-why-calibrate-bod-at-all">So why calibrate BOD at all</h2>
<p>Because sometimes a number is the point.</p>
<p>Biochemical oxygen demand is the parameter that drives sewer surcharges, appears in <a href="https://www.ecfr.gov/current/title-40/chapter-I/subchapter-N/part-133" rel="noopener">discharge permits under 40 CFR Part 133</a>, and determines whether a load is going to cause a problem downstream. "The spectrum moved" is not an answer when the question is whether you are about to exceed a threshold that gets billed by the pound.</p>
<p>For the first industrial deployment, BOD is the parameter we plan to validate against the customer's own reference samples. The reference test is <a href="https://www.ecfr.gov/current/title-40/chapter-I/subchapter-C/part-136" rel="noopener">Standard Method 5210B, as codified in 40 CFR Part 136</a>. Their lab stays the method of record. The goal is to fill the hours between the samples the lab already takes, rather than replace them.</p>
<p>Other outputs begin as change analysis unless and until they are validated against an appropriate reference method.</p>
<h2 id="why-be-this-specific-about-it">Why be this specific about it</h2>
<p>Partly because it is true, and a claim that is not true gets found out in a technical review.</p>
<p>But mostly because the honest version is a better product story. "We measure sixteen parameters" invites a comparison against instruments from companies with thirty years of calibration work behind them, on a spec sheet, where we would lose. "We watch the whole spectrum for change and tell you what it correlates with in your plant, and we plan to validate BOD against your lab" is a different claim, and a parameter list does not answer it.</p>
<p>It also sets expectations that survive deployment. An operator who was told they were getting continuous COD and finds the number drifting stops trusting the instrument. An operator who was told they were getting an early warning that something in the water changed, and gets exactly that, keeps using it.</p>
<h2 id="what-this-means-in-practice">What this means in practice</h2>
<p>For a plant considering continuous optical measurement, three questions are worth asking any vendor:</p>
<ol class="post-list"><li><b>Which parameters are calibrated against my samples, and which are inferred?</b> The answer should be short and specific.</li><li><b>What happens to the calibration when my feedstock changes?</b> If the answer is vague, assume it drifts.</li><li><b>What is the instrument useful for if I never calibrate anything?</b> A good spectral instrument still earns its place as a change detector. A repackaged parameter list does not.</li></ol>
<h2 id="where-this-sits-in-the-stack">Where this sits in the stack</h2>
<p>Measurement is only half of it. A spectrum that moved is not an action: it becomes useful when it is correlated with plant state, attributed to a unit, and priced.</p>
<p>That is the division of labour: the probe reads the stream, the platform works out what the change means and what it is worth, and the operator decides what to do about it. What that process looks like against a real plant dataset is in <a href="/blog/what-25-days-of-historian-data-shows.html">what 25 days of historian data actually shows</a>. Working from a production historian to find cost events is the practical version of the correlation step described above.</p>
<p>That distinction between a measurement that holds against an independent reference and one that drifts undetected applies to every process sensor that drives a control decision. <a href="/blog/cooling-tower-blowdown-conductivity-sensor-drift.html">Cooling tower blowdown: what conductivity sensor drift costs in water and chemicals</a> is a direct example from a different application.</p>
<p>When the same drift problem affects parallel process units running side-by-side, the divergence appears in energy use and process outputs before it shows in any individual sensor alarm. The methodology for catching that pattern -- and what it typically costs before it surfaces -- is in <a href="/blog/why-parallel-equipment-drifts.html">why parallel equipment drifts and what it costs</a>.</p>
<p>More on what the sensor is used for by sector: <a href="/industries/food-beverage.html">food and beverage</a>, <a href="/industries/industrial-water.html">industrial water</a>, <a href="/industries/pharma-biotech.html">pharma and biotech</a>. For more on the measurement hardware, including the AquaSpectra spectral water probe and edge connectivity: <a href="/hardware.html">AquaMesh hardware</a>.</p>
<p><em>Author: Alex Towfigh, AquaMesh</em></p>]]></content:encoded>
      <category>UV-Vis</category>
      <category>spectral analysis</category>
      <category>BOD</category>
      <category>COD</category>
      <category>online water quality monitoring</category>
      <category>surrogate parameters</category>
    </item>
  </channel>
</rss>
