A refinery and petrochemical complex finds $3.0M a year in the plant it already runs
From 150 days of the plant's own records, with no new instruments and no shutdown, AquaMesh traced 9 physical causes of avoidable water, chemical, power, fuel and product loss, plus two instrument readings worth a day of checking. Each was priced from measured quantities, with the check that confirms it.
Integrated refinery and petrochemical complex (a model)
Process
Crude and vacuum distillation, hydrocracking, aromatics, ethylene cracker, polyethylene and glycol, utilities, wastewater
Scale
about 380,000 bpd of crude · about 1.55 Mt/y of ethylene · four cooling towers, four boilers
Data used
1,896 measuring points · 150 days · 13.7M values
New hardware
None for the audit
Scale reference. Aramco reports SATORP at roughly 460,000 barrels a day, its Amiral expansion is designed around a 1.65 Mt/y ethylene mixed-feed cracker, and Sadara reports 26 integrated plants making over 3 Mt/y of chemicals and plastics. This model is sized in the same range. It is a model, not their data, and none of them is an AquaMesh customer.
Annual savings
$3.0M/yr
Range $1.9M–$4.2M across the price ranges tested. No capital spend.
Fresh water saved
1.1M m³/yr
5% of the site's raw-water draw, and the same volume off effluent treatment.
Product recovered
377 t/yr
Phosphate that was being washed to the waste stack instead of leaving as product.
Steam and power
153k t steam
13.1 GWh less electricity, and heat recovery restored on one acid train.
The plant, in 3D1,896 measuring points · 150 days of records. Press play, then click a ringed unit to see the loss it marks.
Thousands of readings, and no way to ask the right question
Your site runs from crude and vacuum distillation through the hydrocracker, aromatics, the ethylene cracker and the polyethylene lines. Four cooling towers, the steam and boiler system and the water treatment plants serve all of it. Wastewater from every area ends up at one biological treatment plant, and the air blowers there keep its bacteria alive.
Most of what this audit found sits in the utilities: cooling water, steam and wastewater air. Three of your four towers waste water or chemical in different ways. One steam area sends too little condensate home. One boiler feed pump is wearing faster than its twins, the identical units beside it. One cooler is fouling quietly behind a valve that is almost out of room.
The largest single flow of trouble runs one way. Polyethylene line 1 flushes product to the sewer at every grade change, and the wastewater plant's air supply cannot follow the load that arrives. Two instrument pairs have also split apart, so some readings you rely on need a check.
Most large process sites are in this position. The control system keeps every reading inside its limits, and every unit is individually reasonable. What it cannot do is notice that two units built to behave alike have quietly stopped doing so. That is where these losses sit, and it is why none of them announced itself.
The approach
Find the loss, find the cause, price it, confirm it
AquaMesh does not assume that software saves a fixed percentage. It starts from the plant's existing control-system and historian data and works through five steps.
01Detect
1,896 measuring points were mapped from the plant's own records. Every unit was compared with its identical twin, day by day, across 150 days. 17 symptoms surfaced.
02Diagnose
An engineering review tested each symptom against what the instrument measures and how that instrument typically misleads, to separate real process problems from instrument problems. The symptoms resolved into 9 distinct causes and 2 leads that need a field check.
03Predict the consequence
Each cause was priced from measured quantities and the plant's own unit costs, against the plant's own healthy twin rather than a design figure.
04Route an approved response
Each cause became one specific action with an owner and a timing. AquaMesh recommends; operators decide. Nothing in the plant is actuated.
05Verify the result
Each cause carries a metric and a target, tracked monthly against a baseline built from this same record and normalised for production and weather.
What the data showed
9 causes, each against its own twin
Each card shows the unit with the problem (red) against its identical twin (green) over the whole record, what the operator does about it, and how the saving is confirmed afterwards.
CAUSE 1
Cooling tower A drains about 2.0× the water it needs to
$1,148,223/yrestimated
Cooling tower A reuses its cooling water 3.0 times before draining it. Cooling tower C, on the same supply, manages 5.0. Every extra drained cubic metre is bought water, treatment chemicals and a trip through effluent treatment.
Cooling tower ACooling tower C× reuse
What the data shows
Tower A reuse
3.01×
times the water is reused before draining
Tower C reuse
4.98×
same supply water, so it is achievable
Tower A drain
190 m³/h
2.0× its twin's
Fresh water needed at twin level
496 m³/h
against 594 m³/h today
Evaporation per MW rejected
1.45 vs 1.45 m³/h
heat load 274 vs 277 MW, same weather: the difference is in the drain, not the duty
What it means
Tower A is held to a water-reuse limit that was set as a temporary measure while the fill was inspected. The inspection found the fill clean, but the limit was never revised, so the tower keeps paying for fresh water, chemicals and effluent handling it does not need. Its sister tower on the same supply shows the higher reuse can be run.
What the operator does
Raise cooling tower A's reuse target toward cooling tower C's, after your water chemist confirms the scaling limits.
How it is confirmed
Cooling tower A reuse count and fresh-water draw. Target: 5.0× reuse, about 496 m³/h of fresh water at the same cooling duty.
How the value is built
Item
Quantity
Price
$ / yr
Fresh water not bought
98 m³/h × 8,059 h
$0.95/m³
$749,514
Less water sent to effluent treatment
98 m³/h × 8,059 h
$0.40/m³
$315,585
Scale and corrosion inhibitor not used
98 m³/h × 0.0358 kg/m³ × 8,059 h
$2.6/kg
$73,423
Biocide not used
98 m³/h × 0.0030 kg/m³ × 8,059 h
$4.1/kg
$9,701
Total
$1,148,223
CAUSE 2
The ARO area returns far less condensate than comparable areas
$498,245/yrmeasured
Steam gives up its heat in the process and should come back as hot, clean condensate. Comparable areas return 88% of their heating steam; ARO returns 63%. At 76 t/h of steam that is about 19 t/h of condensate going to drain instead of back to the boilers, and the boilers have to make it up from demineralised water, re-heat it and treat it.
ARO returnCDU returnt/h
What the data shows
ARO return
63 %
of its heating steam
CDU return
88 %
comparable area
VDU return
87 %
comparable area
HCU return
89 %
comparable area
Condensate lost
19 t/h
at 76 t/h of steam
What it means
Hot, clean condensate that goes to drain takes its heat and its treated water with it. The boilers then replace it from demineralised water, which costs treatment and fuel every hour. The data shows the shortfall but not where the leak or diversion is, so a walk through the area's condensate traps and return lines is the place to start.
What the operator does
Survey the steam traps and condensate pots in the ARO area for traps blowing through or failed open, and check for condensate being routed to drain.
How it is confirmed
ARO condensate return ratio. Target: about 88% of heating steam.
How the value is built
Item
Quantity
Price
$ / yr
Demineralised boiler water not needed
19.0 t/h × 8,059 h
$2.1/m³
$321,275
Fuel to re-heat the replacement water
19.0 t/h × 262 kJ/kg ÷ 91% boiler × 8,059 h
$3.5/GJ
$154,022
Boiler-water treatment on that replacement water
19.0 t/h × 8,059 h
$0.15/t
$22,948
Total
$498,245
CAUSE 3
Cooling tower D circulation pumps are throttled against themselves
$349,540/yrmeasured
Cooling tower D circulation pumps run 4.0 pumps at a time where the comparable towers run 3, and hold the header pressure up by choking their discharge valves (the valves absorb 1.02 bar). Energy spent pushing water through a half-closed valve is heat. Measured at the pump terminals the pumps convert 80% of electrical power into pressure and flow, but 25% of that is then thrown away across the valves, against 9% on the other towers. The tower uses 0.180 kWh per cubic metre circulated against 0.148.
Tower D pumpsTower A pumpskW
What the data shows
Pumps running
4.0 vs 3
this tower against its peers
Pump valve position
37 % open
peers 86 % open
Pressure lost across valves
1.02 bar
peers 0.34 bar
Pump electrical to hydraulic
80 %
at the pump terminals
Energy per m³ circulated
0.180 kWh
peers 0.148 kWh
Power thrown across valves
1,280 kW
904 kW above the controllable minimum
What it means
Running a fourth pump follows a summer instruction that is still open. The pumps then fight their own half-closed valves to hold pressure. The extra pump and the choking are paying for pressure the tower can already get in the way its peers do, so the waste continues for as long as the instruction stands.
What the operator does
Run 3 pumps instead of 4, or trim impellers or fit a variable-speed drive, so the discharge valves open as far as the comparable towers'. Confirm the far-end header pressure stays above the users' minimum while you do it.
How it is confirmed
Cooling tower D circulation pumps: kWh per cubic metre circulated. Target: about 0.148 kWh/m³ at unchanged circulation.
How the value is built
Item
Quantity
Price
$ / yr
Pump power dissipated across throttle valves, above the minimum a control valve needs
904 kW × 8,059 h
$0.048/kWh
$349,540
Total
$349,540
CAUSE 4
Every PE1 grade change flushes hydrocarbon to the sewer
$271,687/yrmeasured
The PE1 sewer line normally carries about 83 m³/h at about 116 mg/L of organic carbon. 15 times in 150 days it jumped to 450 m³/h at up to 5,497 mg/L for about 6.7 hours. 100% of those events begin within three hours of a grade-code change on the PE1 unit, and its flush drain valve is open each time; 100% of the grade changes produce one. The carbon that leaves is product you made. At this rate about 444 t/yr of hydrocarbon goes to the sewer, 1,095 t/yr of chemical oxygen demand lands on the wastewater plant, and it arrives in a few hours at a time. Each slug pulls dissolved oxygen in the aeration basins down to 0.2 mg/L and pushes effluent ammonia to 13.3 mg/L.
PE1 branch organic carbon (daily average)mg/L
What the data shows
Events in the record
15
100% of grade changes
Organic carbon, measured
168.1 t
11.2 t per event
Other slugs in the record
4 events
1.2 t of carbon with no repeating cause, so not priced
Hydrocarbon per year
444 t
376 t carbon × 1.18
Chemical oxygen demand per year
1,095 t
2.9 g of oxygen demand per g of carbon, as the plant measures it
Dissolved oxygen after a slug
0.2 mg/L
against 4.2 otherwise
Effluent ammonia, peak
13.3 mg/L
against 1.0 otherwise
What it means
The slop line that would catch these flushes has been out of service, so every grade change goes to the sewer. You lose product, and the wastewater plant takes the shock of a large load in a few hours. The cause is a steady one tied to grade changes, so it will keep happening at every one until the flush has somewhere else to go.
What the operator does
Route the PE1 grade-transition flush to slop recovery instead of the sewer, and fit a UV or total-organic-carbon trip on the branch so a flush can never reach the treatment plant unnoticed.
How it is confirmed
PE1 sewer-branch organic carbon during grade changes. Target: no excursion at the next grade change; flush volume recovered at the slop tank.
How the value is built
Item
Quantity
Price
$ / yr
Hydrocarbon recovered instead of lost (85% of 444 t/yr)
376.2 t C/yr × 1.18 t/t × 0.85
$720/t
$271,687
Total
$271,687
CAUSE 5
Cooling tower B's controlling conductivity probe has slid out of calibration, so the tower drains water it should keep
$252,879/yrmeasured
Cooling tower B has two conductivity probes on the same water. They agreed at the start of the record and have since moved 20% apart. Independently, the water balance (water in, water drained, water evaporated) and the incoming water's conductivity both say the second probe is right. The first is reading about 20% off, and it is the one the tower's own reuse calculation and drain control follow, so the plant displays 5.3× reuse while the tower really reuses its water 4.4 times. The tower drains extra water to hold a number that is not real.
The tower controls its drain from the probe that has slid, so the error is not just on a screen. It drains water to hold a number that is not real. The maintenance record would point at the other probe, but the water balance and the incoming water both disagree. A bench check of both probes will settle which one to trust.
What the operator does
Calibrate or replace the controlling probe on cooling tower B against a bench conductivity reading, and have the drain controller read the average of the two probes.
How it is confirmed
Cooling tower B: probe agreement and measured reuse. Target: probes within 2%, measured reuse back to 5.0×.
How the value is built
Item
Quantity
Price
$ / yr
Fresh water no longer drained
21.8 m³/h × 8,059 h
$0.95/m³
$166,710
Less water to effluent
21.8 m³/h × 8,059 h
$0.40/m³
$70,194
Treatment chemical not used
21.8 m³/h × 0.0350 kg/m³ × 8,059 h
$2.6/kg
$15,976
Total
$252,879
CAUSE 6
Cooling tower C is still dosing scale and corrosion inhibitor at the level set for a bad week
$238,157/yrmeasured
Cooling tower C used 35 g of inhibitor per cubic metre of fresh water at the start of the record, the same as its peers. The fresh water's hardness then rose from about 118 to 251 mg/L between 2026-05-17 and 2026-05-23, and all the towers' doses were raised. The hardness came back 124 days before the record ends and the other towers' doses came back with it. Cooling tower C's did not. It now doses 58 g/m³ against 35 g/m³ for its peers. Its corrosion rate (1.07 mils per year) is no better than theirs (1.15), so the extra chemical buys nothing.
Cooling tower CCooling tower Akg/h
What the data shows
Cooling tower C dose, start
35 g/m³
of fresh water
Cooling tower C dose, end
58 g/m³
65% above its peers
Peers, end
35 g/m³
back to normal
Fresh-water hardness
118 → 251 → normal
excursion 2026-05-17 to 2026-05-23
Corrosion rate
1.07 vs 1.15
mils per year; no better than the peers
What it means
When the fresh water went hard, all four towers were dosed up, and the other towers were put back when it passed. Tower C was left out of that step, so it is a carry-over from the operations log and not a need of the tower. Its corrosion readings show no benefit from the extra chemical, and it will keep being bought until the dose is reset.
What the operator does
Return cooling tower C's dose to its peers' level, and add a rule that any temporary dose increase carries an end date.
How it is confirmed
Cooling tower C inhibitor dose per cubic metre. Target: about 35 g/m³, with corrosion rate unchanged.
How the value is built
Item
Quantity
Price
$ / yr
Inhibitor no longer over-dosed
(57.7 − 35.0) g/m³ × 501 m³/h × 8,059 h
$2.6/kg
$238,157
Total
$238,157
CAUSE 7
The aeration blowers deliver the same air whether the plant needs a little or a lot
$156,327/yrmeasured
The wastewater plant's air supply varies by only 0.6% while the oxygen the bacteria need swings between about 868 and 1,283 kg/h. The result is dissolved oxygen that averages 4.0 mg/L, about double the 2.0 mg/L that is usually enough for the biology, for most of the time, and a sag to 1.1 mg/L when a heavy load arrives. Air that is not used is paid for in blower power; air that is missing when the load arrives puts treatment at risk. Letting the air follow the oxygen demand, with a mixing floor of half today's flow, would use 404 kW less on average.
Air deliveredBlower powerNm³/h
What the data shows
Air delivered
46,002 Nm³/h
varies only 0.6%
Air the load needs (10th–90th %)
23,001–32,205 Nm³/h
for oxygen demand 868–1,283 kg/h
Dissolved oxygen
4.0 mg/L
low point 1.1 mg/L
Blower power
931 kW
20.2 kW per 1,000 Nm³/h
Saved at the load-following level
404 kW
net of the 3% of the time the plant is short of air today
Effluent ammonia, peak
13.3 mg/L
when the air could not follow a heavy load
What it means
The blowers were set to a fixed speed with the oxygen controller in manual, so they cannot follow the load. Most of the time you pay for air the bacteria do not use. When a heavy load arrives, the biology runs short of oxygen and treatment is knocked back. Left alone, every large slug will repeat that.
What the operator does
Put the blowers on a dissolved-oxygen cascade targeting 2.0 mg/L with a mixing floor, and review the load-following envelope with your process engineer before the change.
How it is confirmed
Blower power and dissolved oxygen. Target: dissolved oxygen about 2.0 mg/L, blower power following the load.
How the value is built
Item
Quantity
Price
$ / yr
Blower power saved by following the oxygen demand
404 kW × 8,059 h
$0.048/kWh
$156,327
Total
$156,327
CAUSE 8
Boiler feed pump 2 is wearing out compared with its twins
$87,740/yrmeasured
Boiler feed pump 2 does the same job as the pump beside it. Measured at the terminals it now turns 11% less of its electrical power into pressure and flow than they do, down from level with them at the start of the record. That is accompanied by a drive-end vibration that has grown to 2.2× its twins' level and a bearing running 22 °C hotter. Wear rings and bearings failing together look exactly like this, and it ends in a trip of a machine the steam system depends on.
Pump 2Twin pumpkW
What the data shows
Efficiency vs its twins
100 → 89 %
ratio, start to end
Vibration, drive end
4.7 vs 2.1 mm/s
125% relative change
Bearing temperature
88.4 vs 66.4 °C
33% relative change
Motor current
210.0 vs 187.5 A
12% relative change
Extra power
227 kW
at the end of the record
What it means
The records point to a real mechanical fault, not a bad instrument, because vibration, bearing heat, oil and current all moved together on this pump only. It is never idle, since it keeps running while its twins share the rest of the duty, and a deferred service on one of them adds to its load. A lube-oil sample asked for a resample, and no follow-up is recorded. If it is left, the likely end is a trip of a pump the steam system depends on.
What the operator does
Plan pump 2's overhaul at the next boiler-feed window: bearings, wear rings and a vibration spectrum first. Do it while its twin carries the duty.
How it is confirmed
Pump 2 efficiency, vibration and bearing temperature. Target: efficiency level with its twin, vibration back to its twin's.
How the value is built
Item
Quantity
Price
$ / yr
Extra power against a healthy twin at the same duty
227 kW × 8,059 h
$0.048/kWh
$87,740
Total
$87,740
CAUSE 9
Cooler E4112 is fouling, and nothing will alarm until its valve runs out of travel
$34,941/yrmeasured
Cooler E4112's heat-transfer ability is 37% below what a clean cooler gives at the same cooling-water flow, while 6 comparable coolers on the site's cooling water held steady. The controller has been hiding it by opening the cooling-water valve from 38% to 98%, so the process outlet temperature has stayed at its target and no alarm has fired. With the valve at the end of its travel the next stage is the outlet temperature climbing, which makes the compressor behind it use more power. To deliver today's duty it now pulls 432 m³/h more cooling water than a clean cooler would.
E4112 cooling-water valveE4112 process outlet%
What the data shows
Heat-transfer ability
95 kW/K
a clean cooler at this flow gives 150 kW/K (-37%)
Comparable coolers
within ±0.0 %
6 other coolers on the site's cooling water
Cooling-water valve
38 → 98 %
the controller compensating
Process outlet temperature
43.0 → 45.6 °C
target 43.0 °C; it has begun to climb once the valve ran out of travel
Cooling water drawn
983 m³/h
a clean cooler needs 551 m³/h
Fouling resistance
3.9 ×10⁻³ K/kW
extra resistance to heat flow from the fouling layer
What it means
The controller is hiding the fouling by opening the valve wider, so nothing alarms while the margin runs out. Once the valve has no travel left, the outlet temperature climbs and the compressor behind it works harder. It is also drawing far more cooling water than a clean cooler would, which may limit the cooler's neighbours on the same water. A cleaning planned now is cheaper than one forced later.
What the operator does
Plan a clean of E4112 at the next opportunity, and check it for waterside scale or biofouling. Add a fouling-factor trend so the next one is seen at 10%, not 30%.
How it is confirmed
E4112 heat-transfer ability and cooling-water valve position. Target: heat-transfer ability back near its start-of-record level; valve back to mid-range.
How the value is built
Item
Quantity
Price
$ / yr
Extra cooling-water pumping
432 m³/h × 0.152 kWh/m³ × 8,059 h
$0.048/kWh
$25,376
Extra compressor power from the warmer outlet
3,054 kW × 2.6 K ÷ 319 K × 8,059 h
$0.048/kWh
$9,566
Total
$34,941
Two leads are not counted. They are shown because they are worth a day of checking, and kept out of the total until a field check confirms them.
LEAD
Two organic-carbon analyzers at the wastewater plant inlet have separated by 13.3%
$0/yrmeasurednot counted in the total
The two organic-carbon analyzers at the wastewater plant inlet agreed at the start of the record and now differ by 13.3%. It is a steady slide and not a process event, because a real change in the stream would move both. An independent estimate, sum of the sewer branches (flow-weighted organic carbon), agrees with analyzer A to within 2.9% and with analyzer B only to 9.1%, so analyzer B is the one to check. This matters because a reading that slides low understates the organic load that reaches the biology.
analyzer Aanalyzer B
What the data shows
Pair ratio, start → end
1.000 → 1.134
steady slide
analyzer A against the estimate
+2.9 %
Sum of the sewer branches (flow-weighted organic carbon)
analyzer B against the estimate
−9.1 %
the one that is wrong
What it means
A steady slide means one analyzer has gone out of true, and the independent estimate points to analyzer B. If B is the one used for operating decisions, you are underestimating the organic load that reaches the biology. A bench check of B against a known sample would confirm it. The data cannot say when it began to slide.
What the operator does
Calibrate or replace analyzer B against a bench or prover reference before anything uses it for control, billing or accounting.
How it is confirmed
The two readings and the independent estimate. Target: the pair within 1% of each other and of the estimate.
How the value is built
Item
Quantity
Price
$ / yr
Total
$0
LEAD
Two polyethylene export meters have separated by 1.0%
$0/yrmeasurednot counted in the total
The two polyethylene export meters agreed at the start of the record and now differ by 1.0%. It is a steady slide and not a process event, because a real change in the stream would move both. An independent estimate, sum of the extruder pellet meters, agrees with meter A to within 0.1% and with meter B only to 1.1%, so meter B is the one to check. If this were the metering that bills product, a gap of this size is a revenue-accounting question, not just an instrument one.
meter Ameter B
What the data shows
Pair ratio, start → end
1.000 → 0.990
steady slide
meter A against the estimate
+0.1 %
Sum of the extruder pellet meters
meter B against the estimate
+1.1 %
the one that is wrong
What it means
The gap is small, but it has a clear direction and it matches no real change in the product stream. The independent estimate points to meter B. If this is the meter that records product leaving the site, the gap is a question for accounting as well as for instruments, and a check against the pellet meters would settle it.
What the operator does
Calibrate or replace meter B against a bench or prover reference before anything uses it for control, billing or accounting.
How it is confirmed
The two readings and the independent estimate. Target: the pair within 1% of each other and of the estimate.
How the value is built
Item
Quantity
Price
$ / yr
Total
$0
The value, in detail
$3.0M a year, and where each dollar comes from
The same $3.0M two ways: by cause, and by the kind of cost it removes. Every priced line belongs to exactly one category, so no dollar is counted twice.
By cause
Cooling tower A drains about 2.0× the water it needs to$1.1M
The ARO area returns far less condensate than comparable a…$498k
Cooling tower D circulation pumps are throttled against th…$350k
Every PE1 grade change flushes hydrocarbon to the sewer$272k
Cooling tower B's controlling conductivity probe has slid …$253k
Cooling tower C is still dosing scale and corrosion inhibi…$238k
The aeration blowers deliver the same air whether the plan…$156k
Boiler feed pump 2 is wearing out compared with its twins$88k
Cooler E4112 is fouling, and nothing will alarm until its …$35k
By kind of cost
Cost removed
Annual change
Of what the plant uses
$ / yr
Comes from
Water consumption
1,117,434 m³
5% of 23,924,589 m³
$1,237,499
Cooling tower A, Cooling tower B's controlling conductivity probe has slid out of calibration, so the tower, The ARO area returns far less condensate than comparable areas
Wastewater sent to treatment
964,446 m³
—
$385,778
Cooling tower A, Cooling tower B's controlling conductivity probe has slid out of calibration, so the tower
Treatment chemicals
128,349 kg
17% of 774,273 kg
$360,205
Cooling tower A, Cooling tower B's controlling conductivity probe has slid out of calibration, so the tower, Cooling tower C, The ARO area returns far less condensate than comparable areas
Electrical energy
13,095 MWh
1% of 956,321 MWh
$628,548
Cooling tower D circulation pumps are throttled against themselves, Boiler feed pump 2, The aeration blowers deliver the same air whether the plant needs a little or a lot, Cooler E4112
Steam and heat recovered
152,988 t
—
$154,022
The ARO area returns far less condensate than comparable areas
Production and yield
377 t hydrocarbon
—
$271,687
Every PE1 grade change flushes hydrocarbon to the sewer
Overall annual saving
$3,037,739
"Of what the plant uses" compares each saving with the quantity measured in this record, scaled to a year of 8,059 operating hours. Electrical energy is set against the metered load in the data, not the whole site.
The water balance still closes
A saving that does not respect conservation of water is not a saving. For the largest water cause, evaporation is fixed by the heat the tower has to reject, so only the drained water, and the fresh water that replaces it, can change.
Tower A, per hour
Today
At the twin's reuse
Change
Evaporation, set by the heat the tower must reject
397 m³/h
397 m³/h
unchanged
Drift loss from the tower
7.2 m³/h
7.2 m³/h
unchanged
Water drained to keep salts down
190 m³/h
92 m³/h
−98 m³/h
Fresh water drawn
594 m³/h
496 m³/h
−98 m³/h
Times the water is reused
3.0×
5.0×
matches its twin
Check: fresh water = evaporation + drift + drain
594 m³/h
496 m³/h
the balance closes
How confident to be
62% of the total rests on quantities read directly from the plant's instruments, such as concentrations, flows and power. The rest uses a physical assumption, such as the tower's drift loss. Prices are the plant's own unit costs; the table shows how far the total moves if each is wrong by the range tested.
Price
Used
Range tested
Total moves by
Water price
$0.95/m³
$0.57–1.33
−$630k to +$630k
Electricity price
$0.048/kWh
$0.034–0.062
−$189k to +$189k
Effluent treatment cost
$0.40/m³
$0.20–0.60
−$193k to +$193k
Hydrocarbon price
$720/t
$540–900
−$68k to +$68k
Steam and fuel value
$3.5/GJ
$2.10–4.90
−$71k to +$71k
The biggest single uncertainty is the water price. It drives 52% of the total. The quantities are measured either way; only their price is in question, and one invoice settles it.
What is not counted
The headline excludes the two instrument readings that need a field check, any benefit from the instrumentation below, avoided failures, operator time and the effect of cleaner operation on downstream production. All are real; none is claimed.
The sensor
Where the plant is blind, a probe closes the gap
AquaSpectra™ smart probe
Everything above came from instruments the plant already owns. What those instruments cannot do is act on what they see. The organic-carbon analyzer on the PE1 sewer branch records every flush, and nothing is connected to it. AquaSpectra sits in the pipe and reads the water's full spectral fingerprint continuously. Hydrocarbons absorb strongly in the ultraviolet, so a flush is recognised as it starts, and the same reading can close the sewer valve and open the slop line. It also learns what normal looks like on that branch, so the next unplanned release, from a source nobody has found yet, shows up too.
Capital$25,000per measuring point, to be firmed with a quote
Loss it guards$272k/yrthe recovery in the finding above, already counted there
Warning timeMomentsagainst the end of a five-hour flush
Counted in the total$0so the same dollar is never counted twice
What the plant can see today, and what the sensor adds
Cooling-water returns
The plant sees today
conductivity, pH and chlorine, none of which would show a hydrocarbon leak early.
AquaSpectra adds
It watches for a hydrocarbon or process-chemical leak into the circuit from any cooler, as a change in the water's fingerprint, in hours.
Early warning · $80,000 capital · counts no savings
PE1 sewer branch
The plant sees today
an organic-carbon analyzer that records every slug, with nothing connected to act on it.
AquaSpectra adds
It watches for hydrocarbon in the water as a flush starts, as a change in its light-absorption fingerprint, whatever the source.
Protects $272k/yr · $25,000 capital · counts no savings
Wastewater plant inlet
The plant sees today
a pair of organic-carbon analyzers, one of which has drifted, and a daily laboratory measure of oxygen demand.
AquaSpectra adds
It watches for any organic slug reaching the biology, from any unit, before the equalisation tank mixes it away.
Early warning · $25,000 capital · counts no savings
Condensate return headers
The plant sees today
conductivity on each return, which sees salts and not oil.
AquaSpectra adds
It watches for hydrocarbon or process chemical getting into returned condensate, before it reaches the boilers.
Early warning · $30,000 capital · counts no savings
One grade change, hour by hour
This is the largest of the 15 flushes in the record, exactly as the plant's own instruments recorded it. The organic carbon on the branch jumps by a factor of 51. Hours later, after the equalisation tank has mixed the slug into the whole plant's flow, the dissolved oxygen in the aeration basins collapses and ammonia escapes into the treated water. A reading at the branch is the only place the cause is visible.
grade changeflush to the sewer
What it does and does not claim. AquaSpectra's BOD reading is calibrated to your own lab results. For everything else it watches the whole spectrum for change and learns, from your lab results, what that change means at your plant. It is a site-calibrated early warning, not a certified compliance method, and it does not replace any test your permit specifies.
Making it stick
From finding to money on the bill
Nine of the causes are settings or planned maintenance that the plant already schedules. The value arrives as each is done, and is confirmed against a baseline built from this same record, normalised for production and weather.
Cause
Type
When
$ / yr
Share
Cooling tower A drains about 2.0× the water it needs to
Setting change
Weeks 1–4, after the water chemist signs off the limits
$1,148,223
38%
The ARO area returns far less condensate than comparable areas
Planned maintenance
Weeks 1–6: trap survey, then repairs
$498,245
16%
Cooling tower D circulation pumps are throttled against themselves
Setting change
Weeks 1–3, with operations
$349,540
12%
Every PE1 grade change flushes hydrocarbon to the sewer
Setting change
Next grade change: route the flush; trip within 4 weeks
$271,687
9%
Cooling tower B's controlling conductivity probe has slid out of calibration, so the tower drains water it should keep
Planned maintenance
Week 1: a bench check settles it
$252,879
8%
Cooling tower C is still dosing scale and corrosion inhibitor at the level set for a bad week
Setting change
Week 1, with the water chemist
$238,157
8%
The aeration blowers deliver the same air whether the plant needs a little or a lot
Setting change
Weeks 2–6: controls change, then confirm
$156,327
5%
Boiler feed pump 2 is wearing out compared with its twins
Planned maintenance
Plan within 4 weeks; vibration spectrum this week
$87,740
3%
Cooler E4112 is fouling, and nothing will alarm until its valve runs out of travel
Planned maintenance
Plan within 8 weeks; trend weekly
$34,941
1%
Two organic-carbon analyzers at the wastewater plant inlet have separated by 13.3%
Field check
Week 1
$0 (not counted)
—
Two polyethylene export meters have separated by 1.0%
Field check
Week 1
$0 (not counted)
—
Confirmed by
Target
Cooling tower A reuse count and fresh-water draw
5.0× reuse, about 496 m³/h of fresh water at the same cooling duty
ARO condensate return ratio
about 88% of heating steam
Cooling tower D circulation pumps: kWh per cubic metre circulated
about 0.148 kWh/m³ at unchanged circulation
PE1 sewer-branch organic carbon during grade changes
no excursion at the next grade change; flush volume recovered at the slop tank
Cooling tower B: probe agreement and measured reuse
probes within 2%, measured reuse back to 5.0×
Cooling tower C inhibitor dose per cubic metre
about 35 g/m³, with corrosion rate unchanged
Blower power and dissolved oxygen
dissolved oxygen about 2.0 mg/L, blower power following the load
Pump 2 efficiency, vibration and bearing temperature
efficiency level with its twin, vibration back to its twin's
E4112 heat-transfer ability and cooling-water valve position
heat-transfer ability back near its start-of-record level; valve back to mid-range
What instrumentation adds
The full instrumentation package
Everything above came from data the plant already collects. Instrumentation comes after, and only where a costly condition cannot already be seen. Items that are early warning rather than savings are shown with no value, so the payback is not flattered by them.
Item
Capital
$ / yr
Why
Water-chemistry check before raising reuse
$12,000
—
Raising a tower's reuse target concentrates the dissolved salts in its water, so scaling and corrosion limits have to be checked first. Its twin already runs at the higher level on the same supply, which is strong evidence it can be done. A corrosion-coupon rack and a cross-check of the conductivity and pH probes turns that evidence into something your water chemist can sign off.
AquaSpectra on the cooling-water returns
$80,000
—
Raising a tower's reuse concentrates whatever gets into its water, and a hydrocarbon leak in any one of dozens of coolers feeds the biology and fouls the fill. Your instruments would not see one until the water smelled. AquaSpectra on each return flags it in hours and tells you which tower, which is what makes running at higher reuse safe.
A third conductivity probe on each cooling tower, with the drain controller reading the middle value
$18,000
—
Two probes can tell you they disagree but not which one is wrong. It took a water balance to settle it here. A third probe on each tower, read by the controller as the median of three, makes the next slide harmless on its own and flags the odd one out for calibration.
AquaSpectra on the PE1 sewer branch, with a trip to the slop line
$25,000
—
Routing the flush to slop recovery is a procedure, and a procedure is only as reliable as the next shift. AquaSpectra reads the branch water's full absorption spectrum continuously. Hydrocarbons absorb strongly in the ultraviolet, so a slug is recognised as it starts, and the same signal that raises the alarm can close the sewer valve and open the slop line. It catches the next unplanned release too, from any source on that branch.
AquaSpectra at the wastewater plant inlet
$25,000
—
The flush found here is the one that repeats. The next one will come from a drain or an exchanger nobody has looked at yet. A spectral reading at the inlet sees the whole plant's water at once, learns what normal looks like on your site, and flags a change in hours, with the branch analyzers pointing to where it came from.
AquaSpectra on the condensate return headers
$30,000
—
Fixing the traps brings more condensate back to the boilers, so the quality of what comes back matters more. A tube leak in a reboiler puts hydrocarbon into condensate that conductivity cannot see, and boilers do not tolerate it. A spectral reading on each header catches it at the source.
Dissolved-oxygen cascade on the blowers, with new probes in the basins
$35,000
—
The blowers were set to a fixed speed after the oxygen controller hunted. Hunting is a tuning problem, and the cure is a slower loop with a floor, not a fixed speed. New optical probes (which hold calibration better than membrane probes) and a cascade that moves air with the oxygen demand turn the saving in the aeration finding into a setting.
Vibration and motor-current monitoring on the boiler feed pumps, cooling-water pumps, aeration blowers
$70,000
—
A boiler feed pump's wear was visible only because it was compared against its twin. Permanent vibration and current monitoring turns that into a direct reading of bearing and wear-ring condition, with a spectrum, before a failure forces a boiler off. It also gives the oil-analysis alert a number to stand next to.
Instrumentation package
$295,000
$0
A caution we would rather state plainly. AquaSpectra's BOD reading is calibrated to your own lab results. For everything else it watches the whole spectrum for change and learns, from your lab results, what that change means at your plant. It is a site-calibrated early warning, not a certified compliance method, and it does not replace any test your permit specifies.
The audit
The full report
The audit is the working behind every figure above: each finding, the evidence for it, how it was priced, and the check that confirms it.
Your actual rate, including treatment and delivery
Effluent treatment
$0.40/m³
$0.20–0.60
Depends on how much of your treatment cost scales with volume
Electricity
$0.048/kWh
$0.034–0.062
Industrial tariff; demand charges would change it
Demineralised water
$2.1/m³
$1.26–2.94
Cost of making it: raw water, membranes, regeneration chemicals and power
Fuel gas
$3.5/GJ
$2.10–4.90
What it would otherwise earn if not burned in your boilers
Hydrocarbon lost to sewer
$720/t
$540–900
Realised value of light naphtha and diluent, not a market quote
Operating hours
8,059 h/yr
fixed
92% availability, applied to every figure
Quantities are read from 1,896 measuring points over 150 days. The record covers one season, and cooling load and water use change through the year, so a full year of records would tighten every figure. Savings are ceilings that assume each fix is made and kept; they are not netted against the cost of the fixes, which is mostly labour and downtime the plant already plans for. Findings indicate; they do not prove, and each names the check that settles it. This is an example case study: the plant and its figures illustrate the method, and no named company is an AquaMesh customer.
Bring your own history.
Send a historian export, as far back as it easily goes and every tag. You get an audit like this one, priced from your own numbers, and you see the value before you buy anything.
