Why a chemical treatment program that used to work fails
Summary
A chemical treatment program that used to work usually fails from slow plant changes: production rate, raw water, equipment, operators or the process itself. Each of those changes moves one reading first, such as dose per unit of output, make-up conductivity or measured pump output. So compare current readings with the site's own records before blaming the product.
Cooling tower cycles of concentration are the ratio of dissolved solids in blowdown to make-up water, and many systems run at 2 to 4 cycles (US DOE FEMP, accessed 2026).
In a 1962-65 survey of streams feeding Lakes Superior and Michigan, hardness, alkalinity and conductivity were lowest in spring runoff and highest in late-summer low water and winter (USGS, 1968).
A metering pump should be recalibrated after every service or repair and checked at high, low and typical flow, because flow is not always linear with pump speed or stroke length (Blue-White Industries, 2025).
NIST/SEMATECH lists the Western Electric rule of 8 consecutive points on one side of the center line and a separate trend rule of 6 points in a row rising or falling (NIST/SEMATECH, accessed 2026).
This answer covers continuously dosed programs, such as cooling water, boiler water, process water and coolant treatment, that drift over weeks or months. It does not cover sudden failures from a spill, a pump trip or a wrong delivery, biological wastewater upsets, or drinking water disinfection, where regulators set the required response.
What does a chemical treatment program look like before it fails?
A chemical treatment program that is about to fail usually still looks normal on the daily log. Setpoints have not changed, product use is near budget and each reading sits inside its range. The trouble shows as a slow lean, not a jump: residuals a little lower each month, a controller that blows down a little more often, corrosion coupons that come back a little heavier, or operators topping up the feed tank more often than in the same months of earlier years.
Because no single reading crosses its limit, nobody calls it a problem until a deposit, a leak, a foam event or a quality complaint arrives. The usual reaction then is to blame the product or raise the dose, and both moves hide the cause. A higher dose can cover a pump that is losing output, or a harder raw water, for a while. It adds cost without fixing why the program stopped keeping pace.
For cooling towers, the US Department of Energy advises keeping a log of make-up and blowdown quantities, conductivity and cycles of concentration, and to "Monitor trends to spot deterioration in performance" (US DOE FEMP, accessed 2026). That trend log is the main tool for catching any dosed program before it fails.
Which slow plant changes break a chemical treatment program?
When a chemical treatment program that used to work starts to fail, five slow plant changes are the usual suspects. They are listed in the order that is cheapest to check, not ranked by how often each occurs.
Production rate
More output brings more heat load, more make-up water and more contaminant load. A dose fixed per day then falls per unit of output, even though the setpoint has not moved.
Raw water quality
Source water shifts with the season and with the source. In streams feeding Lakes Superior and Michigan, hardness, alkalinity and conductivity were lowest during spring runoff and highest in late-summer low water and winter (USGS, 1968). A new well, a blended supply or a switch to reclaimed water causes the same kind of shift in one step.
Equipment wear
Feed pumps lose output, conductivity probes foul and valves leak. A program dosed by a pump that delivers less than its setting looks like a weak product.
Operator turnover
New staff may sample at a different point or time, use a different test kit, or adjust the dose by feel. The readings then shift without any change in the water.
Process change
A new raw material, product mix, cleaning chemical or operating temperature changes what the program has to control, often without anyone telling the treatment supplier.
Which readings show a chemical treatment program drifting first?
The first readings to show a chemical treatment program drifting are the ones normalized to the plant, meaning scaled to what the plant is doing, not the raw setpoints: dose per unit of output, make-up water quality, measured pump output and residual against demand. The site check table below links each reading to the slow change it most likely points to.
Check | What you see | Points to |
|---|---|---|
Dose per unit of output (dose divided by production) | Falling while dose per day is flat | Production rate rose and the program did not scale |
Make-up conductivity or hardness against the same month in past years | Outside the site's own range for that season | Raw water change, seasonal or new source (USGS, 1968) |
Make-up to blowdown flow ratio against the conductivity ratio | The two ratios are not about the same (US DOE FEMP, accessed 2026) | Leak, unmetered draw-off, or a controller or valve fault |
Pump drawdown against the setpoint | Measured output below the setting | Pump off its calibration, for example after a change in suction lift or pressure; recalibrate (Blue-White Industries, 2025) |
Product residual against the residual expected from the feed | Residual low while the drawdown is right | Higher demand from contamination, a process change or raw water |
Corrosion coupon rate (ASTM D2688-23, 2023) | Higher than past pulls at the same rack | Inhibitor level or chemistry no longer matched to the water |
Results grouped by tester or shift | A step on the day the tester changed | Operator or test method change, not the program |
Check dose per unit of output first. When it falls while the daily dose looks steady, the plant has grown past the program. The pump drawdown, a timed measurement of what the pump actually delivers, comes next, because a pump delivering less than its setting makes every other reading look like a product or water problem.
Read the table from the top down and stop at the first row that has moved; the rows below it often move as a consequence. A fall in dose per unit of output, for example, will lower the residual and raise coupon rates later, so those rows confirm the cause rather than reveal a second one.
Record which row moved first and on what date. That date is what you will line up against deliveries, production changes and staff changes.
How do you compare current treatment readings with past records?
Chart each current reading against a baseline taken from a period when the chemical treatment program worked, using the same sampling point, test method and season. A Shewhart control chart plots the value against time with a center line at the mean of the in-control period and limits usually set at 3 standard deviations, so that "almost all of the data points will fall within these limits as long as the process remains in-control" (NIST/SEMATECH, accessed 2026).
Slow drift rarely breaks a 3-sigma limit, so use the run rules too: 8 consecutive points on one side of the center line, or 6 in a row trending up or down, signal a change even when every point is inside the limits (NIST/SEMATECH, accessed 2026).
Adding the Western Electric rules costs false alarms: NIST/SEMATECH puts them at about one in every 92 points, against one in 371 for the 3-sigma limit alone (NIST/SEMATECH, accessed 2026). So confirm a signal with a pump drawdown or a repeat test before changing the program.
Chart dose per unit of output instead of dose per day. Compare raw water readings with the same months in earlier years, so a seasonal swing is not mistaken for drift.
Keep the baseline fixed. NIST/SEMATECH lists a major process change as one valid reason to recalculate limits: that calls for a new baseline after a change you have confirmed, not a quiet reset that hides drift.
The comparison is only as good as the records, and readings that customer operators send in different formats are the first to go missing; see how to collect site readings from customers without retyping.
How do you tell plant drift from a chemical product problem?
To tell plant drift from a chemical product problem, first prove that the feed pump delivers its setting. Then line up the date the readings turned against the dates of deliveries, production changes, staff changes and raw water seasons. Work in three checks.
First, run a drawdown on the pump: Blue-White recommends pumping into a calibration cylinder over a timed period and calibrating at high, low and typical flow (Blue-White Industries, 2025). If the measured output is low, the cause is equipment, not chemistry.
Second, check the product. Compare the certificate of analysis for the current lot with earlier lots, and test a retained sample if active content or appearance looks different.
Third, check demand. If the residual measured right after the feed point is correct but falls faster across the system than it used to, something in the plant is consuming more product.
The timing tells the two apart. A product problem usually starts on the day a new lot arrives and affects every system fed from that lot. Plant drift usually starts before any delivery change, builds over weeks and follows output, season or a process change. If the turn in the chart predates the new lot, the lot is not the cause.
How do you keep a chemical treatment program from drifting again?
To keep a chemical treatment program from drifting again, tie it to the plant's own change and maintenance routines. The program is then adjusted when the plant changes, not months later. Five habits cover most cases.
Recalibrate feed pumps after every service or repair and log each calibration, so the next one shows the pump's health over time (Blue-White Industries, 2025).
Review dose per unit of output and the control charts at every service visit, not only when a complaint arrives.
Check make-up water quality at each season change and whenever the source changes.
Write the sampling point, time and test method into a procedure, so a new operator produces readings comparable with the old ones.
Add the treatment supplier to the plant's change checklist.
For processes covered by US process safety management, OSHA requires written procedures to manage changes to process chemicals, technology, equipment and procedures, except replacements in kind (US OSHA 29 CFR 1910.119, accessed 2026). Outside that scope, a one-line "notify water treatment" step in the plant's own change form serves the same purpose.
The US Department of Energy also expects treatment programs to include routine chemistry checks with regular service reports (US DOE FEMP, accessed 2026).
When should you call a water treatment specialist about a chemical treatment program?
Call a water treatment specialist about a chemical treatment program when a run-rule signal persists after the pump drawdown checks out, or when coupons, deposits or a new water source show that the chemistry itself no longer fits. Call early in four cases:
The residual stays low at a verified feed rate, which means demand has changed and the program may need a different product, not more of the same one.
Corrosion coupon rates or deposit inspections are worse than past results at the same location.
The raw water source or blend is changing, or the plant is adding reclaimed water.
A process change brings in a new contaminant, temperature or material that the original program was not designed for.
Do not raise the dose of acids, oxidizers or biocides beyond the program's design to compensate: the DOE warns that acid overdoses can severely damage a cooling system (US DOE FEMP, accessed 2026). Check any change in hazardous chemical feed against the product's safety data sheet (SDS) and the site's chemical handling rules first.
A specialist needs the history, so bring the charted readings, calibration log and change dates. In Lubinpla, a customer can enter readings in a Field Care site data request; once the supplier's team approves them, they join the project's measurement history, which is read as a table or queried through the agent rather than shown as a trend chart.
Frequently asked questions
How many past readings do you need to build a baseline for a treatment program?
Use enough routine readings from a period when the program worked to show its normal scatter, ideally covering every season. NIST/SEMATECH lists having at least 30 more data points, with no known process change, as one reason to update control limits (NIST/SEMATECH, accessed 2026).
Should you raise the dose when treatment residuals start to fall?
Not before checking why. Run a pump drawdown first, then compare the residual right after the feed point with the residual across the system. Raising the dose before those checks can hide a failing pump or a demand change, and it raises chemical cost without fixing the cause.
Is a seasonal change in raw water a treatment program failure?
A seasonal change is not a failure if the program was designed for the site's full seasonal range. It becomes one when readings move outside the range seen in the same months of earlier years, or when the program was set up in one season and never checked in the other.