Recapturing Process Water Under Variable Load: Bullmark’s Adaptive Control Protocol

Process water recapture systems face a fundamental challenge: production loads rarely stay constant. When flow rates fluctuate by 40% or more within a single shift, traditional fixed-setpoint controls waste energy, overflow holding tanks, or send partially treated water to drain. This article presents Bullmark’s adaptive control protocol—a decision framework for engineers and plant managers who need to maintain high recapture rates despite erratic demand.

We assume you already understand the basics of recirculation loops, blowdown, and conductivity-based control. What we cover here is how to make those systems respond intelligently when the incoming load changes faster than a human operator can react. The protocol is built around three core principles: measure true demand in near-real time, predict short-term trends, and adjust recapture setpoints proactively rather than reactively.

Who Must Choose and by When

This protocol is designed for facilities that already have a process water recapture system in place but are seeing diminishing returns as production variability increases. Typical candidates include food and beverage plants with multiple product changeovers, chemical batch processors, and metal finishers with fluctuating rinse tank usage. If your recapture rate has dropped below 70% despite adequate hardware capacity, or if your holding tanks frequently overflow during peak production, you are the audience for this protocol.

The decision window is often driven by a capital project or a sustainability target. Many sites face a choice: upgrade the control system now, or accept higher water and sewer costs for another year. A common trigger is a 10% annual increase in water utility costs, which makes the payback period for a control upgrade shrink from three years to eighteen months. Another trigger is a regulatory deadline—for example, a new discharge permit limit that forces tighter control of blowdown volume.

We recommend starting the evaluation process at least six months before the target implementation date. This allows time for data collection (flow and conductivity trends over at least three production cycles), vendor selection if you are not building the control logic in-house, and a phased commissioning that avoids disrupting production. Waiting until the last minute often leads to rushed tuning and suboptimal performance.

The protocol itself is vendor-agnostic. It can be implemented on most modern PLCs with analog input cards and a PID or model-predictive control block. The key is not the hardware but the logic: how you define the setpoint and how fast you allow it to change. We will detail that logic in the implementation section.

Who Should Not Use This Protocol

If your production load is essentially constant (variation less than 10% day-to-day), a simple fixed-setpoint controller with a well-timed blowdown schedule will likely suffice. Similarly, if your recapture system is undersized for peak flow—meaning you can never capture more than 60% even at steady state—fix the hydraulic bottleneck first. Adaptive control cannot compensate for a pipe that is too small.

Option Landscape: Three Approaches to Variable-Load Control

When production flow varies, operators typically choose among three control strategies. Each has a different cost, complexity, and responsiveness profile. We describe them here without naming specific vendors, focusing on the logic and hardware requirements.

Time-Based Scheduling

The simplest approach: use a timer to change recapture setpoints at fixed times of day. For example, from 8:00 to 10:00 the setpoint is 80% recapture; from 10:00 to 12:00 it drops to 60% because a high-flow product is running. This works well when production schedules are predictable and repeat daily. The downside is that it cannot respond to unscheduled changes—a line stoppage or a rush order will throw off the timing. It also requires manual updates whenever the production schedule changes, which can be every week in some facilities.

Hardware requirements are minimal: a PLC with a real-time clock and a few setpoint registers. Implementation cost is low, but the labor cost to maintain the schedule can add up. We have seen plants where the schedule file grew to 200 lines and operators stopped updating it after six months, effectively reverting to fixed control.

Flow-Paced Control

Here, the recapture setpoint is a function of the instantaneous influent flow rate. A flow meter upstream of the recapture system sends a 4-20 mA signal to the PLC, which adjusts the target recapture percentage using a linear or piecewise curve. For example: at 100 gpm, target 90% recapture; at 300 gpm, target 70% recapture (to avoid overloading the treatment stage). This approach responds immediately to flow changes and does not require a schedule.

The main challenge is choosing the right curve. If the slope is too steep, the system may oscillate—a sudden drop in flow causes the setpoint to jump up, which increases recirculation and may cause the level in the holding tank to drop too fast, triggering a makeup water addition. If the slope is too shallow, you lose the benefit of adaptive control. Tuning the curve typically requires a week of data logging and iterative adjustments.

Flow-paced control is a good middle ground for facilities with moderate flow variation (20–50% range) and a reasonably stable relationship between flow and contaminant load. It does not account for changes in contaminant concentration independent of flow—for example, a batch with higher chemical oxygen demand (COD) at the same flow rate.

Predictive Adaptive Control

This is the most advanced option and the core of Bullmark’s protocol. Instead of reacting to flow alone, the controller uses a short-term forecast (15–30 minutes ahead) based on historical patterns and real-time inputs such as production schedule data, tank levels, and conductivity trends. The setpoint is adjusted proactively: if the model predicts a surge in flow 20 minutes from now, the controller begins to lower the recapture target gradually, avoiding a sudden overflow event.

Implementation requires a PLC or edge computer capable of running a simple regression or time-series model. Many modern PACs (programmable automation controllers) have built-in math functions that can handle a moving-average forecast. The model does not need to be a neural network—a linear regression on the last 60 minutes of flow data, combined with a schedule feed, often performs well enough.

The trade-off is complexity. The model must be trained on historical data and retuned periodically (every few months or after major process changes). If the model drifts, the controller may make worse decisions than a simple flow-paced loop. We recommend this approach only for sites with a dedicated automation engineer or a strong relationship with a system integrator.

Comparison Criteria Readers Should Use

Choosing among these three strategies requires evaluating your facility against five criteria. We list them in order of importance based on our experience working with dozens of recapture system retrofits.

1. Predictability of Production Schedule

If your production schedule is fixed and repeats weekly, time-based scheduling is viable. If it changes daily or hourly, you need either flow-paced or predictive control. Rate your schedule on a scale of 1 (completely unpredictable) to 5 (identical every day). Below 3, skip time-based.

2. Flow Variation Magnitude

Measure the ratio of peak flow to average flow over a typical day. If the ratio is below 1.5, fixed control may suffice. Between 1.5 and 2.5, flow-paced control is usually adequate. Above 2.5, predictive control offers significant advantages because the system must handle extreme swings without oversizing the holding tank.

3. Correlation Between Flow and Contaminant Load

In some processes, flow and contaminant concentration are tightly linked (e.g., rinse water in a plating line where the drag-out rate is proportional to the number of parts). In others, concentration varies independently (e.g., a batch reactor where one batch has high COD and the next low, both at the same flow). If the correlation is weak, flow-paced control will miss important changes. Predictive control can incorporate conductivity or TDS sensors to adjust for load independent of flow.

4. Available Automation Expertise

Be honest about your team’s skill level. If you have no one who can write a PID loop from scratch, time-based or a simple flow-paced curve (which can be implemented with a function block) is safer. Predictive control requires someone comfortable with math blocks and model validation. Overreaching on complexity is a common cause of project failure.

5. Tolerance for Oscillation

Adaptive controllers can introduce cycling if not tuned properly. If your process is sensitive to sudden changes in recirculation rate (e.g., because it affects temperature or chemical dosing), you may prefer a slower, more conservative tuning. Time-based control inherently avoids oscillation because setpoints change only at scheduled times. Flow-paced and predictive control can be dampened with rate limiters and deadbands.

Trade-Offs Table and Structured Comparison

To help you weigh the options side by side, we have compiled a comparison across seven dimensions. Use this table as a quick reference during your team’s decision meeting.

One nuance not captured in the table: predictive adaptive control can also reduce chemical consumption if you are dosing coagulants or pH adjusters in the recapture loop. By anticipating load changes, the controller can adjust chemical feed rates proportionally, avoiding overfeed during low-load periods. This is a secondary benefit that may tip the scales for facilities with high chemical costs.

Another trade-off is the risk of model failure. If the production process changes fundamentally—for example, a new product line with different water usage patterns—the predictive model will need retraining. During the retraining period (which can take days to weeks), the controller may perform worse than a simple flow-paced loop. We recommend having a fallback mode that reverts to flow-paced control if the model’s prediction error exceeds a threshold.

Implementation Path After the Choice

Once you have selected a control strategy, the implementation follows a consistent sequence regardless of which option you chose. We outline the steps below, with specific notes for each strategy.

Step 1: Baseline Data Collection

Install a data logger on your existing recapture system for at least two weeks. Record influent flow, effluent flow, conductivity (or TDS), tank levels, and recirculation pump status at one-minute intervals. This data will serve as the training set for predictive models and the validation set for tuning flow-paced curves. If you cannot log data at that frequency, use the highest resolution your SCADA system supports—at least one sample per five minutes.

Step 2: Define Setpoint Boundaries

Determine the minimum and maximum recapture rates your system can sustain. The minimum is usually set by discharge permit limits (you cannot send water to drain that exceeds a certain conductivity). The maximum is set by the treatment stage capacity (e.g., membrane flux or clarifier loading). These boundaries will clamp the adaptive setpoint to prevent the controller from demanding impossible performance.

Step 3: Implement the Control Logic

For time-based control, program a schedule table with start times and setpoints. Include a manual override that lets operators pause the schedule during unscheduled events. For flow-paced control, program the curve as a piecewise linear function with at least three breakpoints (low, medium, high flow). Use a rate limiter (e.g., maximum setpoint change of 5% per minute) to avoid sudden jumps. For predictive adaptive control, implement a moving-average forecast of flow over the next 15 minutes. The forecast can be as simple as a weighted average of the last 30 minutes, with more weight on the most recent data. Use the forecast to adjust the setpoint via a proportional gain that you tune during commissioning.

Step 4: Commission in Phases

Start with the controller in manual mode, observing its recommendations without letting it change the actual setpoint. After one week of logging, review the recommendations and adjust the tuning parameters. Then switch to automatic mode but with a narrow setpoint range (e.g., ±5% of the fixed setpoint). Gradually widen the range over two weeks while monitoring for oscillations or overflow events. Only after four weeks of stable operation should you enable full adaptive range.

Step 5: Set Up Alarms and Fallbacks

Program alarms for: (a) setpoint at boundary for more than 10 minutes (indicates the system is saturated), (b) forecast error exceeding 20% (for predictive control), (c) tank level approaching high or low limits. The fallback logic should revert to a safe fixed setpoint (e.g., 70% recapture) if any alarm persists for more than five minutes. This prevents the adaptive controller from driving the system into a dangerous state.

Risks If You Choose Wrong or Skip Steps

Every control strategy has failure modes. Understanding them upfront can save you from a costly retrofit or a production outage.

Risk 1: Oscillation from Overly Aggressive Tuning

This is the most common problem with flow-paced and predictive control. If the gain is too high, the setpoint changes faster than the process can respond, causing the recirculation rate to cycle up and down. The result is wasted pump energy and inconsistent water quality. We have seen a facility where the recapture rate oscillated between 40% and 90% every 15 minutes, causing the downstream treatment system to trip on high conductivity. The fix was to reduce the gain by half and add a 2-minute moving average filter to the flow signal.

Risk 2: Sensor Drift Leading to Wrong Setpoints

Flow meters and conductivity sensors drift over time. A flow meter that reads 10% low will cause the flow-paced controller to select a setpoint meant for a lower flow, potentially overloading the system. Predictive controllers are even more sensitive because they rely on accurate historical data. We recommend calibrating flow meters every six months and conductivity sensors every three months. Install redundant sensors on critical loops so that a single drift event does not go unnoticed.

Risk 3: Schedule Changes That Break Time-Based Control

If you choose time-based control and your production schedule changes, you must update the schedule file. In practice, this often gets forgotten. One plant we worked with had a schedule that was three months out of date; the controller was trying to recapture at 80% during a period when the actual flow was 50% higher, causing repeated overflows. The solution was to switch to flow-paced control, which required no schedule maintenance.

Risk 4: Model Drift in Predictive Control

Even a well-tuned predictive model can drift if the process changes gradually—for example, if a new heat exchanger reduces cooling water demand, or if a different raw material changes the contaminant profile. Without periodic retraining, the model’s forecasts become inaccurate, and the controller may make poor decisions. We recommend retraining the model every three months, or whenever the mean absolute forecast error exceeds 15% over a one-week period.

Risk 5: Ignoring Hydraulic Constraints

Adaptive control cannot fix a pipe that is too small or a pump that is undersized. If your peak flow exceeds the recapture system’s hydraulic capacity, no control strategy will prevent overflow. Before implementing any adaptive control, verify that your pumps, valves, and holding tanks can handle the maximum expected flow. If they cannot, the first step is a hydraulic upgrade, not a control upgrade.

Mini-FAQ

We close with answers to the questions we hear most often from engineers evaluating adaptive control for process water recapture.

Can I retrofit an existing fixed-setpoint system with adaptive control?

Yes, in most cases. The hardware changes are minimal: you need a flow meter (if not already present) and possibly a conductivity sensor. The PLC logic can be updated without replacing the controller, as long as it has spare analog input channels and supports math functions. The biggest effort is tuning, which typically takes two to four weeks of iterative adjustments.

How do I handle chemical dosing under variable flow?

If you are dosing chemicals (e.g., pH adjustment or coagulant) in the recapture loop, the adaptive setpoint changes will affect the required dose rate. The simplest approach is to pace the chemical feed proportionally to the recirculation flow rate, which is a direct function of the recapture setpoint. More advanced: use a feed-forward loop that adjusts dose based on the predicted load from the adaptive controller. This can reduce chemical waste by 10–20% compared to fixed dosing.

What if my flow varies faster than the controller can respond?

If flow changes occur in seconds rather than minutes, the adaptive controller may not be fast enough. In that case, consider adding a small buffer tank upstream of the recapture system to dampen the fluctuations. A tank with 5–10 minutes of residence time at average flow will smooth out most rapid transients, allowing the controller to operate on a more stable signal.

Do I need a dedicated engineer to maintain the system?

For time-based and flow-paced control, the maintenance burden is low and can be handled by a skilled instrument technician. For predictive adaptive control, we recommend having at least one person on staff who understands basic time-series modeling and can retrain the model when needed. If that is not possible, consider outsourcing model maintenance to a system integrator on a quarterly basis.

Is adaptive control worth it for small systems (under 50 gpm)?

The payback period depends on water and sewer costs. For a 50 gpm system running 16 hours per day, a 10% improvement in recapture rate can save 4,800 gallons per day. At $5 per 1,000 gallons, that is $24 per day, or about $6,000 per year. The implementation cost for flow-paced control might be $5,000–$10,000, giving a payback of one to two years. For time-based control, the payback is even faster if you already have a PLC. So yes, adaptive control can be worthwhile even for small systems, especially if water costs are high.

What is the single most important thing to get right?

Data quality. If your flow meter is inaccurate or your conductivity sensor is drifting, no control strategy will work well. Invest in good instrumentation and a regular calibration schedule before you invest in control logic. We have seen too many adaptive control projects fail because the sensors were not reliable. Once the data is trustworthy, the control logic is relatively straightforward.