Nonlinear pH Control Loop Cycling
A municipal wastewater plant uses two-stage pH neutralization. Influent pH swings between 2 and 11 at 800 gpm; 93% sulfuric acid and 25% caustic are added. The final pH controller is a PI loop on a 6 in air-operated diaphragm valve; the setpoint is 7.0. The pH electrode is two-point calibrated daily with pH 4 and 7 buffers; slope is 91%. The loop cycles between 6.2 and 8.5 with a period of about 12 minutes, and the acid valve saturates at 0% then 100%. The operator has tried P=5, I=0.5 min and P=0.5, I=5 min; both cycle. The influent flow is measured and could be used for feedforward. The mixing tank has a 3-minute hydraulic residence time, and the electrode is 20 ft downstream of the reagent injection point.
1. Why can the loop not be fixed by tuning alone?
Read the worked answer
The pH titration curve is highly nonlinear. Near the equivalence point, a very small reagent change produces a large pH change, while far from neutrality the same reagent change produces almost no pH change. The process gain may change by one or two orders of magnitude across the operating range. A single PI tuning set therefore cannot be stable everywhere: settings that are calm near pH 7 will be sluggish at low pH, and settings that respond at low pH will cycle near neutrality.
The 12-minute period and valve saturation are consistent with a limit cycle driven by variable gain plus dead time. The electrode is 20 ft downstream and the tank has a 3-minute residence time, so there is transport delay and mixing lag that further reduce phase margin. The daily calibration and 91% slope suggest the electrode is usable, though it should be cleaned and checked; calibration is not the primary cause. Tuning can reduce the amplitude, but it cannot remove the gain nonlinearity or the saturation.
2. What control strategy would you recommend, and what would you verify during commissioning?
Read the worked answer
I would recommend gain scheduling or adaptive control based on pH, combined with feedforward from influent flow and, if available, influent pH. A practical scheme is to split the reagent addition into coarse and fine valves, or to use a ratio controller on flow with a pH trim loop. A static mixer or better injection point would reduce dead time. If the plant has the capability, a model-based controller or a scheduled gain table with several breakpoints is preferable to a single PI loop.
During commissioning I would step-test the reagent flow at several pH levels to measure process gain and dead time, then build the gain schedule from those results. I would verify valve characterization and fail-safe action, confirm the pH electrode response time and cleaning cycle, and test the loop at the worst-case gain point. I would also check that the controller output does not saturate during normal influent swings and that alarms and permit limits are independent of the control loop. The final tuning should be validated over a full diurnal cycle, not just at one pH value.
3. What immediate operational measures would you take before the control upgrade is complete?
Read the worked answer
I would put the pH loop in manual or a conservative automatic mode with tight operator surveillance, and I would use the influent flow and pH trends to anticipate reagent changes rather than waiting for the feedback loop to react. I would verify the electrode calibration and cleaning, check for fouling or a partially plugged reference junction, and confirm the reagent valve is not sticking. If the effluent permit is at risk, I would reduce throughput or increase buffer capacity until control is stable.
I would also add or verify independent high/low pH alarms and a fail-safe reagent shutdown, because a cycling loop that saturates the acid valve can overshoot into the permit limit. Any temporary tuning change should be documented and time-limited. The operational goal is to keep the plant inside permit while the permanent fix-gain scheduling, feedforward, and mixing improvements-is engineered and tested.