The short answer
Membrane Control
A membrane skid is controlled by a handful of loops, a sequence, a set of trips, and a set of calculations, and the quality of the plant depends on getting each of them right. The permeate flow loop sets the feed pump speed; the recovery loop sets the concentrate valve from the permeate flow and the recovery target; the dosing loops pace the antiscalant, acid, and dechlorination chemicals to the feed flow with analyzer trims; the sequence takes the skid through flush, ramp, run, and flush with permissives at each step; the trips stop the skid on pressure, conductivity, oxidant, pH, and chemical feed failures; and the normalization calculations turn the raw readings into the trends that say when to clean. Several trains share feed, chemicals, and permeate collection, so a plant controller coordinates which trains run, staggers their starts, and balances their production to the plant demand. The historian keeps the normalized trends and the operating conditions for the years the membranes live.
Key points
- Four loops: permeate flow by pump speed, recovery by concentrate valve, dosing by feed flow with trims, and permeate quality by divert.
- A sequence with permissives at every step; a pressure ramp at the manufacturer rate; a flush on every shutdown.
- Trips on pressure, conductivity, oxidant, pH, and chemical feed, each latched and reset by a person.
- Normalization in the controller from measured values, shown beside the raw values, trended for the life of the membranes.
- Multi-train coordination: which trains run, staggered starts, production balanced to demand, shared systems interlocked.
Loops
| Loop | Measured | Manipulated | Notes |
|---|---|---|---|
| Permeate flow | Permeate flowmeter | Feed pump drive speed | Setpoint from the plant demand; ramp limited; output limited by pressure |
| Recovery | Concentrate flowmeter | Concentrate control valve | Setpoint computed from permeate flow and recovery target; slow tuning |
| Feed pH | pH after acid injection | Acid metering pump | Flow-paced with pH trim; high pH trip |
| Antiscalant | Feed flow | Antiscalant metering pump | Pure flow pacing; low flow and low tank alarms; no analyzer |
| Dechlorination | Feed flow and oxidant analyzer | Bisulfite metering pump | Flow-paced with a residual target; oxidant trip |
| Permeate quality | Permeate conductivity | Divert valve | Divert above a limit; stop after a time |
| Interstage boost | Second stage permeate flow or pressure | Boost pump speed | Where fitted |
Sequence
- 1
Ready
Permissives: feed pressure, prefilter differential, chemical tanks and pumps, oxidant analyzer healthy, valves in position, no unreset trips, train selected for automatic.
- 2
Flush to waste
Feed at low pressure through the vessels to waste for a set volume or time; the concentrate valve open; permeate to waste.
- 3
Ramp
Feed pump ramps at the set rate while the concentrate valve moves to its run position; pressure and flow rise together; permeate stays diverted until conductivity is within limit.
- 4
Run
Loops in automatic; trends live; the divert valve returns permeate to service.
- 5
Stop
Ramp down; concentrate valve opens; pump stops.
- 6
Flush
Permeate or low-pressure feed displaces concentrate from the vessels for a set volume; then the train is idle.
- 7
Idle
A periodic flush on a timer; a preservation procedure for long shutdowns; the train remains available.
Trips and alarms
Trips stop the train and latch: high feed or concentrate pressure, low suction pressure, oxidant detected, high feed pH, low antiscalant flow, and sustained high permeate conductivity. Each is reset by a person after the cause is known. Alarms warn without stopping: prefilter differential approaching its limit, a normalized value approaching its cleaning trigger, a chemical tank low, a divert in progress. The design separates an instrument fault from a membrane problem where it can: a conductivity analyzer with a stale or out-of-range signal raises an instrument alarm and holds the last good divert state; an oxidant analyzer fault stops the train, because the risk is asymmetric.
Normalization in the controller
The controller computes the normalized permeate flow, salt passage, and differential pressure every scan from the measured pressures, flows, conductivities, and temperature, using the membrane manufacturer method and a baseline recorded at commissioning after the first stabilization. The values are displayed beside the raw values, trended in the historian, and alarmed against the cleaning triggers. Doing the calculation in the controller rather than in a spreadsheet makes the trend continuous and the cleaning trigger an alarm rather than a monthly discovery. The baseline is re-established after a membrane replacement and recorded.
Several trains
- A plant demand setpoint is divided among the trains in service, equally or by capacity; each train runs its permeate flow loop on its share.
- Trains start staggered so that the feed system, the chemicals, and the electrical system see one ramp at a time.
- Shared systems, the feed pumps, chemical feeds, and the permeate and concentrate headers, are interlocked to the trains that need them.
- Cleaning takes one train out of service while the others carry the demand; the cleaning skid is interlocked so that a train cannot be in service and in cleaning at once.
- Train selection rotates so that hours balance, with a train held out for cleaning or maintenance by selection.
Data
- Every pressure, flow, conductivity, pH, and temperature at a resolution that captures the ramp and the run.
- The normalized values with their baseline.
- The sequence step and its duration; the trips and alarms with time stamps.
- Chemical usage and tank levels.
- Cleaning events with their conditions, for the membrane supplier and the warranty.
Frequently asked questions
- Why compute normalization in the controller?
- So that the trend is continuous, the cleaning trigger is an alarm, and the calculation is done the same way every time by something that does not go on vacation. The spreadsheet remains a check.
- How should the recovery loop be tuned?
- Slowly. The concentrate valve loop interacts with the permeate flow loop, and a fast recovery loop makes both oscillate. Tune the flow loop first with the valve fixed, then the recovery loop slowly enough that the flow loop settles between its moves.
- What happens when the oxidant analyzer fails?
- The train stops. The membranes are the most expensive part of the plant and chlorine destroys them in hours; an analyzer fault is treated as chlorine present until proven otherwise. A redundant analyzer is the way to avoid nuisance stops.
- Can operators change the recovery?
- Within limits set from the membrane projection, with the change logged and the antiscalant and pH programs checked. A recovery above the projection scales the last stage within weeks.
Related topics
- Reverse OsmosisReverse osmosis from the controls side: what the skid contains, the instruments that matter, the numbers an operator watches, recovery, rejection, normalized permeate flow, and the trends that say when the membranes need cleaning or replacement.
- NanofiltrationNanofiltration from the controls side: how it differs from reverse osmosis in pressure, recovery, and what it removes, the softening and color removal applications where it dominates, and trends that are shared with reverse osmosis and the ones that differ.
- Clean-in-PlaceCleaning membranes in place: the triggers from the normalized trends, the low pH clean for scale and the high pH clean for organics and biofouling, the sequence of flush, heat, recirculate, soak, and rinse, and how automation makes cleaning repeatable.
- Sequential Function ChartThe IEC 61131-3 language for sequences: steps, transitions, actions and their qualifiers, alternative and simultaneous branches, where it fits in water and wastewater, and the rules that keep a chart from sticking in a step.
- PID Control for Pumps and ValvesWhat each term actually does, how to tune a loop on a real pump station, and the handful of mistakes that cause most of the oscillation people blame on tuning.
- Historian Data CollectionDeciding what the historian stores and how often: polled versus exception collection, the deadband per tag, rates for control loops, levels, totals, and compliance values, timestamps at the source, quality codes, calculated tags, and the tag list review that keeps a historian from filling with noise and missing what matters.
Direct contact
Have a controls question?
Reach Eric Sullivan directly about anything on this site, a controls or automation topic, or one of his personal projects.