Pump Reliability & Predictive Maintenance (PlusSmart) · 10 min read · Updated November 2026
Parallel Pump Stations on VFDs: Why Equal Speed Fails and How Flow Balancing Fixes It
Four identical pumps, four identical drives, one setpoint, all running at the same speed — and one of them fails its seal every quarter while the HMI says everything is normal. The physics of parallel pumps, the control mistake almost every station makes, and the control that took a port from 52 unplanned stops a year to zero.
Parallel pumping is how most stations are built: duty/standby or duty/assist groups feeding a common header, with a pressure or level setpoint and a PLC that starts and stops pumps and sends the same speed reference to all the running drives. It is simple and it is wrong in a way that only shows up after a few years, when the pumps are no longer identical. The iron-ore port case — continuous seal failures, AUD 350,000 a year in maintenance, 52 unplanned stops, 700 kPa discharge-pressure swings — was a textbook equal-speed station. After ReliaPumpSync replaced equal-speed control with flow balancing, it has had zero seal failures since 2015, maintenance under AUD 5,000 a year, no unplanned stops and 50 kPa pressure swings.
The physics in one diagram you can draw by hand
Draw two pump curves on head-versus-flow axes, one slightly lower than the other (a worn pump has a lower curve: wear rings leak, the impeller is eroded). Draw the system curve through them. At equal speed both pumps see the same header pressure, so each operates where its own curve meets that pressure. The pump with the higher curve delivers more flow; the worn one delivers less — and as the header pressure rises (more pumps running, system resistance up), the worn pump's operating point slides left along its curve towards shut-off. At the limit it dead-heads: it runs, draws power, heats the water in its casing and produces no flow. Its seal faces flash. The newer pump, meanwhile, is pushed right towards run-out, cavitation and overload. The HMI shows both "running" at the same speed and the same, perfectly normal, header pressure.
Symptoms of an unbalanced station
| Symptom | What is happening |
|---|---|
| Seal failures concentrate on one or two pumps | Those pumps run near dead-head; flashing at the faces |
| Bearing failures on the newest pump | It runs out to the right, cavitating, with high radial load |
| Energy per m³ rising year on year | Pumps operating far from BEP, more pumps running than needed |
| Pressure oscillation, check-valve chatter, water hammer on starts | Pumps fighting each other; hard starts into a pressurised header |
| "We replaced the pump and it failed too" | The new pump is now the strong one and pushes a different pump to dead-head |
What flow balancing does
ReliaPumpSync estimates each pump's flow from its own drive data (speed, torque, power) against that pump's actual curve — characterised in operation, so a worn pump has a worn curve in the model — and then adjusts individual speeds so each pump carries its share of the station flow within its preferred operating region (roughly 70–120 % of BEP, per ANSI/HI 9.6.3). The worn pump gets a slightly higher speed; the strong pump a slightly lower one. The module also decides how many pumps should run for the current demand, which is usually fewer than the equal-speed logic starts, and ramps changes so the header pressure moves smoothly instead of in steps. No flow meters are added; no PLC is replaced.
Protection comes with it
The same operating-point model provides ReliaPumpIQ's protections per pump: dead-head detection (flow near zero at speed), dry-run detection that ignores transient air on start-up, run-out limits, and "right fault, right alarm, right trip" so an operator sees which pump has which problem rather than a generic drive trip. ReliaPump360 adds the monitoring layer — reliability, prediction, energy and performance analytics from the drive data, local or in the cloud.
Energy: the cube law with a static-head caveat
Affinity laws say power scales with speed cubed, so 80 % speed needs about 51 % power. That is true against a purely frictional system; against a static head (lifting water 40 m) the savings are smaller, because the pump must always make at least that head. Correct optimisation therefore works against the real system curve — which the drive data also reveals — rather than simply reducing speed. The geothermal plant with 24 pentane feed pumps exceeded its energy-saving target by 30 % because the control was optimised against the actual system, with the right number of pumps running at the right points.
Implementation on an existing station
- Send PlusSmart the DCS or drive trends (speed, current, power, header pressure, level) for a few weeks and the pump datasheets if available.
- Assessment: which pumps are being pushed where, whether Sync or IQ is needed, which drive features are available (sensorless flow control depends on hardware).
- Characterise each pump in operation; build the model; configure protections.
- Commission alongside the existing PLC with a cut-over window of hours; the 24-pump plant did it without shutdown and without original documentation.
- Verify: seal failures, pressure swings, kWh/m³, pumps running at each demand level — before and after.
Pair the control with the right seal on the pumps that have suffered — Flexaseal reviews the failed seals — and with i-ALERT on the bearings so the improvement is measured, not assumed.
Need engineering help with this?
Our engineers apply this to real projects every week — selection, specification, commissioning. Send your question or a part number.
Quick answers
Can this be done with pumps of different sizes?
Yes. Flow balancing works from each pump's own curve, so mixed sizes and mixed ages are handled; the model simply gives each pump a different speed to keep it in its preferred region.
Do we need to replace the PLC?
No. ReliaPump integrates with the existing control system and drives; the station PLC keeps its role. Some advanced features depend on the drive hardware, which the assessment checks.
How quickly do seal failures stop?
From the moment the pumps stop dead-heading. Seals already damaged will still fail; after they are replaced, the port case has recorded none since 2015.