I have seen this happen many times. The result depends on the particular details of the system. Normally, I would expect the useful outgoing flow to be reduced or even to drop to zero. The total flow through the running pump may increase, but not because of an increase in suction pressure. As you noted, this would normally be impossible. The pump flow only increases as the pressure at the discharge drops. With lower discharge pressure, the outgoing flow would be reduced. However, if there are automatic flow controls, they may respond and open control valves that attempts to restore the outgoing flow. Unless the running pump is drastically oversized, it will probably not be able to provide normal outgoing flow with this bypass wide open.

Most of the time in our plant when I have seen this, the outgoing (useful) flow drops to zero and all the flow from the running pump is spilling backwards through the off-line pump.

Johnny Pellin

The process flow will decrease. Depending on conditions it can decrease substantially or nearly stop.

The water wants to flow through the path of least resistance. Because you're essentially short-circuiting the loop, the pump isn't doing as much work as it was before- it doesn't need to produce the same head to push the water.

Let's assume a few things, for the sake of illustration;
No elevation changes, no automatic controls
Piping is 2" (DN50), and the main loop is 200ft (61m) (equivalent)
The "bypass" is 4ft (1.2m) (equivalent)
The pump flow range is 10-100 GPM (37-378 lpm)

Your pump is normally setup to run through 200ft of 2" pipe and does so around 80 gpm, right in the peak efficiency, eye of the curve. Flowing along at about 7.5 ft/s, and it takes about 8.5psi to overcome the friction losses.

Whatever happens, your maintenance team rotates pumps over the weekend and the check valve stays open. The bypass is opened. That short length of pipe offers almost zero resistance, comparatively. If your pump now runs out to 100 gpm that's only 0.25psi to get through the short 4ft section and back to the pump suction. To match that 0.25psi in your 200ft main section we work the numbers backwards and get to about 10-12gpm. Actual values will change as the pump finds its happy spot in the new setup. Actual flow through each run and actual head will be a nice balance with equal inlet pressure and the flow split between the two runs according to that resistance.

So your loop was getting 80gpm (Quseful). Now your bypass (Qrecirc) is getting over 90gpm and your process (Quseful) is getting less than 10gpm as your pump finds that balance. The pump is running off its curve at 100+gpm (Qpump). Your discharge head on the pump went from 8.5psi to 0.25psi.

These are not real-world numbers, just a quick exercise to demonstrate how I would work through some of these numbers.

Thanks for the replies, now I get it better.

It does depend though on the level of bypass through the pump and how your downstream system responds or is sensitive to changes in head produced by the pump.

Large scale flow through the NRV - forward flow can be reduced to zero quite easily if there is a static pressure or head which needs to be overcome before you get flow.

In practice though, if though you have some sort of flow control, either VFD or control valves, and only say 10 to 15% bypassing the pump they might be able to compensate for the extra flow going around the pump with lower head.

More common are minimum flow bypasses operating when they shouldn't or passing flow due to seat wear or poor controls. I really don't like them but are sometimes required when you get large scale flow ranges.

Remember - More details = better answers
Also: If you get a response it's polite to respond to it.

Yes, good point about the controls. I will be considering flow-controlled scenarios separately.

There are two operating failures that are commonly seen in these situations.
1. The back spinning pump has the impeller unscrew from the shaft. When this pump is then started it often seizes because the impeller is not in the correct position.
2. If you hard start a pump while it is spinning backward you can bend or break the shaft. In some cases the shaft defection is enough to destroy seals, and in some cases serious motor damage can happen (winding and/or bearings).

We used to have back EMF sense on all pumps in parallel service. It was a quick way to find leaking check valves before they became a big issue. And we used this signal to inhibit motor start.
This back flow can also happen where you have a pump a low elevation discharging into a lot of elevated piping.

= = = = = = = = = = = = = = = = = = = =
P.E. Metallurgy, Plymouth Tube