What are some reasonably reliable ways of detecting that a water pump (electrical motor based) is "dry running", i.e. no water is passing through the pump either because there is no water available at the inlet, or due to "air-block" ?

I'd prefer mechanisms that are least intrusive, i.e. ideally - do not require re-plumbing. However, if all practical mechanisms involves some plumbing work, then it might as well.

An indication regarding possible cost of sensors, installation skill required, would be very useful.

  • \$\begingroup\$ What is the plumbing configuration? Is this a well pump? submerged? in-line pump? etc... \$\endgroup\$ Jan 19, 2013 at 4:53
  • \$\begingroup\$ Not familiar with the typical plumbing terminology, but it is certainly not submerged. Pump inlet is fed by an underground sump (about 10 feet deep), pump itself is at ground level, and it pumps up water into an overhead tank, which is about 40 feet above ground. Typically it is a 1.5HP 2000W 220VAC monoblock pump. I am just guessing that the configuration is "in-line", I could be wrong. \$\endgroup\$
    – bdutta74
    Jan 19, 2013 at 8:43
  • \$\begingroup\$ Yeah, that's an inline configuration (it's between the well and the tank in elevation and not submerged). \$\endgroup\$ Jan 19, 2013 at 9:47
  • 1
    \$\begingroup\$ Do you specifically want to prevent dry running, or are you looking or prevent other situations where the pump can become damaged, like a stalled rotor? \$\endgroup\$ Jan 19, 2013 at 12:59
  • \$\begingroup\$ While not part of my original question, it would be very good to prevent situations where pump can be damaged. We've had cases of stalled rotor, following prolonged period of dry-running and total loss of lubrication, but I've to admit that I'd forgotten all about it. \$\endgroup\$
    – bdutta74
    Jan 19, 2013 at 13:22

3 Answers 3


Given this additional detail:

Not familiar with the typical plumbing terminology, but it is certainly not submerged. Pump inlet is fed by an underground sump (about 10 feet deep), pump itself is at ground level, and it pumps up water into an overhead tank, which is about 40 feet above ground. Typically it is a 1.5HP 2000W 220VAC monoblock pump. I am just guessing that the configuration is "in-line", I could be wrong.

I've got a very simple, cheap, reliable solution for you: remote monitoring.

Nothing says your sensor has to be at the pump's inlet. You could place your sensor at the system's inlet. Submerge a cheap float sensor at the level of the inlet to the feeder pipe in your well (or whatever water source). Run wires back to the pump's power controller.

When the float sensor says the water level is below the feeder pipe, turn off the pump. Ten bucks, totally reliable. Done. =)

Multiple float switches are increasing depth in a tank

This figure shows multiple float switches at increasing depths in a tank to illustrate operation. You just need one at the bottom of your feed pipe. You do not need to disturb any existing plumbing to do this.

  • \$\begingroup\$ Excellent. I might have accepted this as an answer, but for one doubt. Is the only reason that pumps encounter dry-running, because there is no water at pump's inlet ? If I am confident that this is absolutely the only reason, I think I can accept this answer. Else, the flow-sensor seems to be more reliable. \$\endgroup\$
    – bdutta74
    Jan 19, 2013 at 12:37
  • \$\begingroup\$ A definite vote-up for the excellent diagram. \$\endgroup\$
    – bdutta74
    Jan 19, 2013 at 12:38
  • \$\begingroup\$ @icarus74 No, there are other reasons, including trapped air, pipe blockage, broken impellers, suction pipe leakage etc. The float method addresses one of the causes, whereas the current profiling mentioned in another answer, addresses the superset pattern of free-running of the pump motor. \$\endgroup\$ Jan 19, 2013 at 17:50
  • \$\begingroup\$ @icarus74 -- If your system isn't leaking, then no water at system inlet = no water at pump inlet (or at least you've lost your fluid seal so the water in the inlet pipe will begin to fall back away from the pump due to gravity). You actually may not want to monitor at the pump inlet because you will need to run the pump briefly during dry conditions to prime it after the well refills (e.g. while the top of the inlet pipe is still dry and you are sucking up the new well water). \$\endgroup\$ Jan 19, 2013 at 18:05
  • 2
    \$\begingroup\$ I would also add that these systems are additive. The float sensor is the "best" approach in that it gives you the most performance/accuracy for the least cost, but you could certainly add current monitoring in addition. Knowing that there is water in the well helps make current profiling way more accurate. For example, you could detect blockages in this way. \$\endgroup\$ Jan 19, 2013 at 18:08

If you can sense motor current this might be the least intrusive method, but you must test it first to make sure that there is a reasonable difference between "pumping water" vs "spinning freely". Note that AC motors all have a current surge at the very start, so your circuit should probably look for the current level for a few seconds before making its determination.

As far as how to wire it up -- a current transformer would be slipped over ONE of the motor leads -- preferably the hot lead. You might want to also sense whether the motor is being asked to pump or not so that you could also detect a loss of power or a broken motor lead.

The circuit could also easily detect a jammed or stalled motor by a current draw above the normal pumping current level, but this too would have to be tested to ensure that the "pumping water" vs "stalled" conditions were different enough to reliably detect.

Alternately, you could just count the time that the motor is on... if the motor has been running for more than x minutes, it's likely that it's not actually pumping water since the water level hasn't fallen enough to turn the motor off. That's of course assuming that the pump duty cycle has bounds that are acceptable to your "dry" condition test.

  • \$\begingroup\$ Thanks @Andrew. Both mechanisms are wonderfully explained and pretty clear. As for the "current" detection, I am guess other methods s.a. hall-effect sensor would be equally applicable ? Since my question is indeed in context of another one I asked earlier today (water level detection), so the two can be blended together to work on the rate of level-drop, to determine if it is dry running or not. \$\endgroup\$
    – bdutta74
    Jan 18, 2013 at 13:03
  • \$\begingroup\$ You could also measure motor speed, presumably it spins slower under load, this would also allow you to detect motor jams if debris gets in there. Current measurement is the #1 idea though as it allows for over-current (burnout) protection. \$\endgroup\$
    – John U
    Jan 18, 2013 at 13:05
  • 1
    \$\begingroup\$ If you go with the home brew application of using a current transformer be aware of the safety hazard of using one of these with out its load attached. A current transformer without a load can generate an extremely high voltage that could be fatal - to both the human playing with the transformer and to the transformer itself. In the latter case the voltage can get high enough to punch through the winding insulation inside the transformer coil. \$\endgroup\$ Jan 18, 2013 at 15:44
  • \$\begingroup\$ @MichaelKaras: Does anyone make current transformers with an integrated, specified-value, resistor? If using 1K of resistance would yield a 0-3 volt signal which is what one wants, a coil with a 10K resistor permanently wired across it would limit voltage to a relatively-safe 30 volts without an external load; compensating would simply require using 1.11K instead of 1K. Perhaps power dissipation might be an issue (at a given current level, a 10K resistor would need to dissipate 10x as much power as a 1K) but I'd expect turns ratios to be chosen such that even 10x power would be no biggie. \$\endgroup\$
    – supercat
    Jan 18, 2013 at 16:50
  • \$\begingroup\$ I know that CTs can generate large voltages, but in 13 years of industrial motor control I've never seen a CT do more than short itself out if left unburdened. I'm not advising ever leaving the burden resistor off, mind you, but the danger is there. As far as using a hall effect device: yes, LEMs are more expensive but can work with AC or DC motors and don't have the same potential for harm (of a person or the device itself). \$\endgroup\$
    – akohlsmith
    Jan 18, 2013 at 19:04

Sensing motor current may not distinguish "spinning freely" as the motor ages and lubrication starts to fail.

A flow switch is a minor plumbing job but will definitely tell you if water is running.

  • \$\begingroup\$ Select the flow switch carefully. These will require a certain pressure to activate but if your pump does not generate that much pressure the switch will never activate. \$\endgroup\$ Jan 18, 2013 at 15:38
  • \$\begingroup\$ Thanks Brian. Something of this sort is what I had envisaged, although I didn't know what these are called. Now I do have a good lead. Also thanks to @Michael, I need to worry about the minimum pressure as well. \$\endgroup\$
    – bdutta74
    Jan 18, 2013 at 17:04
  • \$\begingroup\$ Obviously you need to delay checking the flow switch until after the pump has been running for a few seconds. You might want to incorporate the float switch idea too if it is inconvenient to reprime the pump. \$\endgroup\$
    – pilotcam
    Jan 19, 2013 at 15:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.