Thermostat behavior and failure modes

Question

If Zaid's excellent How can I test my thermostat? is "Thermostats 101" then this might be the beginning of "Thermostats 201"…

My VW Jetta TDI (diesel) has been running cold (about one or two marks on the temperature gauge) instead of being pegged to at 190° – courtesy of the ECU, the gauge doesn't really read temperature anything between roughly 170° F and 210° F is display as 190°. The heater output has also been feeling low and the engine has been taking longer to warm up.

All in all this seemed to me like it was a thermostat problem (as opposed to a temperature sensor problem) since the heater wasn't putting out as much heat as I expected and also because a sensor problem usually goes with a somewhat erratic temperature gauge and mine was nice and stable, just low.

So, I got a new thermostat from my friendly (really) local NAPA store and put it in last night. After a test drive the problem seemed resolved – gauge is back to 190° and I've got plenty of heat. But I couldn't leave well enough alone and I put the old thermostat in a pan of water with a thermometer and started heating it up. The spec on the thermostat is for it to open at 195° and that's exactly what it did. Took it off the stove and as the temperature dropped past 190° the thermostat was well on its way to closed, by 160° it was closed completely.

Now I'm puzzled. Seeing this happen makes me think that the old thermostat might not be bad. Or, is it taking too long to close? So I went for a drive, I needed a haircut anyway, the new thermostat really does appear to be working better.

So, what I'd like to understand is:

• What should the hysteresis curve of a good thermostat look like? If it starts to open at say, 195° on a rising temperature, should it be finishing closing at the same point as the temperature falls?

• Are there failure modes besides "stuck open" and "stuck closed?"

Answer 1

Are there failure modes besides "stuck open" and "stuck closed?"

Absolutely. This has been the bane of many an E39 M5 owner.

If the coolant is able to navigate through an alternative flow path besides the path it is expected to take, a mechanical thermostat will modulate temperature at slightly lower than expected.

So what could cause this?

The thermostat will invariably be seated in a gasket/seal of some sort. If this gasket is pinched, the wrong size or loose due to old age/thermal fatigue (real, honest-to-God hysteresis) some of the coolant flow will be able to bypass the thermostat and run the overall cooling system at a slightly higher flow rate.

Answer 2

I'm sure that the answer to this question is specific to the make and model, but if you're really curious you could easily set up an experiment to get an approximation to the correct answer. Really I would doubt that hysteresis plays a part in a car thermostat the same way that it does in a HVAC thermostat (since the thermostat is not just fully opened of fully closed, but can be anywhere between that spectrum).

You could probably see the difference just by finding the steady state gap for various temperature levels as described in the following experiment.

Experiment: finding the "infinity curve"

Simply buy another new thermostat from NAPA, and get a micrometer to take precise measurements, then heat water in different containers to several different temperatures (say 160° to 200° in increments of 5°). Leave the thermostat in each container long enough that it reaches a steady state, and measure the gap with your micrometer. Repeat the experiment with the old thermostat and then you will be able to compute two discrete curves as a function of the ambient temperature.

My hypothesis is that the difference between the two thermostats will likely be seen when comparing these two curves, the old thermostat having a smaller gap at 190° than the new one. In this paradigm, the failure mode would be "open to the incorrect gap at a given temperature".

If the curves happen to be the same, then that means that thermostat hysteresis really is essential to the regulation of temperature, and you would have to devise an experiment to measure the opening and closing time constants as a function of temperature as well. Then a linear differential equation model of the gap would be given by

dg/dt = (g_inf - g)/g_tau

where g_inf is the curve computed in the first experiment and g_tau is the curve computed is the time constant curve.