Why solenoid force drops after thermal cycling

Why this happens

A unit that is measurably weaker after a long run is a common enough complaint, and the first instinct is almost always wrong. A part that measured 12 N when it was built and measures 7 N two years later has lost 42 % of its force, and how the reason is found depends entirely on which of three mechanisms took it. The magnetic material is the least likely cause and the one that gets changed first.

Soft magnetic steel does not decay in service. Low-carbon steels and electrical irons are used well below saturation, and their saturation flux density is a property of the alloy that does not drift with time or with thermal exposure at ordinary temperatures. What does change is the residual stress left in the part by machining, punching or forming, which raises coercivity and lowers permeability. Heating relieves that stress, so the effect of thermal exposure is a small improvement, not a loss. Low-carbon steel saturates somewhere between 1.6 T and 2.0 T, and that ceiling belongs to the alloy rather than to its service history. A working point at 1.2 T still has 0.4 T of headroom before the alloy runs out, and nothing about a thousand hours at 100 C moves that figure. On a unit that has been cycled, the material is if anything slightly better than when it was made. That is the opposite of the assumption that drives most material changes, and it is set out alongside the surface treatment question in plunger material selection.

The real mechanisms are mechanical, and there are three of them. They produce the same symptom and have three different fixes, which is why measuring before changing anything matters. One of the three is worth well under 10 % of the force on a typical design, the other two are worth tens of percent, and telling them apart costs 3 measurements.

The first is creep in the thermoplastic bobbin. Glass-filled thermoplastics lose stiffness at temperature, and under a sustained load they continue to deform for as long as the load and the temperature are held. The published heat deflection temperature for a glass-filled polyamide is often quoted around 240 C, but that figure describes short-term stiffness under a defined load and is not a long-term working limit. Long-term behaviour is governed by the creep curves at the working stress and temperature, and for a glass-filled polyamide those curves are useful well below 150 C. A bobbin creeping 1 % over a thousand hours at 100 C moves a 13 mm dimension by 0.13 mm. Creep rate on a glass-filled polyamide scales roughly with the cube of stress, so a 50 % increase in stress more than triples the rate, and running the same part 20 K hotter doubles it again. Those two effects together are the difference between a part that stays where it was built and a part that drifts out of specification inside its warranty.

That tenth of a millimetre is the whole story when it lands in the air gap. If the surface that has moved sets the end of stroke, the working gap has grown. Take a 0.30 mm gap that becomes 0.43 mm. Force in the unsaturated region varies as the inverse of the gap squared, so it falls to about 49 % of what it was. Nothing electrical has changed, no material has degraded, and the coil resistance is exactly where the design put it. The unit is weak because a plastic part moved, and it moved slowly enough that no incoming inspection would have caught it.

The second is stress relaxation in the return spring. A spring held compressed at temperature loses force over time, a few percent in a thousand hours, depending on material, initial stress and temperature. On a design with a substantial force margin this is invisible. On a low-power design where the spring occupies a third of the available force, it shows up as a unit that starts reliably when new and hesitates after a year. A spring that gives up 5 % of its force costs under 2 % of the net force when it held a third of the budget, and costs the entire margin when it held the entire margin. The mechanism and the pre-load arithmetic that makes it matter are set out in how a return spring eats usable stroke.

The third is settling at the joints, and it is driven by differential expansion. A stack of aluminium, steel, brass and glass-filled plastic has a different thermal expansion coefficient in every layer. Aluminium expands at around 23 ppm per kelvin, steel at 12, and a glass-filled thermoplastic at 30 or more along the flow direction. Across a 70 K swing on a 20 mm stack, those differences produce relative movements of tens of microns at every clamped interface. Repeated cycling works those interfaces against each other, the joint settles, the pre-load drops, and the assembled gap opens very slightly. This is fretting, and it is why a unit can be tight when it is assembled and loose after a hundred cycles. It is also cumulative: a joint moving 0.04 mm per cycle has travelled 40 mm of relative motion by the thousandth cycle, every millimetre of it spending the pre-load that holds the assembly together.

Measurement is what separates them. A magnet that has genuinely weakened would show a changed inductance for the same gap and the same winding. A gap that has opened shows a changed inductance at the same plunger position, a different free gap, or both. A relaxed spring shows nothing electrical at all. Three measurements, taken on the same unit and compared against a new one from the same batch, identify the mechanism in an afternoon. An inductance bridge, a depth gauge and a spring gauge are the whole instrument list, and the three together take under an hour on a unit that is already apart. A gap that has opened from 0.30 mm to 0.40 mm lowers the seated inductance by roughly a quarter, which is far above the repeatability of an ordinary bridge.

Check these in order

1. Measure inductance at a defined plunger position, not resistance. Resistance tells you about the wire. Inductance tells you about the magnetic circuit, which includes the gap. Measure on the suspect unit and on a new unit of the same design, with the plunger clamped at the same position, because inductance varies strongly with gap and an unclamped comparison measures the fixture rather than the part. A 0.05 mm difference in clamping position moves the reading further than the defect usually does, so the fixture has to be repeatable before the number means anything.

2. Measure the free, unassembled gap. With the spring out and no load applied, measure the gap between the pole faces at the seated position. Compare with the drawing and with a new unit. This is the measurement that finds creep and settling, and it is the one most often skipped because it requires disassembly. A free gap 0.05 mm wider than the drawing is worth investigating; 0.10 mm wider is a different product.

3. Measure the spring on its own. Take it out and measure force at the installed length and at the stroke-start length. A spring that has relaxed shows it immediately. This is a two-minute test and it eliminates a third of the possibilities. A spring specified at 0.50 N that now reads 0.46 N has given up 8 %, which is within range for a thousand hours at 100 C.

4. Measure the assembled gap, then the gap after a hot hold. The difference between the two on a single unit, measured before and after a few hours at operating temperature, quantifies how much of the loss is time-dependent rather than temperature-dependent. Log the gap after 1 hour, after 4 hours and again overnight; a reading that is still opening at the last point has not reached equilibrium.

5. Check the pre-load on any bolted or crimped joint. Torque the fixings to specification and check the seating of any crimp. Settled joints are usually visible on close inspection once you know to look for witness marks at the interface. A crimp that has settled 0.02 mm has given away 0.02 mm of gap, and on a 0.30 mm working gap that costs about 12 % of the force.

6. Look at the bobbin, dimensionally and visually. Measure the axial dimensions against the drawing, and look for the discolouration or gloss change that indicates the part has been at or above its design temperature. A part that has been running hotter than intended will show it.

7. Compare the resistance of the suspect winding with the drawing, hot and cold. A winding that has deformed or lost turns shows up here. A winding that reads exactly as designed has not changed, which shifts the investigation to the mechanism. Copper sits about 31 % higher in resistance at 100 C than at 20 C, so a suspect winding should read high when hot, and a reading that has fallen points at shorted turns rather than at the mechanism.

8. Establish whether the loss is monotonic or stepped. A gradual loss over thousands of hours points at creep or relaxation. A step change after a particular event points at settling or at a single thermal excursion. The shape of the trend is diagnostic in itself.

What actually to change

What the measurements show	What to change	Why not the obvious thing
Free gap wider than drawing, inductance low	Change the bobbin material or add a metallic stop that sets the gap	Grinding the plunger moves the stop and does not restore the gap
Spring force below specification	Reduce working stress, change material or reduce the hold time at temperature	A stiffer spring costs force at the start of stroke
Gap correct when cold, wider when hot	Take the gap reference off the plastic and put it on metal	Increasing the coil power treats the symptom and adds heat
Pre-load lost at a joint	Change the joint to one that does not rely on friction, or use a positive stop	Retorquing restores it only until the next hundred cycles
Resistance or inductance abnormal	Investigate the winding rather than the magnetics	A winding fault is an electrical problem even when it presents as weak force
Force constant, behaviour changed	Look at friction at the guided surfaces	A uniform force loss has to come from the gap, not the guide
Loss only on the oldest units	Add creep and relaxation to the qualification test	A short soak test cannot see a mechanism that runs for thousands of hours
Material changed and nothing improved	Put the original material back	The loss was geometric and the change was aimed at the wrong variable

The last row is worth stating plainly. Material substitution is the most expensive single change available and it addresses the least likely cause. Because the measured effect of a core material change is usually a few percent, while a gap change of 0.1 mm on a 0.3 mm gap is worth about half the force, a programme that changes material after a force loss has moved in the wrong direction at the highest cost. Force varies with the square of the gap, so 0.10 mm of extra gap on a 0.30 mm working gap costs about 44 % of the force, while changing the core material is usually worth well under 10 %. The comparison of the two effects is worked through from the other side in calculating holding force and the corrections everyone skips.

When it IS the harder problem

The mechanism runs close to the long-term temperature limit of its plastic. Then creep is not a defect but a design outcome, and the fix is to move the load path off the plastic. The useful move is to find every dimension that sets the air gap and put a metal part in charge of it, even at the cost of an extra component, because the alternative is a specification that depends on a creep curve. Published creep data for glass-filled polyamides usually runs to 1,000 hours and sometimes to 10,000, so extrapolating it to a 10-year life is a modelling exercise rather than a lookup. Distinguishing this case from an ordinary thermal problem is the subject of works cold, fails hot.

The part passes every bench test and fails only in the field. Then the test does not reproduce the load history. Creep is driven by time at temperature under load, not by the number of cycles, so a life test that counts operations without holding the unit energised and hot for long periods will not see it. A unit that passes 24 hours of hot soak testing has seen 24 hours of a mechanism whose characteristic time runs into thousands of hours. Adding a sustained hold at operating temperature to the qualification plan costs a few days and catches the mechanism that a cycle count misses, as set out in what a life test should measure.

The loss is real but small, and the margin was never stated. This is the common and awkward case: the force has dropped by 8 %, the design had no stated margin, and there is no way to say whether that matters. The productive response is not to redesign but to establish the margin now, by measuring the load and the available force on the same unit, and then to decide whether the loss is within it. An 8 % loss on a design that carries a 30 % margin is a non-event, and the same 8 % on a design that carries 5 % is a redesign. A force figure without the load it works against is not a specification, and the same discipline applies to batch spread, as set out in inconsistent force across a batch.

Force is not the only thing that has changed. Sometimes the complaint is about timing, noise or release rather than strength, and all three can drift for the same reasons. A relaxed spring changes release speed. A settled joint changes the end-stop impact and therefore the noise. A winding that has crept into a different position changes inductance and therefore response time. When more than one symptom moves together, treat it as one geometric change rather than several faults, and re-measure the three quantities above before assuming the mechanisms are independent.