How to estimate life from an application duty cycle

The mechanism runs a repeating cycle. How do you turn that cycle into a statement about how long the motor will last, and what part of that statement can only come from a test?

A duty cycle is the sequence of loads and rests the mechanism repeats. It decides two different things, on two different clocks, and the mistake that makes life estimates useless is treating them as one.

The thermal clock is fast and calculable. It asks whether the winding reaches an acceptable temperature, and it resets whenever the motor cools. The wear clock is slow and cumulative. It counts revolutions, starts, reversals and stalled seconds, and it never resets. Both have to be evaluated, and only the first can be closed with arithmetic.

Start by writing the cycle down

0123405101520Time within one machine cycle (s)Shaft torque (mN·m)RMS torque 5.32 mN·macceleraterundeceleratedwell, power off
One machine cycle, tabulated as four phases and plotted at the scale of its own axes. The RMS line is computed from the phases below it. Torque during deceleration is shown as a magnitude; its sign does not change the heating.
The cycle behind the plot
Phase Torque mN·m Duration s Motor current A
accelerate18.00.151.165
run6.01.200.415
decelerate12.00.150.790
dwell0.02.500 (unpowered)

The thermal clock

Trms = √( Σ Ti² · ti / Σ ti )

Trms
root mean square torque over one whole cycle
Ti
torque during phase i
ti
duration of phase i, including the rest phases
For the cycle above: √(113.4 / 4.0) = 5.32 mN·m. The rest period is included in the denominator, which is what makes an intermittent duty thermally acceptable at torques well above the continuous rating.

RMS torque is a convenient proxy, but the quantity that actually heats the winding is RMS current, and the two are not proportional because current carries the no-load offset. For this cycle the current-weighted result is 0.355 A, which gives a copper loss of 0.57 W and, at an illustrative 10 K/W thermal resistance, a steady rise of about 6 K. Comfortable, and comfortably calculable.

That calculation is the part of a life estimate that can be done honestly at a desk. It answers whether the motor will overheat. It says nothing about how long it will last.

The wear clock

Wear accumulates against counters, not against hours. Converting the cycle into counters is arithmetic, and it is the step most life discussions skip.

At 6 mN·m the illustrative motor turns at about 6047 rpm, so the 1.2 second run phase is roughly 121 motor revolutions. Add the ramps and one cycle is on the order of 130 motor revolutions, one start, one deceleration and no stalled seconds.

One cycle, scaled to an illustrative 100,000-cycle product life
Counter Per cycle Over 100,000 cycles What it wears
Motor revolutionsabout 130about 13 millionBrush sliding distance, bearing revolutions, gear tooth contacts
Starts1100,000Brush and commutator erosion from inrush current
Reversals or braking events1100,000Brush contact, endplay impacts, gear backlash impacts
Stalled seconds00Nothing here, and this row is the one to check first
Powered hours0.0004 habout 42 hVery little. This is why quoting life in hours misleads
Elapsed hours0.0011 habout 111 hGrease oxidation and evaporation, which run on elapsed time and temperature

The last two rows are the argument against hour-based life figures. This mechanism accumulates 13 million revolutions in 42 powered hours. A different mechanism could accumulate 42 powered hours at a tenth of the speed and do a tenth of the sliding. The same "hours" figure describes both.

What limits life, and what closes each one

Life-limiting mechanisms and the evidence that settles them
Mechanism Counter it runs on Can it be calculated?
Brush and commutator wearRevolutions, starts, current density, contaminationNo. Ordering between candidates can be reasoned; the number requires a life test at the real duty
Bearing lifeRevolutions under a stated radial and axial loadPartly. Standard bearing life models apply when the loads are known, and small-motor bearings are frequently below the range those models were built for
Gear tooth wear and fatigueTooth contacts and load per contactPartly, for metal gears with published material data. For small molded gears, test
Grease degradationElapsed time at temperatureDirectionally, using an oxidation rule of thumb. Not to a number
Winding insulationCumulative time at temperatureYes, once the temperature rise is measured, using published insulation class data
Connections and leadsFlex cycles, if the lead moves with the mechanismNo. This is a mechanical fatigue test on the harness, and it is regularly forgotten

Designing the life test

  1. Reproduce the counters, not the hours

    The test has to deliver the same number of starts, reversals and revolutions as the field. A continuous-running test with the same total hours understates brush wear severely.

  2. Compress dwell, carefully

    Shortening rest periods is the legitimate way to accelerate. Verify that the compressed cycle does not raise the steady temperature above the field case, because that changes the failure mechanism rather than accelerating it.

  3. Do not accelerate by raising the load

    Higher torque changes current density, temperature and contact behavior at once. The result is a different failure mode, tested faster.

  4. Measure continuously, not at the end

    Log current, speed and temperature throughout. The trend curve tells you the mechanism. A pass or fail at the end tells you almost nothing.

  5. Define the end-of-life criterion first

    A percentage change from each unit's own initial measurement, agreed before the test starts. Deciding afterwards what counts as failure is how a life test produces an argument instead of a result.

  6. Tear down the units when the test ends

    Including the ones that passed. The wear distribution on a survivor tells you how much margin the design actually had.

A life test run this way produces something more useful than a number. It produces a curve of degradation against counted cycles, which can be extrapolated to a different duty with defensible assumptions. A single pass or fail at one duration cannot.

Where the cycle data comes from

A duty table written from the design intent is a starting point. The one that predicts life is measured on the real mechanism, because mechanisms do things the design did not describe.

  1. Put a shunt in series with the motor

    A low-value resistor and an oscilloscope give current against time for the whole cycle. Current converts to torque through the torque constant, so this single trace produces the torque column of the duty table directly.

  2. Capture the whole sequence, including what happens between cycles

    Homing moves at power-up, retries after an error, jog moves during service and cleaning cycles all add starts. They are usually absent from the design duty and present in the field.

  3. Capture the worst unit, not the nominal one

    Mechanism friction varies across a production population and rises over life. The duty that matters is the one at the high end of that distribution, at the end of its service interval.

  4. Record the ambient the mechanism actually sits in

    Enclosed instruments run warmer inside than the room they stand in, and the temperature inside is what sets both winding rise and lubricant life.

The difference between the designed cycle and the measured cycle is often a factor of two or more in start count, and start count is the dominant wear counter for a brushed machine. Measuring it is a day of work that changes the life answer more than any component selection.

Record

Who wrote it, who reviewed it, when it was last checked, and what has been corrected.

Written by
Micro Motion application engineering
Reviewed by
Micro Motion manufacturing engineering
First published
Last reviewed
Reading time
8 minutes

Corrections

No corrections have been issued for this article. When a figure, a number or a statement here is found to be wrong, the correction is recorded in this block with its date rather than edited silently into the text.

Report an error in this article