Accelerated Life Test Plan Under Schedule Pressure
A team must demonstrate a 5,000-hour B10 life for a new bearing in a gearbox. A full-life test at use conditions would take too long, so the team proposes an accelerated test at elevated load and temperature. They have 12 test units and 8 weeks of test time. Preliminary analysis suggests an inverse-power load-life exponent of about 3 and an Arrhenius temperature acceleration. Marketing wants a single number for the datasheet; the test engineer wants to run at the highest acceleration possible to finish on time. A customer requires evidence that the B10 claim is supported.
1. What are the key risks in running at the highest possible acceleration, and how should the acceleration level be chosen?
Read the worked answer
Risk of over-acceleration.
- Failure-mode shift. At high load or temperature the dominant failure mechanism can change (e.g., from subsurface rolling-contact fatigue to surface distress, or from bearing wear to lubricant breakdown). The test then measures a different mechanism than the field and the extrapolation is invalid.
- Invalid acceleration model. The inverse-power exponent (~3) and Arrhenius factor are assumptions. If the exponent is wrong, extrapolation from high stress to use stress is badly biased. Over-acceleration amplifies extrapolation error.
- Physical limits. Lubricant film, cage speed, and thermal limits may be exceeded, producing artifacts.
How to choose the level.
- Run a preliminary screening at two or three stress levels with a few units to confirm the same failure mode (post-test metallography, wear pattern) and to estimate the acceleration exponent from the data rather than assuming it.
- Set the acceleration so that the failure mechanism is unchanged and the extrapolation distance is modest.
- Document the model, the assumed exponent, and the confidence bounds; state that the B10 claim depends on those assumptions.
Bottom line. Highest possible acceleration is the wrong objective; the right objective is the highest acceleration that preserves the failure mechanism and supports a defensible model.
2. Given 12 units and 8 weeks, outline a test plan that produces defensible evidence for the 5,000-hour B10 claim, and state what you would tell marketing.
Read the worked answer
Plan elements.
- Screening phase: 2-3 units at two stress levels for 1-2 weeks to confirm mechanism and estimate the load-life exponent.
- Main test: remaining units at the chosen acceleration level, run to failure or to a defined suspension time, with time-to-failure recorded (not just pass/fail). Use a Weibull analysis with censoring.
- Acceleration model: apply the inverse-power load relation (exponent from screening, ~3) and the Arrhenius temperature factor; combine into an acceleration factor and translate test hours to equivalent use hours.
- Statistics: with ~10 units, the B10 estimate will have wide confidence bounds. Report B10 with a lower confidence limit, not a point estimate.
- Verification: include at least one unit at or near use conditions as an anchor, if schedule allows.
What to tell marketing. Do not publish a bare number. State the B10 estimate, the confidence level and interval, the acceleration model and assumptions, and the test conditions. If the lower confidence bound does not support 5,000 hours, the claim is not yet demonstrated and more test time or units are needed. A single datasheet number without the model and confidence is not defensible evidence for the customer requirement.