Degradation Models
Degradation Models (Accelerated Degradation) · Also known as: Accelerated Degradation Testing, Degradation Path Models, Performance Degradation Analysis, Bozunma Modelleri
Degradation models estimate product lifetime by tracking measurable performance characteristics—such as crack length, light output, or insulation resistance—over time rather than waiting for outright failure. Introduced in rigorous form by Meeker, Escobar, and Lu (1998), these models fit a stochastic degradation path to repeated measurements and define failure as the first time the characteristic crosses a predetermined threshold, enabling reliable lifetime inference from accelerated test data with very few or no observed failures.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use degradation models when units rarely or never fail during a feasible test period yet exhibit a measurable, monotone performance decline. Key assumptions are that the degradation path follows a parametric form, the critical failure threshold is physically meaningful and pre-specified, and degradation measurements are obtainable without destroying the unit. The method is less appropriate when degradation is non-monotone, when multiple competing failure modes exist, or when no suitable measurable characteristic can be identified. Alternatives include accelerated life testing (when failures occur) and Wiener-process or gamma-process models for non-linear stochastic degradation.
Strengths & limitations
- Enables lifetime inference from tests with few or zero observed failures
- Makes efficient use of continuous degradation measurement data rather than only pass/fail outcomes
- Allows extrapolation from accelerated stress levels to normal use conditions through well-defined acceleration factors
- Provides physically interpretable model parameters tied to the actual degradation mechanism
- Requires a measurable degradation characteristic that correlates reliably with the failure mode of interest
- Parametric path assumptions (linearity, power-law) can be misspecified, leading to biased lifetime extrapolations
- Extrapolation to use conditions assumes the acceleration model holds across the entire stress range tested
- Threshold d_f must be defined a priori; wrong threshold choice directly distorts all lifetime estimates
Frequently asked
How many test units are needed for a reliable degradation model?
There is no universal rule, but Meeker and Escobar recommend at least 10–20 units per stress level to estimate both fixed path parameters and unit-to-unit random effects with adequate precision. Fewer units lead to wide confidence intervals on lifetime quantiles, especially in the tails. Sample size planning software or simulation studies are advisable before committing to a test plan.
Can degradation models handle non-linear degradation paths?
Yes. The mean path μ(t; β) can be any parametric function—power-law, exponential, logarithmic, or Arrhenius-based—as long as it is monotone and the parameters are identifiable from the data. Alternatively, Wiener-process and gamma-process models provide fully stochastic non-linear frameworks. The key requirement is that the chosen functional form be physically motivated, not selected purely for empirical fit.
What is the difference between accelerated degradation testing and accelerated life testing?
Accelerated life testing observes actual failures at elevated stress and models time-to-failure directly; it requires enough units to fail. Accelerated degradation testing measures a performance characteristic over time without requiring failure, then extrapolates to a failure threshold. Degradation testing is preferred when failures are extremely rare, but it requires a reliable measurable surrogate for the failure mode.
Sources
- Meeker, W. Q., Escobar, L. A., & Lu, C. J. (1998). Accelerated degradation tests: modeling and analysis. Technometrics, 40(2), 89–99. DOI: 10.1080/00401706.1998.10485191 ↗
How to cite this page
ScholarGate. (2026, June 2). Degradation Models (Accelerated Degradation). ScholarGate. https://scholargate.app/en/reliability/degradation-models
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Maintenance OptimizationReliability↔ compare
- Reliability AnalysisReliability↔ compare
- Weibull RegressionSurvival↔ compare