Reliability assessment based on degradation measurements: How to compare some models?

• We propose a comparison of reliability estimation methods for mechanical parts.
• The comparison method uses simulation to generate linear degradation data.
• We study five methods to assess reliability using single linear degradation data.
• The assessment accuracy depends on censoring time and on the variation coefficient.
• Our method allows optimizing validation tests of mechanical parts for vehicles.

An important issue in the validation of mechanical parts for vehicles is reliability assessment for high mileages, by means of tests. Since the tests should be as short as possible, and for parts subjected to degradation mechanism, such as wear or crack propagation, it would be appropriate using degradation measurements (such as mass loss or crack length) in order to estimate reliability. In this study, we present some statistical approaches responding to this concern and propose a method to compare theses models. Different types of data can be available; in this paper, we only consider the case in which one measure is available for each part. Only linear degradation is studied.