ABSTRACT
This chapter discusses several models that, combined with the acceleration modeling methods of Chapter 8, can be very useful for designing reliability life test experiments and analyzing reliability test or —eld data. Sections 9.1 and 9.2 deal with step stress experiments and degradation data analysis. These approaches extend the methods of Chapter 8 to situations where the analyst is uncertain of obtaining suf—cient failure data at the stress levels chosen for experimentation or where time and equipment exclude the multistress cell experiments described in Chapter 8. Section 9.3 extends the maximum likelihood estimate (MLE) modeling techniques used
in Chapter 8 to situations where the analyst needs to determine which of one or more categorical factors are inµuencing the T50 or α scale parameters. For example, is vintage a signi—cant factor and how by much does it change lifetime? Are there signi—cant differences in lifetime based on the factory of manufacture or the vendor from which a component was procured? Actions based on answers to questions like these can have signi—cant —nancial relevance, and as the software examples will show, lifetime regression provides a powerful tool for obtaining these answers. Section 9.4 looks at a model that is very popular in biomedical survival analysis modeling: the proportional hazards model. Section 9.5 explores data analysis when the failures seem to stop occurring long before
all the units on test (or observed in the —eld) have failed. Are there subpopulations which are defective in some way, causing some units to fail relatively early, while the rest of the (nondefective) units continue to operate until they reach their far-off wear-out times? These situations can be modeled using defect subpopulations. The accelerated life models of Chapter 8 can still be applied to the defect subpopulation failure data. Based on the authors’ extensive industrial experience, defect models are often appropriate and may be the most neglected and underused analysis approach in the literature.
