The failure rate dynamics in heterogeneous populations

Most populations encountered in real world are heterogeneous. In reliability applications, the mixture (observed) failure rate, obviously, can be considered as a measure of 'average' quality in these populations. However, in addition to this average measure, some variability characteristics for failure rates can be very helpful in describing the time-dependent changes in quality of heterogeneous populations. In this paper, we discuss variance and the coefficient of variation of the corresponding random failure rate as variability measures for items in heterogeneous populations. Furthermore, there is often a risk that items of poor quality are selected for important missions. Therefore, along with the 'average quality' of a population, more 'conservative' quality measures should be also defined and studied. For this purpose, we propose the percentile and the tail-mixture of the failure rates as the corresponding conservative measures. Some illustrative examples are given.