9. Reliability Prediction

Reliability prediction of a system estimates the probability that the system fails in a certain time interval given system context information. To this aim, several standards based on different data, failure modes, and models have been developed.The chapter provides a tabular comparison of standards as well as formal expressions that are necessary to understand their structure. It defines the term reliability as a time-dependent probability expression of a system. Related terms like availability, dependability are distinguished. The following terms are related to each other: failure density distribution, failure probability, reliability, and failure rate. Expressions are provided for the Weibull, exponential, and lognormal distributions. It is shown that most reliability prediction standards only use constant exponential failure rates as opposed to the more realistic bath-tube distribution. Multiplicative, additive, and mixed reliability prediction standards are distinguished, also standards that consider components only or standards that in addition consider system development conditions. Existing standards are compared regarding, inter alia, their coverage of types of environmental stress, component functional loading and mission profiles of systems under consideration as well as failure models used, and component and technologies covered. To this end, a comprehensive list of comparison items for standards and related software tools is provided. The chapter shows that reliability prediction complements the range of system analysis methods by providing frequency of event input for fault tree analysis (FTA), failure mode and effects analysis (FMEA), and Markov analysis. As field data is more easily accessible due to increasing sensing, intelligence, and connectivity, it is expected that scaling, updating, and even generation of empirical reliability prediction become feasible in ever-increasing application domains.