| X-ware systems are defined as complex systems that consist of interacting X-ware components of hardware, software and human. Many of modern systems are X-ware systems.; Though reliability of hardware, software and human components has been extensively studied separately, reliability of X-ware systems lacks necessary attention. Unlike hardware systems whose failures are mainly caused by components, failures in X-ware systems not only come from component failures but also from complex interactions.; In this dissertation, a definition for X-ware systems is given. Characteristics of X-ware systems are discussed. Comparisons between hardware, software and human are made. Big differences in many aspects among them make it hard to integrate them to form system reliability. As a very important element in complex systems such as X-ware systems, interaction is intensively studied. A classification from different angles is given and is supposed to identify as many interactions as possible. Input-output model (IOM) and sharing resource mode (SRM) are proposed to identify failure modes in interactions. Interaction network (IN) as a generalized network model was developed. By matrix representation of interaction networks, interaction paths and cycles could be identified. Algorithms were designed to obtain minimal cut sets. Minimal cut sets from different networks can be linked. As a systematic methodology, a hierarchical failure mode and effect analysis (X-FMEA) of X-ware systems is proposed. X-FMEA is different from traditional FMEA in two things: the procedure is hierarchical in nature and it includes interaction into analysis. To evaluate X-ware system reliability, comparison of modeling powers of widely used models was made. A new approach called simulation model (SM) was designed to evaluate X-ware system reliability. SM is not only able to describe the complex interactions among components but also avoids exponential explosion in size. It is easy to be built. With help of computers, the analysis is conducted through Monte Carlo simulation. |
