Abstract: Software faults are the underlying and important roots which result in the mistake, failure and even break-down of system. Therefore, the fault diagnosis technology is very significant to software quality assurance. Recently, the fault diagnosis technology based on the artificial intelligence theory attracts more and more attention. Because I3ayesian networks theory has some significant advantages, such as easy expression and precise reasoning, a software fault diagnosis model with four layers(namely, reason, mode, fault and watch, short for WCMF) was firstly proposed by combining Bayesian networks with SFMEA(Software Failure Modes and Effect Analysis). Secondly, a software fault diagnosis framework based on the fault information database and the fault diagnosis database was presented. Finally, a case study on navigation software was proposed. The results shown that the fault diagnosis model and presented framework is feasible and effective,which have the following advantages,such as timely and convenient fault diagnosis,improving diagnosis efficiency by utilizing the history information of the faults and their diagnosis.

Key words: Bayesian networks, SFMEA, Software fault diagnosis, WCMF