Study On The Application Of Software Reliability Design Technology

With the rapid development of computer technology, weaponry systems andautomated command systems and other military systems depend much more onsoftware than before. Software plays an increasingly important role in weaponry,space, navigation and other systems which need high reliability. Therefore, militarysoftware reliability becomes bottleneck and key factor of military system.Software reliability is the most important attribute of software quality. Softwarereliability design technology is an important means to ensure and improve softwarequality. Therefore, this dissertation takes the project’s software as main object ofstudy, the application of software reliability design technology is carried out. Basedon the software development cycle, we can start reliability design from three aspects,which are software reliability design requirements, detailed software reliabilitydesign, code written specification. And then failure data are collected, andexperiments and assesses software reliability growth are performed The maincontributions of this dissertation are described as follows:Firstly, the in-depth studies of software reliability design methods and softwarereliability design criteria are performed. At the stage of software requirementsanalysis and outline design stage, software reliability design requirements for theproject software should be proposed. Secondly, based on the Failure Mode, Effects and Criticality Analysis (FMECA)and Quality Function Deployment (QFD), system-level house of software reliability(HoSR) model is presented. At the stage of detailed design stage, a common failuremode and causes of failure library should be established, and it should be constantlyfilled. Based on the library, HoSR is used to analyze software reliability of unitmodule, which help us find potential failure modes and the causes of failure.Common mode failures, common cause failure, the relationships between the failuremodes and the relationships between the causes of failure can be intuitively found,thus module itself defect can be controlled. The failure of transmission between themodules can be controlled, and passed defect can be controlled. Subsequently, defectby the coaction of module itself defect can be controlled. Finally, improvementscould be proposed based on the causes of failure. Reliability design could be guidedby the results of the analysis, for the result that software reliability could beimproved.Thirdly, intelligent optimization algorithm is combined with dynamic fuzzyneural network (DFNN), and the optimal value of DFNN parameters is gotten byimitative prediction (IP), accordingly, software reliability growth prediction model isestablished. Before acceptance of software, software reliability prediction model isestablished by DFNN. DFNN is trained by defect data and IP is proceeded. Duringthe IP, intelligent optimization algorithm is used to get the optimal value of DFNNparameters. And then DFNN which has been trained is used to predict softwarereliability, compared with G-O model, BP neural network and fuzzy neural network,its prediction error is small and steady.