Research On Key Technologies In Component-based Software Heuristic Testing And Reliability

With the development of software, software grows both in size and complexity. Theidea on the software system design is also changed from the self-sufficient to the socialcooperative. The social cooperative software design requires the software functions arecomponent-based and standardized. This makes the software development method onthe component-based design and reuse become important. Effective software testing andreliability evaluation on the component-based software system become the hot topic ofthe software engineering research.The components in the component-based software system are numerous, functionindependent, widely originated, updatable and replaceable. Therefore it is difficult totest the component-based software system and make reliability evaluations. Since thecharacteristics of the high complexity and the low coupling exist in thecomponent-based software system, PSO (Particle Swarm Optimization) software Z-pathcoverage testing method and the back-propagation neural network model for reliabilityevaluation are proposed in this work. Some innovative contributions of the thesis areenumerated as follows:1. A self-adaptive inertia weight PSO test case generation algorithm (AIC-PSO) inclustering is proposed, since PSO test case generation algorithms are suffering fromprematurity led by clustering, which decreases the test case generation efficiency.AIC-PSO algorithm quantizes and monitors the test case clusters in the swarm. Basedon the cluster quantization, the AIC-PSO algorithm adjusts the inertia weights of the testcases dynamically to increase the variation of the test cases in the search space.AIC-PSO prevents the PSO algorithm from trapping into the prematurity and enhancesthe efficiency of the test case generation.2. The problem of computational resource optimization and the lack of test pathinformation exchange exist in PSO software multiple path structural cover testing. Sincethat, a multi-path oriented PSO automatic test case generation algorithm (MPPSO) isproposed in the multi-path software testing. The proposed algorithm can fully exchangthe information in the process of covering test paths, optimize the computational resources and enhance the test case multi-path coverage.3. Most of the available component-based software reliability models consumehigh computational cost and suffer from the reliability evaluating complexity for thesoftware system with high complex structures. Since that, a component-basedback-propagation reliability model (CBPRM) with a low complexity is proposed basedon the artificial neural networks. CBPRM optimizes the software system reliabilitydynamically depending on the artificial neural networks, the component historyreliabilities and the correlative sensitivities. CBPRM has a linear increase evaluationcomplexity related to the component quantity and outperforms the state-based modeland the path-based model of the reliability evaluation with robustness to the softwaresystem structure.4. A component-based hierarchical Web reliability evaluating model is proposedfor the component-based Web system. According to the reliability analysis requirements,the proposed model divides the component-based Web system into the system workinglayer, the log generation layer, the reliability extraction layer, the reliability convergencelayer and the reliability evaluation layer. The reliability evaluation layer uses thecomponent reliabilities in the Web system as the input of the back-propagation artificialneural networks and output the evaluation result of the overall Web system. Theproposed model can satisfy the requirements of the heterogeneity, the distribution andthe loose coupling of the component-based Web system. This model can also evaluatethe overall Web system reliability in the low evaluation complexity and the acceptableevaluating accuracy.