屏蔽数据下带变点的软件可靠性增长模型

SOFTWARE RELIABILITY GROWTH MODEL WITH CHANGE POINT ON MASKED DATA

屏蔽数据是可靠性工程中一种常见的缺失性失效数据。在考虑屏蔽失效和多变点的情况下,建立基于非齐次泊松过程的软件可靠性增长模型。该模型利用C-Chart方法估计软件失效过程中的变点位置;详细推导可靠性模型参数的极大似然过程,利用EM(Expectation Maximization)算法解决似然函数极其复杂的问题;利用Tomcat 5软件的失效数据进行模型性能对比分析,结果显示提出的可靠性模型具有较好的效果。

Masked data is a common missing failure data in reliability engineering. Considering shielding failure and multiple change points, this paper establishes a software reliability growth model based on non-homogeneous Poisson process. The C-Chart technology was used to estimate the position of the change point during the software failure process. The maximum likelihood estimation(MLE) process of the model parameters was derived in detail, and expectation maximization(EM) algorithm was applied to solve the likelihood function complicated problem. We adopted Tomcat 5 software failure data to conduct a comparative analysis of model performance. The result shows that the proposed reliability model has better effect.