Research On Testing And Evaluating Theory Of Reliability And Safety For High Dependability Software

With the rapid development of computer technology and low costing in hardware, kinds of complex and powerful computer systems are applied extensively in safety-critical area, such as aviation and astronavigation, traffics and transports, nuclear power plants and medicine devices, where failures could result in loss of life, significant property damage, or damage to the environment. Therefore, the dependability increasingly becomes the focus, and the related software is called high dependability software. In fact, the software reliability is always so poor contrasting to the rapid development of hardware reliability that software becomes the bottleneck of the system dependability. Software testing plays two important roles in enhancing software dependability support. One of the roles is to eliminate the bugs so as to promote the software dependability truly, the other is to measure the current dependability level. Now, it is a hot topic for researchers including scholars and engineers to advance the software dependability effectively and demonstrate the dependability level objectively.In this dissertation, the origin and meaning of dependability is delivered in detail. In addition, the actions among the defect, error and failure are discussed at length. And the relation among the aspects of dependability is argued, as well as the dependability assurance techniques are neatened. All the work aims to aid understanding the testing target: high dependability software.Due to the fact that different applications focus on different aspects of dependability, the testing and evaluating methods for reliability and safety are mainly considered in this dissertation, as well as some systematic and deep study is conducted. The main contributions of this dissertation are as following:1) It is impractical to demonstrate some high reliability targets, due to the large testing case size and long testing duration. In order to reduce the total number of the testing cases but not to decrease the confidence level on the testing results for the reliability demonstration, a new method, which uses the Bayesian statistical inference with the prior knowledge dynamic integrating, is presented. In particular, the method is expanded in two instances: the discrete executing software and the continuous executing software. Numerical analysis shows that we can use fewer testing cases and shorter testing duration to achieve the same confidence level.2) In order to ensure the high reliability target, a reliability-testing framework is presented for the high dependability software. The main idea of the testing framework is that we can monitor and control the software reliability in the molule...