| This thesis deals with statistical testing as a means for software dependability demonstration for safety-critical military avionics systems. Statistical testing is currently the only method accepted as being mathematically valid for quantitative assessment and demonstration of software reliability, and hence software dependability. Other methods, such as Software Reliability Growth models, Fuzzy Software Reliability models. Error Seeding, Testability Analysis, Coverage Analysis, and Conditional Correctness have theoretical and practical deficiencies, and are controversial to various extents. Various approaches to statistical testing for reliability demonstration of hardware or software are compared: Laplace's rule of succession, TRW Software Reliability Theory, Bayesian estimation, Life Testing, MTBF assurance testing, Bayesian Zero Failure Reliability Demonstration Testing, Probability Ratio Sequential Testing (PRST) and the t-distribution method. It is shown that the most efficient method is PRST. An original approach for statistical testing of safety-critical systems, Single Risk Sequential Testing (SRST), is developed and demonstrated to be more efficient than PRST. An original model for software reliability demonstration of safety-critical systems is developed, based on the SRST method and the Test, Analyse And Fix (TAAF) method. The model treats reliability demonstration as a multi-stage process, and not as a one go/nogo trial, and therefore results in an optimal testing policy, depending on the circumstances. Two optimal testing policies are developed by solving two optimization problems. The necessary conditions for using the model are examined. It is shown, that the model is most suitable for dependability demonstration of safety-critical military avionics systems. Hence an efficient testing methodology is developed, taking into account some specific characteristics of such systems. Finally, the model and the methodology are used to design cost-effective test plans and test set-ups for two real military avionics systems. |
