Reliability demonstration testing method for embedded operating systems

摘要

The reliability target is becoming higher and higher for modern embedded operating systems with their being extensively used in safety-critical situations. In order to solve the problem that the fixed duration testing method, which was based on the classical statistics, can't satisfy the requirements of reliability demonstration testing for modern embedded operating systems due to the long testing duration, a new method based on the empirical Bayesian inference is presented. Firstly, the prior distribution of the embedded operating system failure intensity is derived on the basis of analyzing the testing records of reliability growth testing phase. Then, the reliability demonstration testing duration can be calculated by the prior distribution. Finally, the prior knowledge dynamic integrating approach taking both the specific testing information and the prior knowledge into account is developed. Experiment shows that the method introduced above can reduce the testing duration effectively without decreasing the confidence level in the testing results.