A New Software Reliability Evaluation Model Based on the Probability of the Failure Mode

Abstract

Most existing software reliability models are only suitable for the reliability testing based on the operation profile which has the low fault detection rate. The reliability testing based on the failure modes can improve the fault detection rate. However, it can't describe the reliability accurately for it doesn't have the related reliability models. To solve this problem, we propose a new reliability model for the reliability testing based on the failure modes. First, the probability of the failure mode is obtained according to the operation profile and the input value domain. Based on the probabilities of the failure modes, the failure rate of the function is calculated. Furthermore, the function transition matrix with the failure rate is constructed by Markov process. Then a new software reliability model is proposed based on the probabilities of the failure modes. We apply this new model on the engine control software. The experimental results showed this model can process the failure modes for calculating the reliability accurately. Meanwhile, the fault detection rate can be improved and the testing effort can be reduced significantly.