Software Reliability Assessment of Safety Critical System Using Computational Intelligence

In the recent past, automotive industries are concentrating on software controlled automatic functions for its safety operations. The automotive safety and reliability lie in its design, construction, and software implementation. To assess the software reliability, the hidden design errors are classified and quantified. The temporal characteristic of numerical error is analyzed and its probabilistic behavior is explored using a novel framework called software failure estimation with numerical error (SFENE). Here, a model is devised to assess the probability of occurrence of the numerical error and its propagations from the initial to various other states using a Hidden Markov Model. It is seen that the framework SFENE supports classifying and quantifying the behavior of numerical errors while interacting across its system components and aids in the assessment on software reliability at design stage. The sensitivity and precision are found to be satisfactory. This attempt will support in the development of cost effective and error free safety critical software system.