Software reliability and safety: a real case study in the transportation area

摘要

The aim of this article is to present original software reliability method concerning transport domain with odometer system of safety application. When a new software is created, the phase of software validation is essential to development cycle. That's why some qualitative and/or quantitative analysis .of software potential errors, and of their effect, are achieved to estimate the software criticality. FAIVELEY TRANSPORT, as a software designer, performs this kind of analysis for the software validation and in particular has applied it for a speed measure station ERTMS (European Rail Traffic Management System). The study presented in this project is in particular supported by software of ERTMS odometer system. The SEEA (Software Effect and Error Analysis) is a qualitative method applied to software, for safety, in order to estimate it and to improve it, through the adoption of some recommendation emerging study. In fact, the fundamental principle of SEEA is to consider some software intrinsic hypothesis of design simple mistake, to deduce the effect from mistakes and to analyse the sturdiness of software compared with mistakes. The setting up of a software fault tree, with a top event representing the feared event, enables to extract the first order minimum cut set and to submit it to the SEEA method. In this case, the weak point emphasized by this method are improved with recommendations. In the first time, the program functional comprehension allows to set up the skeleton main part of the software fault tree, with the top event representing the top event and with top down analysis in the program code lines. In fact, some functions are not studied systematically because they are totally beyond the limits of the study. Indeed, the generation of software fault tree has to assume analysis hypotheses such as relevant life phases of the product or its operational process. As a result from this analysis, recommendations to introduce means of protection within the software have allowed to remove the most serious software errors and to assure a high level of software safety. This method has allowed to validate an original reliability process of real time data-processing application, through a joint approach with the SEEA and the modelling by fault tree, based on the rereading of source code.