On Automatic Crash Model Translation

This paper discusses some issues relevant to automatic crash model translation (conversion), based on SimTech experience. Automatic crash model conversion is in many ways similar to automatic language translation. The first level of conversion is “literal” translation, where each entity in the source model is translated into a corresponding entity in the target model. This issue is already complicated because such a target entity may not exist or may not be uniquely defined. Such a literal translation may not run on the target code, exactly as a literal translation of a text can be incomprehensible. The second level of conversion is therefore to obtain a model which runs, in the sense of not containing contradictory cards which would block the target code. The third level of conversion seeks to obtain a model which gives the same result on the target code as on the source code. The methodology described in the paper is implemented in an ENKIDOU application. We show the conversion of a relatively complex model and the comparison between the results of the original (source) and converted (target) model.