Model reduction algorithm for fast neutrality tests and fault localization of Simulink models
A minor change of a Simulink model can result in an unexpected consequence, so the Simulink model is usually required to be rerun and tested, which increases the development cost and time. Compared with the reference model, only the changed parts of the updated model could result in a failure at the outputs. So, a two-stage model reduction algorithm is designed to isolate the changed parts, that speeds up the processes of neutrality test and fault localization. The first reduction is based on the changed parts, the second reduction is based on the bad outputs. The changed parts and the bad outputs are the blocks of interest of the reduction. The blocks related to the blocks of interest are reserved, the others are deleted. The thesis proposes a way of conversion of the Simulink model to a digraph based on extended data dependence to find the related blocks. After the model reduction, the faults are located with the help of signal comparison