Search

On-board subsystem is a key of guaranteeing traffic security and enhancing operational efficiency for highspeed railway (HSR). Thus, the test of the subsystem is the main means of ensuring both the functionality and reliability of the train control system. To deal with the low efficiency of existing manual mode, an automated test data generation method combining back propagation neural network (BPNN) and multi-population genetic algorithm (MPGA) is presented. First, colored petri net (CPN) models are designed according to the relevant requirements specifications of the on-board subsystem. The feedback information of each model is taken as the fitness function. Second, the BPNN method is adopted to simulate the solving process of individual fitness value, which is to reduce the running cost of the designed model. Third, to overcome the shortcomings of genetic algorithm (GA), the MPGA is used to search for test data in the input space. Finally, we take the startup procedure of the on-board subsystem as a case study. The results reveal that the novel method has the advantages of short running time, higher efficiency, and better stability compared with the traditional GA. [ABSTRACT FROM AUTHOR]