Research On Software Fault Diagnosis Method For On-board Equipment Of CTCS Based On Timed Automata Theory

The on-board equipment software fault is an important factor affecting the train operation safety in Chinese Train Control System(CTCS),thus it is of great significance to perform fast,accurate,and different granular fault diagnosis on the logic functions of the on-board equipment software in the development cycle.The traditional on-board equipment software fault diagnosis mainly relies on methods such as offline testing and manual debugging based on test case sets,and in which there are problems such as insufficient real-time and incomplete fault detection,and low fault location efficiency.In this paper,according to the functional requirements of the on-board equipment software in design and realization stage of Next Generation Train Control System(NGTC),a set of software fault diagnosis method driven by model and data is proposed.The specific research contents include:(1)Fault analysis of the on-board equipment software.Combined with the functional structure and working principle of NGTC,the key functions and related interfaces of the on-board equipment are deeply analyzed.According to the typical software failure modes and the key variable behaviors of the interfaces,the fault modes related to the automatic train protection function are qualitatively analyzed.(2)Research on the fault diagnosis method of the on-board equipment software.Focusing on the functions and timing requirements of the on-board equipment software of NGTC in the design and implementation stage,the on-line fault detection method of the on-board equipment software is studied and it is based on the timed automata and timing consistency theory.Furthermore,the model mutation operator is used to analyze the diagnosability of the key function faults of the on-board equipment software and then a code-level fault location method for the on-board equipment software based on BP(Back Propagation)neural network is proposed.On this basis,a set of fault diagnosis framework driven by model and data is constructed for the on-board equipment software.(3)Research on the online fault detection of the on-board equipment software function.Combining the speed protection function requirements of the on-board equipment,the observable interface behavior of the fault detection process is defined.For the on-board ATP(Automatic Train Protection),a hierarchical function model including environmental functions such as train location update and MA(Movement Authority)is constructed and a variant fault model is built.According to the actual simulation software,an adaptation environment for fault detection applications based on the consistency theory is built.Fault diagnosability and online detection analysis are performed on the software defect versions of the on-board ATP,and the results are evaluated from the two perspectives of detection efficiency and detection accuracy.(4)Verification and analysis of fault location method for the on-board equipment software.According to the fault location method proposed in the paper,an application framework for analyzing the statement execution characteristics of the on-board equipment software is established.Based on the execution characteristic data and the BP neural network model,the fault location method is compared with three classic methods of Tarantula,Jaccard and Ochiai from the three aspects of single mutation version,multiple mutation versions and improved BP neural network algorithm.The experiment results show that the fault diagnosis method driven by the timed automata model and data can quickly and accurately find the on-board equipment software faults,and its fault detection effect is far greater than the offline mutation testing.At the same time,the fault location method based on the BP neural network can solve the problem that the model-based fault detection method cannot perform fine-grained location at the statement level.And this fault location method has a better location effect than the traditional method based on the program spectrum.In large-scale software fault versions,the fault location method based on BP neural network only needs to check 30%of the code to find the hidden defects in 80% of the faulty versions.The related results can provide a certain theoretical basis and technical support for the development of the on-board equipment software of NGTC.Figures 73;Tables 27;References 85.