Research Of Fault Diagnosis Of The Micro-computer Monitoring System Based On Neural Networks

Currently, china railway tends to develop high speedily and heavily. The demands of the reliability of the signal devices and the driving security are increasing more and more. Therefore, it’s urgent to research on the significant issue, in order to ensure driving safely, that realize diagnosing the faults of the signal devices in railway, by new technique and method.In our country, the micro-computer monitoring system can inspect the signal devices in railway and record its major running status, and give an alarm in time when the devices become unusual. By this way, it provides a scientific foundation to the railroad branch in grasping the status of the equipments and analyzing the accidents. To a certain extent, the micro-computer monitoring system contributes to the traffic security, but it still can’t realize diagnosing the faults of the signal devices in railway and lakes the comprehensive function that combines the status supervision and fault diagnosis. It still relays on the skill and experience of maintenance staff to field investigation, judge and deal with fault. Consequently, it’s necessary to realize the fault diagnosis foundation of the micro-computer monitoring system.In connection with the complex, stochastic and nonlinear characteristics of railway signal equipment fault, the paper applies the neural network for the fault diagnosis of the computer monitor, using the nonlinear, self-learning ability and fault diagnosis of many advantages in the field, to achieve fault diagnosis of computer monitoring system.There is designed fault diagnosis process of computer monitoring system which based on BP neural network, through analyzing the failure mechanism of signal equipment to determine characteristic parameters, using status monitoring capabilities of computer monitoring system to obtain fault data. The fault data as training samples to train the neural network, and the network will automatically get the mapping. And then, it real-timely monitors the abnormal changes of characteristic parameters, using the characteristic parameters of the signal equipment as a sample to input the neural network, the network can output the best matching diagnosis.The combination of the fuzzy systems with the neural network is able to mix the advantages of the fuzzy logic and the neural network and can consider the ambiguity of the fault knowledge of the train signal device, in the same time, take advantage of the forced ability of self learning. The fuzzy systems and the neural network can accomplish the fault diagnosis together. This paper combined the fuzzy systems with the neural network. Firstly, it carries out the preliminary fault diagnosis.It used the fuzzy system as a pre-system of the neural network, fuzzed the fault data, to make more precise of training data, and to speed up the learning speed, improve the diagnostic accuracy. Using the concept of the fuzzy weights to calculate the weight of the fault symptoms, and then fatherly determine the weight of the fault, and conduct the precise fault diagnosis.It analyzed fault of switching capacity, switching capacity that requires manual intervention and analog fault diagnosis, described in detail the extent of the fault type and fault diagnosis process, and simulate the diagnostic process. The simulation results show that fuzzy neural network method applied to the fault diagnosis of microcomputer monitoring system is feasible, and can greatly enhance the training speed and diagnostic accuracy. Finally, the paper simulated the on-site examples of fault diagnosis, simulated results show that the fuzzy neural network method applied to the microcomputer monitoring system fault diagnosis is feasible, and the method has some practical value, it is able to promote within a certain range.