Research On Prediction Method Of On-board ATP Failure Rate

On-board ATP is one of the key equipments in railway signal system.ATP failure not only directly threatens train operation safety,but also reduces the transportation efficiency of railway system,so it is particularly important to carry out scientific and reasonable maintenance work for on-board ATP.However,at present,there are some shortcomings in the ATP maintenance work of relevant railway departments,such as unreasonable maintenance spare parts reserve and unbalanced distribution of maintenance resources.These shortcomings may lead to ATP failure can not be dealt with timely and effectively,thus affecting the normal operation of the train.Therefore,in view of the above problems,this paper proposes a failure rate prediction model based on MEA-Chaos-Elman to accurately and effectively predict the variation trend of ATP short-term failure rate in the future,so as to help the relevant railway departments rationally optimize the reserve of ATP maintenance spare parts,dynamically allocate maintenance resources to improve the safety and protection capability of ATP equipment,thereby ensuring the safety of train operation and transport efficiency.In this paper,the original failure rate time series is obtained by using the actual fault statistics of a certain type of ATP equipment in the whole road network,and the time series is denoised by the wavelet threshold denoising method to improve the reliability of the data and the prediction effect of the subsequent model.Then,in view of the chaotic characteristics of the denoised time series,it is difficult to mine the hidden data variation law in one-dimensional time domain.The C-C algorithm and G-P algorithm of chaotic theory are used to calculate the time delay r and embedding dimension m respectively,and the one-dimensional failure rate time series is extended to the high-dimensional space through phase space reconstruction,thus restoring the chaotic attractor and its hidden data variation law of the original sequence.By calculating that the maximum Lyapunov exponent is greater than 0,it is further verified that the failure rate time series is indeed chaotic.After that,the short-term failure rate of ATP is predicted by using chaos theory-related prediction model,Elman neural network prediction model and Chaos-Elman combined prediction model respectively.According to the prediction results,it is proved that the combined Chaos-Elman model has better prediction effect because it takes advantage of two theories at the same time,and its accuracy can reach about 90%.Finally,in order to further reduce the prediction error of ATP failure rate,aiming at the problems existing in the established Chaos-Elman model,the weights and thresholds of Elman neural network are optimized by using the Mind Evolutionary Algorithm,so as to further construct the final MEA-Chaos-Elman failure rate prediction model based on the previous work.According to the comparison of simulation results,MEA-Chaos-Elman model has the best prediction effect and possesses accuracy and stability at the same time,and its prediction accuracy of ATP short-term failure rate can reach about 95%,which can provide a scientific and reasonable quantitative reference for the relevant railway departments.Thesis contains 40 figures,8 tables,52 references.