Research On Fault Diagnosis Method Of Switch Equipment Based On Random Forest

Railway signal equipment is a key facility to ensure the safety of traffic and improve transport efficiency.Switching equipment is a vital but basic equipment in the railway signal system,and its condition directly affects the smooth flow of railway transport,and is closely related to the safety of railway transport.Currently,the maintenance and repair approach taken for turnout equipment is basically a hybrid approach,which is a mixture of daily maintenance combined with cyclical overhaul during the sunroof period.Due to the low level of intelligence of the current railway signal monitoring system,the traditional method of manual analysis of fault alarms by maintenance personnel is still used on site when turnout equipment is faulty.On the one hand,this method relies excessively on the practical experience and professional knowledge of the maintenance personnel on site.On the other hand,it can not meet the needs of today’s rapid railway development in terms of diagnostic efficiency.The study of intelligent switch equipment fault diagnosis method is of great significance to the improvement of signal equipment repair and maintenance level.In this paper,taking ZD6 switch machine as an example,based on the analysis of its failure causes and failure modes,the random forest algorithm is adopted as a fault diagnosis method for turnout switching equipment,and the actual monitoring data is used for verification.The main work of the thesis is as follows:(1)Failure mode analysis and database design for turnout switch equipment.Based on the mechanical and electrical principles of the turnout equipment,the common failure modes are labeled with serial numbers based on historical statistical reports and frequency of occurrence,and the corresponding causes of failure are analyzed under each fault category label to give recommendations for maintenance.After screening to separate faulty and normal data from the monitoring system,the data was pre-processed for sampling imbalances in the faulty data.Based on the historical expert experience database,add the fault and normal data that have been processed and separated,and merge to create a database of fault cases in turnout conversion equipment in this paper.Make data structure definitions as required to meet machine learning data requirements,complete database design and give E-R diagrams.(2)Research on fault diagnosis methods for turnout switch equipment.By analysis the characteristics of turnout equipment,this paper studies the fault diagnosis method of turnout conversion equipment based on random forest,giving the theoretical analysis and model building process.The old-fashioned way of extracting features for action processes in segments is abandoned and a non-segmented approach is used to extract parameters as the feature set to be selected.The model training set,test set and validation set data were selected in the ratio of 6:2:2,and the model was optimized using the generalization error as the tuning benchmark,and finally a comparison experiment with other algorithms was conducted.(3)Design of software for the fault diagnosis function of turnout changeover equipment and its interaction process.This paper establishes the functional requirements for fault diagnosis,and designs the functional software architecture and user interaction process.And the data of actual turnout equipment in a station of Urumqi Bureau is used as a sample for verification.The research results of this paper are a highly accurate random forest-based fault diagnosis algorithm for turnout switching equipment and a front-end software relied on the algorithm.The validation demonstrates that the method provides an intelligent refinement of turnout equipment fault diagnosis and offers an expert experience perspective for the accumulation of experienced faults,which will guarantee for the safe operation of turnout equipment in the future.Figure 43,Table 12,Reference 52.