Failure Prediction Of Railway Turnouts Using Textual Data

Railway turnout is a very important equipment in railway infrastructure,and it is also the most prone to failure.Its status will directly affect the availability and service quality of the entire railway system.Once it fails,it can lead to delay of trains,affect the efficiency of the train operation.More seriously,it may lead to derailment of trains and endanger traffic safety,property damage and casualties.Busy railway transportation also puts forward higher requirements for the maintenance of the turnouts,and the traditional planned maintenance method cannot meet the demand for operational efficiency.As an important means to reduce equipment maintenance costs and improve equipment intact rate,prognostics and health management technology has aroused more and more attention.Therefore,it is necessary to use scientific methods to predict the probability of equipment failure or the number of failures that will occur in a certain period of time.Replacing post-processing with prevention becomes an overwhelming trend.Traditionally,the acquisition of sensor data is costly,and the data is incomplete and contains noise.At present,railway fault records are mostly textual data,and these neglected rich textual data can be used as a good source of signal equipment analysis.Based on the maintenance record data from Guangzhou Railway Group,this paper first uses text mining technology to extract the real failure causes of railway turnouts,and then conduct research on failure prediction based on the main cause.After failure cause extraction,we find that weather is one of the most important causes of turnout failures,and about 30%of turnout failures are weather-related.Therefore,weather is selected as the failure factor.By combining the real-world fault records and the weather data,a Bayesian network model and an improved AdaBoost model for prediction are presented to explore the exact relationship between weather and turnout failures.Considering the rarity of failure events,the paper uses Noisy MAX model to predict the failure numbers of railway turnouts from small data sets.Considering the stochastic nature of failure events,this paper uses Monte Carlo simulations to weaken the randomness of failure events,and get good prediction performances.In addition,we compare the prediction results of models presented in this paper with other popular prediction methods.The results show that the Bayesian network presented in this paper has better prediction performance.Finally,we use the C#programming language to design and improve the railway turnout maintenance management module of the signaling equipment maintenance management system,merging the conditional probability table of the Bayesian network,at last to provide maintenance decision supports for the railway corporations.The main innovations of this paper are:(1)This paper presents a weather-related failure prediction model of railway turnouts for the first time.The weather-related failure causes and the number of failures were extracted from the textual records.The weather data and fault data were fused to analyze the exact relationship between the weather factors and the turnout failures.(2)This paper uses the text mining technology to extract the failure causes of railway turnouts,and obtains the real failure causes from real-world textual data.(3)This paper presents a Bayesian network-based failure prediction method for railway turnouts,designing the entropy minimization discretization technique to discretize the model variables,in order to get better prediction performance.(4)In order to avoid the inaccuracy of forecasting due to the small data set,the Bayesian network is transformed into an independent causal interaction model,and then uses the Noisy MAX model to reduce the number of parameters needed to learn,and uses Monte Carlo simulations to weaken the influence of the random nature of failure events,finally improve the prediction accuracy.The research in this paper can provide guidance for the procurement of equipment and spare parts inventory,avoid unnecessary waste and improve the utilization rate of equipment.In the aspect of equipment maintenance,it can reduce the improper maintenance and improve the service life of equipment.It can also help the signaling&telecommunication department to develop the next period of maintenance management plan and provide guidance to improve the efficiency of train operations and security.