Research On Remaining Useful Life Prediction For Valve Components Of DK-2 Brake Based On Data-driven Method

The DK-2 brake is used to realize the braking and releasing functions of the train,and valve components are its key executive components.The relay valve and high-speed solenoid valve are the two most common valve components in the DK-2 brake.Accurately predicting the remaining useful life of the relay valve and high-speed solenoid valve is meaningful to ensure the safe operation of trains.Due to the complex working mode and failure mechanism of the DK-2 brake,it is challenging to predict the remaining useful life of valve components.Therefore,this paper aims to achieve the accurate prediction of the remaining useful life for valve components of DK-2 brake,and studies the remaining useful life prediction method of relay valve and high-speed solenoid valve based on data-driven.The main contents include the following aspects:Firstly,in order to solve the problems that the early failure characteristics of the relay valve are not obvious and the remaining useful life prediction accuracy is greatly affected by the noise of the monitoring data,a remaining useful life prediction method for the relay valve based on extreme gradient boosting tree is proposed.The pressure signals of the brake pipe and the equalizing reservoir are collected by pressure sensors,which are decomposed into sub-signals of different frequency bands by ensemble empirical mode decomposition.And the noise-containing sub-signals are denoised by threshold noise reduction.Then,the alternative fault features of the relay valve are extracted.The Shapley value method based on game theory,which is used to reduce fault features dimensions.The extreme gradient boosting tree model is constructed,and the mapping relationship between the sensitive fault features and the life state of the relay valve is established,so as to realize the remaining useful life prediction of the relay valve.Secondly,for the problem that it is difficult to accurately predict the remaining useful life of the high-speed solenoid valves through a single type of data,a remaining useful life prediction method for the high-speed solenoid valve based on multi-sensor data fusion is proposed.The current sensor and pressure sensor are used to collect the drive current of the high-speed solenoid valve and the pressure signal of the equalizing reservoir,the timedomain and time-frequency domain features of current signals and pressure signals are extracted by ensemble empirical mode decomposition.The self-organizing map algorithm is used to fuse multi-dimensional features into a new health indicator that characterizes the life state of the high-speed solenoid valve.A hybrid model integrating similarity model and support vector regression model is constructed,and the weight parameters of the hybrid model are adaptively adjusted through the prediction error of the single model to achieve accurate remaining useful life prediction of the high-speed solenoid valve.Finally,this paper designs and develops the remaining useful life prediction system for valve Components of DK-2 Brake.Feasibility of the remaining useful life prediction method of valve components is verified by laboratory DK-2 brake system function simulation test platform.