Research On PHM Method Of Switch Machine Based On Digital Twin

As one of the basic equipment of railway,the normal operation of switch machine affects the safety and efficiency of train.At present,the maintenance method of switch machine still stays in "periodic repair" and "fault repair",which lacks the health management.Equipment cannot be effectively evaluated current operation state.Fault prediction and health management(PHM)has become one of the hotspots of equipment reliability in recent years.Through the monitoring,diagnosis and prediction,the intelligent management of equipment can be realized.With the emergence of digital twin,which provides the possibility for seamless integration between physical space and digital space,PHM receives technical support.Aiming at the problem of intelligent maintenance of switch machine,this thesis studies the PHM method of switch machine based on digital twin.The main work of this thesis is as follows:(1)According to the principle and working characteristics of switch machine,its common faults and manifestations are analyzed.Combined with the definition of digital twin,the feasibility of applying digital twin to switch machine is analyzed,and a PHM framework of switch machine based on digital twin is proposed.(2)Based on Solid Works platform,the bottom-up method is used to model parts.A3 D model of switch machine is constructed through part constraint and assembly;Based on Unity3 D and Visual Studio platform,import 3D parts,analyze the movement process of the switch machine,input the analog start signal into it by C# script,realize the positioning and inversion action of the whole machine and the action of corresponding components such as spindle,action rod and automatic shutter.Finally,the design and implementation of behavior model of switch machine digital twin are completed.(3)Under the models of Convolutional Neural Network(CNN)and Long Short Time Memory(LSTM),the fault diagnosis model of switch machine is established.Taking the historical data and fault data of the current signal of the switch machine as an example,the experimental verification is carried out.The results show that this method has higher accuracy than LSTM model and CNN model.It can fully capture the spatial and temporal characteristics of the signal,and has a better diagnosis effect.As a part of the rule model of the switch machine digital twin,it explains the state of the switch machine.Finally,the diagnosis results are visually displayed in the behavior model.(4)Under the Autoregressive Integrated Moving Average(ARIMA)model and LSTM model,the fault prediction model of switch machine is established.Taking the change of gap of switch machine as an example,the experimental verification is carried out.The results show that the ARIMA-LSTM combination model has higher prediction accuracy than the single model,and can predict the next time of the gap and the failure of the switch machine.The ARIMA-LSTM model combined with entropy weight method is taken as a part of the rule model of switch machine digital twin.It explains the evolution of poor adhesion fault.Finally,the prediction results of the rule model are visually displayed in the behavior model to provide intuitive information for maintenance stuff.