研究生学术英语演讲示范.ppt

Railway Track Circuit Fault Diagnosis Using Recurrent Neural Networks李超李超171110362024/5/25 周六 IEEE Transactions on Neural Networks&Learning SystemsPublished in 2017Tim de BruinKim VerbertRobert BabuskaFault Diagnosis,Track Circuit,Recurrent Neural Network.structureA brief introduction on railway track circuitsThe motivation of this paperThis papers work A brief introduction on railway track circuitsFunction:detect the absence of a train in a section of railway track.No trainA train inTrain outSignal curve when a train pass a track circuit The motivation of this paper(1)Insulated joint defect(2)Conductive object across the insulated joints The motivation of this paper(3)mechanical rail defectThe motivation of this paper(4)electrical disturbanceThe motivation of this paper(5)ballast degradation The motivation of this paper(1)insulated joint defect(2)conductive object (3)mechanical rail defect(4)electrical disturbance(5)ballast degradationCause Failure in Railway Track Circuit,which will lead to disastrous safety problemsThe motivation of this paper(1)insulated joint defect(2)conductive object (3)mechanical rail defect(4)electrical disturbance(5)ballast degradationDepend on manual measurementCost much time on fault diagnosis in an emergencyThe motivation of this paper Therefore,in order to avoid artificial participation and shorten the fault detection time,the author of this paper hopes to use Neural Network to automatically and quickly identify these faults.The motivation of this paper This papers workNeural Network Memory unitInput&Output unitsControl unit This papers workNeural Network This memory neural unit allows the trained neural networks to have better memory effects and faster training speeds Data input1st neural layer2nd neural layerOutput Result1500 sequencess(Simulation and actual data)results of RNN and CNNRNNsAccuracy:99.7%CNNsAccuracy:99.2%(THE ROWS INDICATE THE TRUE CLASS AND THE COLUMNS REPRESENT THE PREDICTED CLASS)Thank you！