As key components of electric multiple units（EMUs）brake system

sensor has significant impact on security and reliability of EMUs operation.Ensemble empirical mode decomposition（EEMD）overcomes the shortcoming of mode mixing of EMD

and the decomposed intrinsic mode functions（IMFs）prominent the local features of fault

and solve energy entropy

to form fault feature vector

then input to least squares support vector machine（LSSVM）for fault identification and classification

at the same time

particle swarm optimization（PSO）algorithm is applied for optimizing the structure parameters of LSSVM.The test result indicates that the fault diagnosis model has high correct classification rate for AS pressure sensor fault of EMUs.