The safety and reliability of the traction seat plays an important role in the safety of locomotive operation,the research on the status identification of locomotive traction seat has important economic value and social significance.In order to carry out the research on the identification of the normal,small crack and large crack states of the traction seat,a simulation experimental platform of 〖JP2〗the locomotive traction seat was built.The acceleration sensor was used to collect the vibration signals of different states,and the time-〖JP〗domain features of the original data were extracted.The time-frequency features were extracted by ensemble empirical mode decomposition (EEMD),and the K-nearest neighbor algorithm (KNN) was used for the status identification of the traction seat.The experimental results show that the EEMD-KNN pattern recognition method can effectively recognize the locomotive traction seat state,and the rate of recognitionreaches 83.3%.After adding time domain features the recognition rate is higher,which reaches 90.5%.