Abstract: To solve the problem of real-time massive data parsing and model computing of EMU PHM, a PHM model processing framework for EMU based on stream computing is presented. The data processing flow of the EMU was analyzed, and the overall framework of the PHM model processing of the EMU was given based on Spark Streaming. For real-time massive data parsing and processing, the in-vehicle data structure before parsing was analyzed, the parsed in-vehicle data structure was defined. A universal data parsing component was designed, and the implementation of stream computing was given. For model computing, the formal definition of PHM model was given in detail at first, including the basic information, input, output and logical body of the model. Based on this definition, the common components of the model were designed, which enabled the rapid development, efficient computing and unified application of the model. Through the effective application of the EMU PHM system, it is proved that the framework can meet the real-time computing demands of large amounts of data very well.