In order to extract the characteristics of the vehicle communication network fault data in different operating states and test the network performance under different performance degradation degrees, so as to facilitate the research of network fault diagnosis and health management, a MVB network fault injection device based on FPGA was designed. By studying the generation mechanism of the vehicle network faults, fault injection such as MVB network impedance mismatch, intermittent disconnection, and data frame error was realized. At the same time, taking latch metastable structure as random number entropy source, a real random number generation model based on LFSR + latch structure was proposed to simulate the occurrence of incidental and random faults.Finally, the MVB network test platform was built to further verify the feasibility of fault injection method by collecting the MVB network waveform data and extracting feature quantities under various fault modes.