Abstract: As an important equipment for simulating the space environment, vacuum pumps are responsible for obtaining and maintaining the high vacuum environment inside the test container. Therefore, timely and accurate diagnosis of the operating status of vacuum pumps is particularly important. In response to the above issues, this article proposes a joint denoising method combining empirical mode decomposition and wavelet packet thresholding, which uses BP neural network to conduct fault diagnosis and signal recognition research on the data of the signal acquisition system. The results show that the average accuracy of this method reaches 90.0%, which can effectively achieve fault diagnosis and state evaluation of vacuum pumps.