In view of the problem of mass data compression generated during the diagnosis of metro vehicle plug door by PHM, a wavelet compression algorithm based on adaptive sectional threshold was proposed. Through the adaptive segmentation of the original data and the automatic adjustment of the amplification factor of the threshold of each segment, data compression with high accuracy and large compression ratio was achieved. To tackle the problems that the energy of the sampling current of the plug door motor was not concentrated and the amplitude varied greatly, the collected original data was adaptively segmented. Amplification factors of each segment of the wavelet threshold were adjusted automatically by a preset threshold, and then one-dimensional wavelet was used for data compression. On the premise of satisfying the accuracy of reconstructed signals, the method achieved high data compression ratio and greatly reduced the data amount to be stored. Data from Chaqiao Depot of Metro Line 2 in Wuxi, Jiangsu Province was obtained, and data compression test for metro vehicle plug door was carried out based on segmented adaptive wavelet threshold by selecting db3 wavelet for 4-layer wavelet compression. Results show that the test realized ultra-high compression ratio of 2.56% with the overall distortion ratio of 0.3%, indicating that the method is feasible and effective and can meet the requirements of actual engineering application.