Abstract: For data automatic monitoring in the vacuum thermal test, the data abnormity detection method based on data mining is studied, and an improved DTW shape-based distance similarity measurement method is proposed. The algorithm reduces the amount of computation by the wavelet transformation and the search width limit, which avoids the distortion of the shape symbol by adjusting the calculation method. Its parameters are universal for the vacuum thermal test data and the best range of parameters is obtained through the statistical analysis of the actual sample data’s similarity clustering accuracy.