In recent years, D-S evidence theory, as an uncertain reasoning method, is flexible in expressing uncertain information, convenient in reasoning mechanism, and close to the thinking habit of human experts in dealing with uncertain information. It is often used as the theoretical basis of information fusion and widely used in fault fusion diagnosis recently. The information fusion technology based on D-S evidence theory can analyze, process and synthesize different measurement information which provided by various types of sensors in the system, and get a more accurate and complete judgment than a single sensor. According to the characteristics of vibration isolation system, this paper optimizes the feature extraction method of vibration signal and completes the feature extraction. On this basis, the information fusion technology of D-S evidence theory is used for multi-sensor fault diagnosis, and the fault type is judged, and the diagnosis results before and after data fusion are analyzed.