To address the problems of over-reliance on manual experience and poor generality of processing complex signals in the traditional rotating machinery health state assessment methods,an error threshold anomaly detection method was proposed based on the adversarial autoencoders model(AAE).Feature extraction and operating state modeling were carried out by using the vibration signal of equipment directly,and distribution model was established by using vibration state data of the equipment in normal state;deep learning was used to learn the inherent characteristics of vibration data,and error threshold was introduced as the decision criterion of fault warning to achieve the efficient evaluation of the equipment operating state;a high speed centrifugal pump was used to verify the proposed method.The results show that the accuracy of the counteracting self-coding model can reach 100% for the judgment of abnormal data,by using this method,the operation state of rotating equipment can be effectively detected based on the monitoring data;compared with the traditional auto-encoder,the method has a significant improvement in the accuracy and precision of diagnosis.