Multi-sensor systems usually have hundreds or thousands of sensors, each of which generates a large amount of parameter data over time. Once some of the parameters state distribution changes, it is likely to cause abnormalities in other parameters in the system, even The entire system is not working properly. This project is based on Bi-LSTM's AutoEncoder deep structure real-time monitoring system for all sensor parameter status. Once an abnormal parameter feature occurs, the system will promptly notify the relevant personnel to timely repair. This method can not only effectively maintain the efficient operation of the system, but also avoid losses caused by system anomalies.