Engraving machine anomaly detection method based on improved particle filter and convolutional network

The invention discloses an engraving machine anomaly detection method based on improved particle filtering and a convolutional network. The method is divided into an offline training stage and an anomaly detection stage. In the off-line training stage, firstly, operation data of the engraving machine are collected, a system model structure under a normal working condition is determined, then an improved particle filtering algorithm is introduced to obtain a system state variable estimation value, an expectation maximization algorithm is used for iteratively updating model parameters, and an accurate mathematical model of the engraving machine system is obtained; and finally, training a one-dimensional residual convolutional neural network by using residual information of the model to realize binary classification of system operation data. In the anomaly detection stage, the mathematical model is used for conducting particle filtering on the operation data to obtain an estimated value sequence of noise, and fi