Fault classification method based on bidirectional long-short-term memory unit and capsule network

Abstract

The invention discloses an industrial process fault classification method based on a bidirectional long-short-term memory unit and a capsule network. The industrial process fault classification methodis composed of the bidirectional long-short-term memory unit, the capsule network, a sensor and a Softmax output layer. The bidirectional long-short-term memory unit can process the sequence data andextract dynamic features from the data; each capsule in the capsule network has size and direction information; the capsule network has stronger expression capability, dynamic characteristics extracted by the bidirectional long-short-term memory unit can be further integrated, the sensor and the Softmax output layer perform fault classification according to fault sample characteristics integratedby the capsule network, and the network is suitable for processing the classification problem of industrial data with nonlinear characteristics and dynamic characteristics. According to the method, the classification precision of samples with dynamic characteristics and other characteristics can be effectively improved.

Description

technical field [0001] The invention belongs to the field of industrial process fault diagnosis and classification, and relates to a fault classification method based on a bidirectional long-short-term memory unit and a capsule network. Background technique [0002] In industrial process monitoring, when a fault is detected, it is necessary to further analyze the fault information, and fault classification is an important part of it. Obtaining the type of fault is conducive to the restoration of the industrial process. [0003] With the development of industrial automation and the popularization of sensors, the amount of industrial data is larger. In addition, due to the complexity of industrial processes, industrial data generally has characteristics such as nonlinearity and dynamics (also known as autocorrelation), and deep learning can quickly It is best to use the large amount of data to extract the nonlinear and dynamic features of industrial data. Therefore, deep learn...

AI Technical Summary

Problems solved by technology

[0005] Aiming at problems such as nonlinearity and strong dynamics in data in the current industrial process, the present invention proposes a fault classification method based on bidirectional long-short-term memory unit and capsule network. (CapsNet), multi-layer perceptron (MLP), Dropout layer and Softmax layer constitute the entire BiLSTM-CapsNet model, which realizes accurate classification of faults in industrial processes

Images