Fault diagnosis methodbased on wavelet convolutional neural network

Abstract

The invention discloses a fault diagnosis method based on a wavelet convolutional neural network, and the method comprises the steps: training a preliminary diagnosis model through employing a CNN algorithm, and enabling faults to be divided into three types: normal faults, easy diagnosis (ETD) faults and difficult diagnosis (HTD) faults; feature extraction is achieved by stacking a convolution layer and a pooling layer, and finally fault diagnosis is achieved through two full connection layers and softmax. Compared with the prior art, the method has the advantages that 1) a multi-model diagnosis framework is developed, the calculation amount of models is reduced, and valuable prior knowledge application in black box process monitoring is facilitated; 2) the rectangular convolution kerneland the pooling function are applied to the chemical data so that the feature extraction capability of the WCNN is improved, and the method can be popularized to other industrial data; and 3) the fault detection and diagnosis performance of the chemical production process is improved, and meanwhile, the calculation burden and the monitoring performance are balanced.

Description

technical field [0001] The invention relates to the technical field of chemical process fault diagnosis, in particular to a method for detecting abnormal images. Background technique [0002] Corrosion, aging, fouling and other changes of important parts or equipment in the chemical production process make chemical industry a powerful time-varying process. Such complex nonlinear and time-varying mechanisms require sophisticated monitoring methods. [0003] The convolutional neural network is applied in the chemical process, making full use of the distributed control system (DCS) data at each point of the entire process, so as to realize fault diagnosis. However, the current convolutional neural network (WCNN) is not closely integrated with industrial data. For some chemical processes with more complex processes, such as chemical processes, it will take too long and the diagnosis rate will drop. Contents of the invention [0004] In order to improve the performance of fau...

AI Technical Summary

Problems solved by technology

However, the current convolutional neural network (WCNN) is not closely integrated with industrial data. For some chemical processes with more complex processes, such as chemical processes, it will take too long and the diagnosis rate will drop.

Images