A method for diagnosing industrial process fault conditions based on deep neural network

Abstract

The present invention proposes a method for diagnosing industrial process fault conditions based on deep neural networks, including: obtaining offline fault video image sequences of industrial production processes; gray scale consistency transformation; extracting time series gradient images; spatial features and short-term time series features; extracting time series Long-term timing characteristics of gradient images; offline training to obtain the weight and bias of the neural network; obtain the online video image sequence of the industrial production process; calculate the probability distribution of the diagnostic results of the online video image sequence; output the diagnostic results; It has strong robustness, and has universal applicability to different industrial processes involving high temperature and image features, and can achieve better fault condition diagnosis results for industrial processes in different environments. Convolutional neural network and cyclic neural network The network is jointly trained, so that each set of training data only needs a set of training labels to complete the training, so that the network as a whole can meet the final fault diagnosis needs and improve the diagnosis accuracy.

Description

technical field [0001] The invention relates to the technical field of computer vision and the field of fault diagnosis, in particular to a method for diagnosing fault conditions of industrial processes based on a deep neural network. Background technique [0002] Many industrial processes involve the use of high temperature to convert raw materials into products, and their industrial processes (such as various abnormal working conditions, etc.) often show a variety of visual features. In particular, taking the refining process of fused magnesia as an example, the process usually uses a three-phase AC fused magnesia furnace (hereinafter referred to as fused magnesia furnace) to heat and smelt powdery raw materials mainly composed of magnesite ore through an electric arc. . The operation of the fused magnesium furnace mainly includes three normal working conditions of heating and melting, feeding and exhausting. Due to the characteristics of low grade, complex mineral compo...

AI Technical Summary

Problems solved by technology

However, many unstable factors in actual production make this method unable to work accurately and stably, such as the continuous production of CO in the molten pool 2 Bubbles cause the melt to roll, causing the distance between the liquid surface and the three-phase electrodes to change continuously, and the resistance will also change accordingly, resulting in irregular fluctuations in current

In addition, artificially formulating expert rules based on current data is also highly dependent on expert experience, which is prone to false negatives and false negatives

Images