A Fault Detection Method for Industrial Processes Based on Parallel Partial Least Squares

Abstract

The invention discloses an industrial process fault detection method based on a parallel partial least square method. The industrial process fault detection method comprises the steps that the process data and the quality data under the normal condition are acquired to act as sample data, a parallel partial least square method data model is established and the input and the output of the parallel partial least square method data model are divided into four subspaces by using the parallel partial least square method: an input-output joint space, an unpredictable output principal component space, an unpredictable residual space and an input residual space; the monitoring statistic quantity index of the four subspaces and the control limit thereof are calculated; and the industrial process data and the quality data to be detected are acquired, the industrial process data and the quality data to be detected are divided into four subspaces and the actual monitoring statistic quantity index of each subspace is calculated, and the judgment result indicates that the fault occurs in the industrial process to be detected when the actual monitoring statistic quantity index exceeds the control limit. More reasonable subspace division is performed on the input and output data so that the accuracy of fault monitoring can be enhanced.

Description

technical field [0001] The invention belongs to the technical field of industrial process fault detection, and more specifically relates to an industrial process fault detection method based on a parallel partial least square method. Background technique [0002] With the continuous growth of industries such as chemical industry, petroleum, and steelmaking, people urgently need feasible and effective fault monitoring strategies for actual industrial production. The partial least squares method is favored by academic research and business circles, and it can detect and diagnose faults in industrial production. Whether a fault occurs can be judged by whether the fault statistics exceed the limit, and whether the fault will affect the output quality of the product. In practice, it can diagnose the specific situation of the fault, and more importantly, tell the producer whether the fault affects product quality. For those faults that will affect product quality, producers will...

AI Technical Summary

Problems solved by technology

Second, the latent variables extracted by the partial least squares method are not in descending order of the variance size, which leads to a large variance change in the residual space, which is not suitable for monitoring with Q statistics

Third, there are disturbances in the main space, which will affect the monitoring results of the main space

Fourth, the partial least squares method (Partial Least Squares, PLS) mainly considers the monitoring of input X, and does not consider the monitoring of Y

Images