Algorithm improvement and software system based on monitoring of data driving process

Abstract

The invention performs deep analysis on fault detection behaviors of process monitoring and studies change conditions of statistics when a fault happens and looks for an effective method to avoid missed detection of the fault and solve problems that monitoring statistics are not uniform and the like and looks for an effective method to improve the performance of fault diagnosis. Based on study work at present, we find that specific hidden variables which cannot be obtained through direct measurement may include characteristic information of a specific process fault and through measurement of a changing tendency of some statistics, the hidden variables which include the characteristic information are determined so that loss of useful information during statistical process monitoring can be prevented effectively and process monitoring and fault diagnosis effects can be improved. The process monitoring improvement algorithm improves the performance of the process monitoring significantly and included are mainly the following aspects: 1. fault detection time is reduced so that it is convenient for a process fault to be found timely and thus system operation risks are reduced; 2. the missed detection rate for faults, particularly some faults which cannot be detected through traditional methods is reduced; 3.loss of useful information can be prevented effectively so that conditions are provided for further fault diagnosis and fault clearing.

Description

technical field [0001] An invention related to chemical process monitoring and process quality management, in particular to a data-driven process monitoring algorithm improvement and software. Background technique [0002] The goal of statistical process monitoring is to identify process faults by constructing statistics, and to implement fault diagnosis by analyzing fault data, so as to ensure the normal operation of the process. Traditional statistical process monitoring is based on probability theory and mathematical statistics. In the early days of statistical process monitoring, due to the limitations of measurement technology, data storage and data analysis technology, people can only judge the state of the production process by measuring some important variables and indicators in the production process. However, in a complex industrial process, there are many variables or indicators that need to be monitored, and there are often correlations among them. The results ...

AI Technical Summary

Problems solved by technology

[0003] This invention deeply analyzes the fault detection behavior of process monitoring, studies the change of statistics when a fault occurs, and finds an effective method to avoid missed detection of faults. Solve the problem of different monitoring statistics, and find effective ways to improve the performance of fault diagnosis