Adaptive soft measurement prediction method based on Bayesian network with sliding window

Abstract

The invention discloses an adaptive soft measurement prediction method based on a Bayesian network with a sliding window. According to the method, the advantages of the sliding window are fully exerted, the soft measurement model is updated by continuously adding new samples and deleting old samples, and the data closest to the sample to be predicted in the aspect of time are constantly selected to perform modeling. The prediction model is established by using the Bayesian network method in each window, the prior probability distribution of each node in the network is acquired through the parameter learning mode, and input information of the sample to be predicted acts as the evidence to be added to the established network to obtain posterior probability distribution of the node to be predicted so that the output mean and variance can be obtained and qualitative variable prediction can be completed. Compared with other present methods, accurate quality forecast can be given to the constantly changing industrial process so that the prediction value can be obtained and the corresponding prediction accuracy can be given, and the soft measurement prediction problem in case of missing of the data set can be greatly solved.

Description

technical field [0001] The invention belongs to the field of industrial process control, in particular to an adaptive soft sensor prediction method based on a Bayesian network with a sliding window. Background technique [0002] The goal of soft sensing is to build appropriate models to predict quality variables that are difficult to measure or have large time delays in measurement using process variables that are easy to measure. Real-time and accurate prediction of quality variables is beneficial to control product quality and improve production efficiency. [0003] Soft sensor models are generally divided into mechanism models and data-driven models. With the development of computer technology, data-driven modeling methods have received more and more attention. There are many common data-driven modeling methods, and the most widely used ones are principal component analysis and partial least squares method, both of which are linear models; considering the uncertainty of...

AI Technical Summary

Problems solved by technology

[0004] However, in the actual industrial process, the model will be degraded due to process drift, catalyst failure, etc. Simply put, the original model is no longer applicable to the existing operating state, so it is necessary to continuously update the model to make it applicable time-varying industrial process

Images