Self-adaptive quality forecasting method based on incremental support vector regression

Abstract

The invention discloses a self-adaptive quality forecasting method based on incremental support vector regression, which is used for adapting to a complex and changeable multi-working-condition production process, realizes the differential learning of a newly added sample by utilizing the KKT condition of the self-adaptive quality forecasting method based on incremental support vector regression on the basis of the original support vector regression model, carries out the incremental learning on samples with information, keeps the model unchanged for samples without new information, and can reduce the updating frequency of the model while ensuring the generalization capability of the model. The self-adaptive quality forecasting method based on incremental support vector regression not onlycan effectively face the nonlinearity in the actual industrial process, but also can continuously update the model aiming at the time-varying characteristic in the process, and the problem of the updating efficiency of the model is improved to a certain extent, so that the purpose of self-adaptive quality prediction is achieved.

Description

Technical field [0001] The invention belongs to the field of industrial process control and soft measurement, and relates to an adaptive quality prediction method based on incremental support vector regression. Background technique [0002] There are many variables that cannot or are difficult to directly measure in industrial processes. These variables are usually closely related to product quality and are parameters that must be monitored. Soft measurement technology is an effective method to solve this problem. It has been obtained in industrial processes. widely used. In recent years, data-driven soft-sensing modeling methods have received widespread attention. Among them, support vector regression is based on the principle of structural risk minimization, which has solved the problems of small samples, nonlinearity and local minima, and has been widely used in many research fields. The current support vector regression learning algorithms are mostly offline and batch proce...

AI Technical Summary

Problems solved by technology

In this way, the model obtained by offline training cannot be guaranteed to be suitable for a long time, so the research on the online learning method of support vector regression has important theoretical significance and practical value

Images