Bayesian semi-supervised robust PPLS soft measurement method based on incomplete data

Abstract

The invention discloses a Bayesian semi-supervised robust PPLS (Bayesian semisupervised robust PPLS, BSRPPLS) soft measurement method based on incomplete data, which is Bayesian semi-supervised robustPLS fault monitoring method based on incomplete data and is different from the existing Bayesian semi-supervised robust probability PLS fault monitoring method based on multivariate student distribution PPLS modeling and is used for modeling each data vector noise by using independent student distribution. An adjustable robust degree-of-freedom parameter is contained in the distribution, so thatthe modeling flexibility is improved, and an estimated posteriori distribution parameter is solved by using a Bayesian variational reasoning method, the model can reconstruct original data by using pollution-free data elements, reduces the influence of polluted elements on the reconstructed data, solves the problems of data loss and influence on the model precision at wild points, has good robustness, and is beneficial to improving the industrial process monitoring performance and the process operation understanding cognition level.

Description

technical field [0001] The invention belongs to the technical field of PPLS soft measurement, in particular to a Bayesian semi-supervised robust PPLS soft measurement method based on incomplete data. Background technique [0002] With the advent of the Industry 4.0 era, the modern industrial automation system is constantly developing towards the trend of complexity, informatization and intelligence. As the key to ensure the stability of product quality and the safe and stable operation of process production equipment, process monitoring has become an indispensable and important part of modern complex industrial systems. In the actual process, due to changes in the external environment, fluctuations in the quality of raw materials, the accuracy of the measuring equipment itself, and the complexity of the equipment, it is difficult to directly establish a mathematical monitoring model for the process. Therefore, the theory and technology of process monitoring based on data dy...

AI Technical Summary

Problems solved by technology

This model not only makes full use of labeled data and unlabeled data, but also uses uncontaminated data elements to reconstruct the original data as much as possible, reduces the impact of contaminated elements on reconstructed data, and solves the problem of data loss and wild points at the same time. It has good robustness. It improves the accuracy of the model, which is conducive to improving the performance of industrial process monitoring and the level of understanding of process operation

Images