Intermittent process monitoring data processing method

Abstract

The invention relates to the field of control science and engineering, especially an intermittent process monitoring data processing method. The method comprises the following steps: a, employing an improved affine propagation clustering algorithm; b, designing a similarity decreasing scanning algorithm; c, determining outliers and wrong classification points, and respectively designing solutions.The improved affine propagation clustering algorithm can reveals a mode switching process between all subperiods of an intermittent process more accurately, and the similarity decreasing scanning algorithm can further reflect the change trend of the internal mode of each subperiod with time. The method for determining the outliers and wrong classification points and respectively designing solutions improves the statistical modeling accuracy, is small in algorithm scale, is simple in calculation, and is easy for engineering implementation.

Description

technical field [0001] The invention relates to the field of control science and engineering, in particular to a batch process monitoring data processing method. Background technique [0002] For a long time, with the market's urgent demand for multi-category, small-batch and high-value-added products, batch production has become the main production method in many industrial fields. However, the complexity of intermittent production will inevitably lead to production reliability and safety issues. In order to capture the mechanism of batch process more accurately to improve monitoring performance and monitor potential safety issues in a timely manner, the academic field has conducted in-depth research on batch processes and proposed many data-driven multivariate statistical analysis methods. For example, the typical principal component analysis method and least squares method and their extended forms have been widely used. However, these traditional methods basically assum...

AI Technical Summary

Problems solved by technology

However, these two methods have different degrees of limitations in the stage division

K-means clustering is relatively random when selecting the initial cluster center, and is sensitive to noise and outliers, which will lead to unstable clustering results and limited clustering quality

Fuzzy C-means clustering is a local search algorithm, which needs to manually determine the number of cluster centers, and its iterative search tends to fall into a local optimal solution, so the clustering results are often unsatisfactory.

In fact, for a complex batch process, it is difficult to accurately estimate the number of cluster centers without prior knowledge or experience, so it is not conducive to the accurate modeling and monitoring of batch processes

Images