Production process quality prediction and control method based on KEPLS

Abstract

The invention discloses a production process quality prediction and control method based on KEPLS, and the method specifically comprises the following steps: building a KEPLS model according to field data, carrying out feature sampling through IFS, and constructing a new algorithm based on IFS-KEPLS to carry out the online monitoring and quality prediction of a production process; and after the IFS-KEPLS performs quality prediction and a predicted value seriously deviates from a standard prediction curve, constructing a second model, namely, performing production process quality control by adopting an SV-KCD and KEPLS combined algorithm. According to the method, the monitoring and prediction performance of the model is improved, so that tiny and slowly changing faults are captured, and the quality control standard in the intermittent production process of the flow industry is well improved.

Description

technical field [0001] The invention relates to the technical field of production process control, in particular to a quality prediction and control method in the industrial production process. Background technique [0002] The modern process industry involves more and more production links, and the complexity is getting higher and higher, and the intermittent production process accounts for a large proportion, for example, the production of pharmaceutical intermediates, steel smelting, glass firing, etc., involving chemical, steel , metallurgy, medicine and other industries, mainly aiming at the process improvement in the intermittent production process, through the use of data mining, data analysis, quality prediction and other means to improve product quality and pass rate. [0003] For the quality prediction of intermittent production processes, it has high added value and importance in production. At present, most of the production equipment is equipped with sensors, w...

AI Technical Summary

Problems solved by technology

However, this type of data-driven algorithm has certain requirements for the real-time performance of online applications in the production process of the process industry, especially for fast-changing processes or when key data in sensitive periods have a strong correlation with product quality. The effect has little effect. If you participate in online prediction, it may even directly affect the quality and pass rate of the product. The applicability of this type of algorithm in such a scenario is yet to be discussed. Secondly, the data in the process industry generally has high dimensions and nonlinearity. , strong correlation and other characteristics also limit the specific application of this type of algorithm; thirdly, this type of data mining method often has a certain lag when processing the quality data of the process industry, and cannot reflect the production status in a timely and effective manner, while the quality Once the data lag is applied to production, it is very likely to lead to a decline in product quality, unqualified products, and even production failures that lead to accidents; finally, in actual production, if slow and imperceptible failures occur on site, Existing algorithms can only extract second-order statistical information, and its sensitivity is difficult to meet the requirements, which will cause the sample variables in the high-order statistics to fail to display the corresponding high-order information due to the small fault information being submerged in the noise

Images