Soft measurement method based on integrated multi-core partial least squares regression model

Abstract

The invention discloses a soft measurement method based on an integrated multi-core partial least squares regression model, which aims to establish and fuse the core partial least squares (KPLS) regression models corresponding to a plurality of core functions so as to avoid the problem of selection of the core functions and implement the online soft measurement on the basis. According to the method, firstly, four common kernel function types are all considered, so that the kernel function selection problem is avoided, and the method is stonger in universality. Secondly, a plurality of different KPLS regression models are respectively established by using a plurality of kernel functions, so that the advantages of a plurality of models are brought into full play. Finally, according to the method, the prediction data of each quality index is integrated into one soft measurement data by using a least square algorithm, so that the soft measurement effect is not weaker than that of any modelusing a single kernel function. By integrating the two advantages, the method provided by the invention overcomes the defects of the traditional soft measurement method based on KPLS, and is a more preferable soft measurement method.

Description

technical field [0001] The invention relates to an industrial soft sensor method, in particular to a soft sensor method based on an integrated multi-core partial least square regression model. Background technique [0002] Ensuring the stability of product quality is the basis of corporate profitability and market competitiveness, and real-time monitoring of product quality data is the direct technical means to achieve this goal. A natural idea is to measure quality data or quality-related data in real time. However, unlike measuring temperature, pressure, flow and other information, instruments and meters for measuring quality indicators are expensive and the cost of later maintenance is not a small expense. Although offline manual analysis can also be used to monitor product quality, the delay in quality monitoring cannot be avoided. Taking the product concentration produced in the process industry as an example, the concentration is the effective data that reflects the ...

AI Technical Summary

Problems solved by technology

However, choosing an appropriate kernel function from the perspective of improving the accuracy of the regression model requires a lot of verification

Images