A weighted Gaussian model soft measurement modeling method with time delay estimation

Abstract

The invention discloses a weighted Gaussian model soft sensor modeling method with time delay estimation, belonging to the field of complex industrial process modeling and soft measurement. The methodcomprises the steps of estimating process delay parameters by using a sliding grey relational degree algorithm and extracting process delay information; when a new sample arrives, reconstructing themodel sample based on the time delay parameters estimated in the offline phase; establishing a weighted Gaussian model to construct the joint probability density function of input and output variablesby the weight relative to the training sample; at last, employing that condition distribution function to estimate the value of the output variable in real time to predict the key variable accurately. In an intuitive and effective way and with the low computational complexity, the time delay information of variables extracted from the process history database is used for the soft measurement modeling data reconstruction, the actual causal correspondence between input and output is corrected, the interference of process random noise is effectively solved, and more accurate prediction results are obtained. Therefore, the product quality is improved and the production cost is reduced.

Description

technical field [0001] The invention relates to a weighted Gaussian model soft sensor modeling method with time delay estimation, which belongs to the fields of complex industrial process modeling and soft sensor. Background technique [0002] In the actual industrial process, the measurement of some key variables is very important to produce high-quality products, but under the constraints of existing technical conditions and economic costs, it is very difficult to directly obtain key variables; based on this background Under the circumstances, soft-sensing technology came into being; it infers and estimates difficult-to-measure key variables by constructing the functional relationship between easy-to-measure variables and key variables in the process, so it has been widely used. [0003] However, the actual industrial process data is usually polluted by random noise due to factors such as measurement variation and transmission interference, thus exhibiting certain uncertai...

AI Technical Summary

Problems solved by technology

Fortuna L et al. used process device process design parameters to estimate the approximate range of time delay. Since the time delay was only roughly estimated, the prediction accuracy of the model was poor; Komulainen T and Zhang J estimated the time delay by calculating the correlation coefficient between input and output variables. Only the linear relationship between variables is considered, and ideal results may not be obtained for nonlinear processes; Li Yanjun et al. use fuzzy curve analysis to estimate time delay parameters, but there are certain requirements for the number of samples; Ruan Hongmei et al. use DE algorithm to optimize The joint mutual information between the process variables of the alkane tower is used to determine the process delay parameters, but the intelligent optimization algorithm is easy to fall into local optimum, and the calculation complexity is high

Images