A Soft Sensor Modeling Method Based on Cooperative Sharing of Noise

Abstract

The invention provides a noise weakening method based on collaborative apportionment in soft sensor modeling, and realizes a soft sensor modeling example based on the method. The noise reduction method uses the correlation coefficient between auxiliary data and key variables to construct the credibility vector, uses the Euclidean distance between each data row and the center point data of the sample space to construct the bias degree vector of each data row, and uses each data and sample space The difference between the center point data constructs the real deviation matrix and deviation direction of each data, and finally combines the credibility, bias degree, deviation and intensity factor to jointly share the noise of the original auxiliary data, and uses different intensity factors to find the time when the noise weakens. The performance of the model is improved to the highest point, thereby balancing the noise data and non-noise data in the collected industrial data, solving the complex problem of noise in the soft sensor modeling data, reducing the difficulty of noise processing, and making the prediction performance of the soft sensor model It has been improved, and a soft sensor modeling example has been realized to show that the method has high adaptability and stability.

Description

technical field [0001] The patent of the present invention relates to a noise processing method and a soft measurement modeling method. It has important application and promotion value in the field of industrial production. Background technique [0002] The industrial process is complex and changeable, and the indicators are diverse, and many indicators are difficult to measure directly with testing instruments, or even impossible to measure. For such indicators, soft measurement technology is widely used at home and abroad for research and testing. Soft sensing technology is to construct a mathematical model with process variables that are easy to measure as input and key variables that cannot be measured or difficult to measure as output. In soft-sensor modeling, not only the high dimensionality, strong correlation, high redundancy, nonlinearity and complex noise of the data need to be considered, but also the model's own structure and related parameters. In terms of dat...

AI Technical Summary

Problems solved by technology

[0003] Soft sensor modeling is of great significance in industrial process control. Data noise cannot be completely removed, but traditional noise processing methods can only deal with specific noises. Therefore, a new noise processing method is needed to realize noise data and non-noise data. to improve the predictive performance of the model

Images