Support vector machine method for measuring overflow granularity distribution of hydrocyclone for solid-liquid separation

Abstract

The support vector machine method for measuring the particle size distribution of solid-liquid separation in a hydrocyclone belongs to the technical field of overflow particle size measurement, and is characterized in that the feed flow rate, feed volume concentration, and feed pipe pressure directly related to the hydrocyclone are selected. , overflow concentration, and overflow particle size data obtained from off-line analysis are used as auxiliary variables, and then the functional relationship between the auxiliary variables and the measured variables is fitted by the computer using the support vector machine method to realize the measurement of the measured variables. Therefore, the present invention has the advantages of low requirement on the number of samples, sufficient precision, and easy realization and popularization.

Description

technical field [0001] The invention relates to measurement technology in automation technology. Specifically, it is a soft measurement method for an important index of a hydrocyclone, which is widely used in solid-liquid separation: the overflow particle size index. Background technique [0002] A hydrocyclone is a device for separating heterogeneous liquid mixtures, which uses centrifugal force to separate two or more phases based on the density difference between the two or more phases. Hydrocyclones are widely used in chemical, petroleum, ore dressing and other fields. According to the different substances to be separated, hydrocyclones can be divided into two types: solid-liquid separation and liquid-liquid separation. This article only discusses hydrocyclones used for solid-liquid separation. The following "cyclones" or "hydrocyclones" all refer to hydrocyclones used for solid-liquid separation. [0003] As a widely used solid-liquid separation equipment, the overfl...

AI Technical Summary

Problems solved by technology

The method proposed in patent A overcomes the shortcomings of the above two commonly used methods to a certain extent, but it also has certain shortcomings: 1. The connection between the ball mill and the classifier, the introduction of a large number of auxiliary variables, the introduction of small or repeated variables On the contrary, it will cause errors in the model; 2. Using the learning method based on empirical risk minimization requires a high number of samples for learning and updating, and it is difficult for companies that lack rapid overflow granularity index detection methods to meet the sample number requirements

Images