Platinum Flotation Grade Estimation Method Based on Image Data Extraction and Neural Network Modeling

Abstract

The invention relates to a platinum flotation grade estimation method based on image data extraction and neural network modeling. Through variable experiments, six variables and flotation rate, pulp concentration, collector, activator, foaming agent, and depressant are obtained. Select the degree of correlation between the grade and the recovery rate; collect and preprocess the platinum foam image, and obtain the default grayscale image, histogram equalization, image contrast enhancement and image binarization from the preprocessed four Five types of image data including energy, entropy, inertia, homogeneity, and gray-level correlation are extracted from the feature image; a multi-layer perceptron neural network model including a three-node input layer, a hidden layer, and a two-node output layer is established. A platinum flotation grade estimation method based on image data extraction and neural network modeling proposed by the present invention effectively realizes the estimation of flotation grade and recovery rate through foam images, and achieves real-time monitoring of grade and recovery in the flotation process rate purposes.

Description

technical field [0001] The invention relates to the field of mineral separation, in particular to a platinum flotation grade estimation method based on image data extraction and neural network modeling. Background technique [0002] With the development of science and technology, the requirements for the automatic monitoring technology of various parameters in the beneficiation process are getting higher and higher, and many disadvantages of the traditional method are gradually emerging. Traditional monitoring technologies mainly include: fluorescence analysis and particle size analyzer, which can obtain product grade in a short time. The disadvantages of the traditional method mainly include: the design and development of related equipment is not practical, the error of detection data is large, the service life of the sensor is short due to the poor beneficiation environment, and the maintenance cost of the control system is high. [0003] Artificial neural network (ANN), ...

AI Technical Summary

Problems solved by technology

The disadvantages of the traditional method mainly include: the design and development of related equipment is not practical, the error of detection data is large, the service life of the sensor is short due to the poor beneficiation environment, and the maintenance cost of the control system is high

Images