Construction Research On Reagent Addition Model Of Mineral Flotation Based On Generative Adversarial Network

Froth flotation is a widely used method of beneficiation,which separates useful minerals from pulp by filling air and adding suitable flotation reagents.In the froth flotation process,the adjustment of reagents addition amount will directly affect the stability of the production process and the recovery rate of mineral resources.At present,a large number of flotation reagent addition modeling methods based on machine vision mainly focus on establishing the relationship between a single feature of the froth image and the reagents addition amount.These methods are difficult to accurately model the flotation reagent addition process.To this end,this paper analyzes the correlation between the flotation reagents addition amount and the features of the froth image,and proposes an intelligent modeling method for the flotation reagent addition process based on generative adversarial network(GAN).The research work of this article mainly includes:Aiming at the problem that the complicated variable status of froth images makes it difficult to effectively extract visual features,this paper uses convolutional neural networks(CNNs)to replace traditional manual feature extraction methods to obtain deep semantic features of froth images.At the same time,in order to solve the problem that the coupling relationship between the reagent addition amount and the froth image features in the mineral flotation process is complex and difficult to accurately model,a correlation modeling method that maximizes the mutual information between the reagent addition amount and the froth image characteristics is proposed.The mutual information between the reagent addition amount and the depth characteristics of the froth image is approximately solved by the lower bound of the mutual information variation,and the neural network parameters are optimized with the goal of maximizing the mutual information,and the correlation between the flotation reagent addition amount and the froth image features is obtained.Considering that the dimensionality of the froth image is too high,it is difficult for traditional modeling methods to directly construct the reagent addition model based on the froth image.This paper proposes a modeling method for the flotation reagent addition process based on generative adversarial network.This method directly uses flotation reagents addition amount and froth image as model input to construct a flotation reagent addition model.In order to realize the prediction of the froth image after adding reagents and the measurement of the feature distribution difference,the image prediction network and the image discrimination network of the model are rationally designed.The flotation reagent addition model uses the game confrontation between the image prediction network and the image discrimination network to realize the accurate modeling of the flotation reagent addition process.Using the working condition data collected from the gold antimony flotation industry site,the model proposed in this paper is used to simulate and verify the single-step reagent addition process and the multi-step reagent addition process.The single-step reagent addition results show that predicted froth image can effectively model the influence of the actual reagent dosage on the froth image.The multi-step reagent addition results further show that the model has a good froth image prediction ability,which can predict the froth image after adding reagents according to the multi-step adjustment of the reagent addition amount.Through the ablation experiment of mutual information loss,it further shows that the flotation reagent addition model fused with mutual information can realize the potential association modeling between the flotation reagents addition amount and the froth image features.