Study On The Extraction Of Dynamic Information Of Flotation Machine Froth And Control Model Of Froth Layer Thickness

Froth flotation is an important beneficiation method,as over 90% of non-ferrous minerals are sorted by flotation methods.Froth flotation uses the hydrophobic differences in mineral particles under the action of flotation chemicals to achieve separation.The flotation process is accomplished with the aid of flotation equipment including flotation machines and flotation columns.The control of the flotation machine is critical to the mineral flotation index and sorting efficiency.In traditional flotation production,the adjustment of these parameters relies on workers to determine the working state of the flotation machine by observing information on the apparent characteristics of the flotation froth and adjusting control parameters such as the thickness of the froth layer(Ore slurry level).Although the automatic control technology of flotation machines has developed rapidly in recent years,including the automatic control technology of froth layer thickness(Ore slurry level)has been applied to a certain extent.However,the determination and setting of these control parameters still needs to be done manually with the help of experience.Due to the different experience and skill levels of each person,it is not possible to ensure that the parameters are set optimally,which can easily cause fluctuations in flotation indexes and ensure that the flotation machine operates in an optimal state.To address these problems,this paper proposes the use of visual and image recognition technologies instead of manual to obtain information on the dynamic characteristics of the flotation froth(froth movement speed,froth stability).Firstly,an improved GMS-based flotation foam image feature matching algorithm is proposed,which is 6-7 times more efficient in operation compared to the traditional matching algorithm.In order to eliminate the clustering of feature points in the matching results,a foam image feature matching algorithm based on the deep learning Super Glue model is proposed,which effectively improves the foam image feature point matching effect.The results show that the distribution of the moving velocity field of the froth feature points based on the Super Glue model is better.In this paper,a foam stability detection method based on the matching rate of image feature points is proposed for foam stability,and the stability value of the foam is successfully extracted.In comparison with the existing foam stability feature extraction results based on the grey-scale difference method,the results demonstrate that the proposed foam stability detection method has a higher sensitivity to detect changes in foam feature structure.On the basis of the above,this paper uses Pearson coefficients to analyse the correlation between the data characteristics of raw ore grade,concentrate grade,influent flow rate,filling volume,froth movement speed and froth stability in the production of industrial flotation machines,and finds that the data characteristics that are more correlated with the froth layer thickness of flotation machines are influent flow rate,froth movement speed,froth stability and concentrate grade.Based on the above,a Particle Swarm Optimization based Gradient Boosting Decision Tree(PSO＿GBDT)froth thickness control model for flotation machines was constructed and compared with other algorithms such as Particle Swarm Optimization based Support Vector Regression(PSO＿SVR)to show that the PSO＿GBDT based control model has the lowest mean absolute error The results show that the PSO＿GBDT-based control model has the lowest mean absolute error and is more stable.This paper combines the above research results with an existing automatic control system for flotation machine froth layer thickness(Ore slurry level)and proposes an implementation of an adaptive control system for flotation machine froth layer thickness based on the dynamic characteristics of the froth.Finally,the improved GMS algorithm,the Super Glue model deep learning algorithm and the PSO＿GBDT based flotation machine froth layer thickness model developed in this paper are tested on a large 160m3 flotation machine in a large copper ore processing plant in Jiangxi.The results validate the feasibility of the proposed flotation froth dynamic feature extraction algorithm and the constructed froth layer thickness control model based on the flotation froth dynamic features.