Research And Application Of Intelligent Control Of Flotation Dosing System

Slime flotation is an important process in coal preparation plant and the intelligent construction of flotation system is of great significance to the overall intelligent construction of coal preparation plant.At present,because there is no accurate measurement method of production index in slime flotation production,the production status can only be judged by the work experience of on-site operators,and the dosage of reagents is usually set based on the subjective judgment of on-site operation experience.This operation mode makes the field work personnel’s labor intensity is high,and site operators often misjudged subjective factors,but also because making reagent adding quantity is too high or too low,lead to plant and tail coal grade instability or depletion of medicament,which affect the production efficiency,and seriously hindered the intelligent construction of coal preparation plant.In view of this problem,this paper proposes an ash soft measurement technology based on machine learning,which can achieve rapid and high-precision online detection of cleaned coal ash and tailings ash.On this basis,an intelligent control method of flotation dosing system combining expert system and Bayesian network is proposed to realize the intelligent addition of flotation agents.This paper firstly proposes a soft measurement technology of ash content based on machine learning in order to solve the problem of lack of high precision production index detection means in the current coal slime flotation production.In view of the fact that it is difficult to use flotation foam image to measure cleaned coal ash in industrial field and susceptible to environmental impact,a more stable image acquisition device for flotation tailings was built.Stable and high-quality images of tailings can be obtained by this device.By extracting gray and color characteristics,and then combining with flotation process variables,the soft measurement technology of cleaned coal ash based on machine learning methods is researched and realized.Four different machine learning methods,namely random forest RF,support vector machine SVM,Gaussian process regression GPR and LSTM,are used for model training and comparison through field data.The results show that among the four machine learning methods,LSTM model has the best prediction effect,and the root mean square error of prediction results RMSE is 0.2231.The R~2 of goodness of fit is 0.963,and the soft sensor model of cleaned coal ash is established based on the algorithm.The soft sensor model can be used to detect cleaned coal ash by slime flotation.The root mean square error(RMSE)and average percentage error(MAPE)are 0.4783 and 3.9345%respectively.The model was also used to optimize the traditional tailings ash prediction technique,which relies only on tailings images,by adding flotation process variables,the results show that the introduction of flotation tailings ash content after the process variables predicted results is superior to the traditional rely on prediction of tailings image characteristics of ash content,RMSE reduced from 5.1114 to3.898,The mean percentage error also decreased from 4.1942 to 3.1058.The stability control method of flotation dosing process based on expert system and Bayesian network is proposed.In this method,feedforward reagent is added by expert system according to flotation feeding condition.Through the combination of on-line detected ash value and flotation condition,the compensation model of reagent dosage was designed by Using Bayesian network,and the feedforward dosage was revised online.Then the intelligent control simulation of flotation dosing system is studied.A PID controller based on BP neural network was designed,and system identification toolbox was used to identify flotation dosing process.The simulation research in Simulink proves that the PID controller based on BP neural network is more suitable for the process control of flotation process,which is a complex nonlinear process system.Finally,on the basis of the above research,the software and hardware architecture of the system is built for Liuwan Coal preparation Plant,and the industrial test is carried out in the industrial site.The results verify the feasibility of the research,reduce the labor intensity of operators,improve the economic benefit and intelligent level of coal preparation plant.