Research On Intelligent Dosing Control Of Coal Slime Flotation Based On Machine Learning

The flotation process has a large number of uncertain factors,with multiple and complex variables and a certain coupling relationship between them,which cannot be described by a simple linear model.At the same time,the production process is not a stable process,but is constantly changing and fluctuating,making it difficult to control.Therefore,the entire process needs to be analyzed as a whole.This article uses machine learning algorithms to establish a model based on a large amount of experimental data to predict it.The thesis takes the flotation process of a coal preparation plant as the research object,analyzes the process flow and influencing factors of the coal slurry flotation process,and determines the variables and detection methods that need to be detected in the process control based on actual production: the input flow rate,input concentration,and inflation rate are obtained using electromagnetic flow meters,mass flow meters,and gas flow meters to obtain data;The tailings ash content is detected using an image recognition based tailings ash analyzer;Manual testing and analysis of feed and clean coal ash content;For the flotation dosing control actuator,a high-precision diaphragm pump is used for dosing.To improve the quality of sample data,in-depth analysis and preprocessing of the original data were first conducted.Since the data does not change much in a short time,the iterative average calculation is carried out for each variable,the box chart method is used to detect the outlier of the sample data,and the data processing methods such as data elimination and data noise reduction are used to process the abnormal data,and then the processed data are subject to principal component analysis and normalization,providing data guarantee for the subsequent model training.Due to the large sample size and high data dimensionality,this article selects multiple hidden layer BP neural networks and LSTM,GRU deep neural networks to perform regression prediction on the dosage of collectors and foaming agents,and establishes a "10 input 2 output" flotation process reagent dosage prediction model.Through a detailed comparison of the prediction results of BP model,LSTM model,and GRU model,the LSTM model has the highest prediction accuracy,with predictions of collectors and foaming agents reaching 95.46% and 85.60%,respectively.Therefore,an intelligent dosing prediction model for flotation process was established using LSTM neural network algorithm.On the basis of the intelligent dosing prediction model in the flotation process,an intelligent control system for the flotation process was constructed.A fuzzy controller was designed with the tailings ash deviation and deviation change rate as inputs and the reagent addition amount as output.The upper computer monitoring interface of the control system was designed,and data communication and dosing amount control were achieved through OPC technology and PLC communication with the lower computer.The on-site application results show that after the intelligent control system is put into use,it can automatically adjust the dosage of collectors and foaming agents according to changes in flotation production parameters,achieving the production requirements of flotation clean coal ash content<11.5% and flotation tailings ash content>65%,achieving intelligent control of flotation process dosing.The thesis consists of 63 figures,28 tables,and 82 references.