Data Prediction Analysis And System Implementation For Flotation Engineering

Froth flotation is considered to be one of the most widely used and important beneficiation methods in mineral processing technology.Flotation and its automatic control technology will become more and more important in the future when rich mineral resources are gradually attenuated.Precisely predicting the purity of the flotation product is of great significance for optimizing the control of various parameters of the flotation process and realizing the automatic control of flotation.Currently,flotation products’ purity prediction mostly uses mainstream statistical learning models to model and analyze flotation historical data,but statistical learning models cannot meet the requirements for prediction accuracy in actual production.Aiming at the problem of flotation products’purity prediction,this thesis studies the algorithm of flotation products’purity prediction which is based on statistical learning and deep learning.This thesis also designs and implements an analysis software which is suitable for flotation data prediction.The main tasks completed in this thesis are as follows:1.The thesis realized the flotation algorithm of products’ purity prediction based on statistical learning model.According to the characteristics of the dataset,this thesis uses ridge regression,decision tree and AdaBoost to model this flotation process.Since there are multiple values to be predicted,the thesis compares each model’s native model and its multi-target regression model,and analyzes its prediction effect and operated efficiency.Experimental results show that the AdaBoost model has the highest prediction accuracy index among the three models,but its modeling and prediction time is also the longest.2.The thesis realized and improved the LSTM-based flotation product purity prediction algorithm.The deep learning model FlotationNet is used to modeling this flotation process based on time series,and its prediction accuracy and operated efficiency are analyzed.It was found that its prediction accuracy was poorer than the AdaBoost model,and the modeling speed was also slower.In order to optimize the prediction accuracy and improve the modeling speed,the model structure was optimized and adjusted for the shortcomings of the model.Experiments show that the improved model has higher prediction accuracy than the AdaBoost model,and the modeling time is reduced by nearly half compared to the original model.3.The thesis designed and implemented a set of systems suitable for flotation data prediction and analysis.Based on the five prediction algorithms studied in the thesis,a B/S system for predicting the purity of flotation products was developed,which realized data preprocessing and prediction simulation,thereby pro viding a basis for parameter optimization.Mainly include purity prediction simulation,parameter statistical analysis,and historical data visualization and analysis functions.