Research On Flotation Index Prediction And Process Parameter Optimization Based On Data Driv

Flotation process is widely used in mineral resources development,because of the increase of labor cost and the demand for cost reduction,efficiency and quality improvement,it has become a key object of transformation in the construction of intelligent mine.At present,the data accumulated in the long-term production process of the flotation section can be used to realize the prediction and multi-step prediction of important flotation indexes such as concentrate grade and concentrate recovery,as well as to find the combination of process parameters that can obtain better flotation indexes.By providing the above information to the flotation section site,it can help the staff to better grasp the production situation and adjust the process parameters,so as to reduce the influence of human factors in the flotation production process and stabilize the flotation benefit at a higher level.This paper takes the copper flotation section of Xintian concentrator of Yunnan Hualian Zinc Indium Co.,Ltd as the research object,establishes the flotation index prediction and multi-step prediction model,and optimizes the process parameters.The specific research contents and results are as follows:(1)According to the process state of copper flotation section and the subsequent modeling requirements,data acquisition and data preprocessing are carried out,and data set B for the establishment of prediction model and data set A for the establishment of multi-step prediction model are obtained.Based on the data set B,which has completed the processing of parameter lag problem,the influence of each process parameter on grade and recovery is evaluated comprehensively.(2)After using different feature engineering methods on data set B,the LightGBM prediction model and DNN prediction model are established,and the prediction effects of the two models on grade and recovery are compared and analyzed.The results show that the prediction model established by LightGBM method is relatively superior.The RMSE and MAE of the grade prediction model were 0.998 and 0.715,and the RMSE and MAE of the recovery prediction model were 1.424 and 1.027.The LightGBM prediction model can be used to optimize the process parameters.(3)After using different feature engineering and data conversion methods for data set A,the LightGBM recursive multistep prediction model and LSTM multi-output multistep prediction model are established,and the prediction effects of the two models on grade and recovery in the next 6 time steps(each time step is 15min)are compared and analyzed.The results show that LightGBM model is more accurate in predicting the next three time steps,while LSTM can achieve stable prediction for further time steps.Therefore,the modeling method should be selected according to the length of time steps to be predicted.The LightGBM one-step-ahead prediction model established during this period can also be used to optimize the process parameters.(4)After the prediction model,optimization parameters,objective function,optimization flow and optimization algorithm parameters were determined successively,the improved particle swarm optimization algorithm and gray Wolf optimization algorithm were used to optimize the process parameters of copper flotation section.The results show that the best optimization effect can be obtained by using LightGBM prediction model and Grey Wolf optimization algorithm to comprehensively optimize grade and recovery.The predicted value of grade and recovery after optimization is significantly improved compared with the real value.