Research On Control Of Cu-Ni Sulphide Flotation Process

With the massive exploitation of mineral resources,high-grade mineral resources are depleted,and the distribution of useful minerals in ores is becoming more and more fine and more complicated.In addition,the requirements and precision of sorting fine-grained and ultra-fine materials in materials and chemical industries are more and more The higher the yield^[1-4],in order to make full use of this part of low-grade mineral resources,the foam flotation process has more and more advantages over other beneficiation methods,and it has become the most widely used and best performing beneficiation method^[5].Different from the traditional gravity beneficiation method,the foam flotation process is a kind of beneficiation method for material sorting by utilizing the difference in floatability caused by the difference in physical and chemical properties of mineral particles.Among them,multiple process flows are involved.In the traditional flotation operation,on-site process personnel judge the production condition of froth flotation based on the shape,color and dynamic characteristics of the flotation foam according to the production experience,and make appropriate adjustments according to the production experience.This kind of operation mode is relatively dependent on the experience of the field personnel,and it is easy to produce large product quality fluctuations.Due to the long time lag of the system itself,when the product quality problem occurs,it takes a long time to adjust to ensure the product quality is stable.Therefore,it is urgent to introduce automation control technology to solve these problems.The froth flotation process is a complex multiphase,polymorphic,multi-input,and output-coupled system.There are many parameters affecting the quality of the final product,including:raw material grade,grinding method,grinding fineness,slurry concentration,slurry particle size,ore feeding speed,flotation liquid level,slurry temperature,pulp PH value,flotation reagent type,Flotation dose,flotation time,intake air amount,intake pressure,etc.Therefore,the traditional mechanism-based modeling method can not effectively fit the flotation process model.In recent years,most experts and scholars tend to use machine learning to fit this complex system,and at the same time achieve relatively satisfactory results.The project is based on the relationship between the control parameters of the flotation flotation process,the process parameters and the image features,and establishes the relationship model between the process parameters and the dosing amount,the relationship between the image features and the dosing amount.This thesis is oriented to industrial applications.Aiming at the technical difficulties of poor predictive ability and generalization ability of the process automation system,a compensation control strategy model based on multi-source data information entropy is proposed.The main work is as follows:?1?The problem of poor prediction ability and generalization ability for machine learning models.Real-time collection of field data through nearly two years.Machine learning samples cover most of the production process,which increases the generalization capabilities of the model.By performing data cleaning and dimensionality reduction on the collected data samples,the model calculation speed can be improved without sacrificing the accuracy of the model.?2?The problem of poor adaptability and low precision for a single model.The multi-source data?process data and bubble characteristics?affecting production indicators are analyzed,and then the prediction sub-model and error compensation model are established in turn.Finally,the drug dosage prediction model is constructed through information entropy integration.?3?A flotation data acquisition system and production guidance system were set up in the Jinchuan Group nickel ore flotation production line.A dedicated foam image analyzer was developed based on the site conditions,and functions such as data acquisition,data analysis,data cleaning,database storage,and production guidance were developed according to functional requirements.