An Intelligent Operation Adjustment Strategy Based On Bayesian Network For Copper Cleaning Process

On one hand,in flotation processes,due to the complex mechanism of flotation processes,uncertainties and a large number of strong coupling process variables,it is difficult to establish an accurate mathematical model.On the other hand,due to the changes of ore sources,multiple ore sources and other reasons,the working conditions of flotation processes change frequently and it is difficult to ensure stable operation of production for a long flotation process.Presently,the adjustment of pulp level and aeration rate mainly relies on the visual observation of the froth state by the operator based on their own experience on the flotation site.However,this adjustment method is subjective and it is difficult to adjust the operation reasonably.Operators often fail to detect changes of working conditions and adjust operating parameters timely,resulting in "over aeration rate","slotting" or "running".Eventually,the grade of the concentrate is fluctuating and the economic benefit is poor,which makes it difficult to meet the actual production requirements.Therefore,we intend to adopt an intelligent operation adjustment strategy to solve the problem that the copper concentrate grade is difficult to meet the standard.In this thesis,the operation adjustment strategy of the copper cleaning process for different working conditions can not only reduce the labor intensity of the operators,but also facilitate them to make correct decisions in time when the working conditions change.The operation adjustment strategy can achieve the goal of guiding actual production,increasing production efficiency and maximizing economic benefit.Firstly,this thesis establishes an offline model based on Bayesian network for the copper cleaning process.Secondly,the online reasoning of operation adjustment strategy for copper cleaning process is carried out through the offline Bayesian network model.Thirdly,we predict the success probability of the qualified copper concentrate grade by reasoning results.Therefore,operators are advised to make adjustment decisions based on the success probability.Finally,the validity of the operational adjustment strategy based on Bayesian network reasoning is verified by test data.The main contents of this thesis are as follows:(1)In view of the different operational adjustment strategies under different working conditions,this thesis proposes a method of working conditions classification based on PCA-AP for copper flotation process.Firstly,the PCA is used to reduce the attributes of ore property parameters and the froth image feature parameters.Then the comprehensive feature parameters are obtained by dimension reduction and are classified by AP clustering.(2)Offline modeling based on Bayesian network for the copper cleaning process is established.Firstly,according to the mechanism of the copper cleaning process,the network nodes are selected and the node levels are divided.Then,according to expert knowledge,we can determine the causal relationship between nodes.Therefore,the Bayesian network structure of the copper cleaning process is determined.Finally,we use actual industrial data to learn the Bayesian network parameters by the maximum likelihood estimation method.(3)Online reasoning based on Bayesian network for copper cleaning process operation adjustment is proposed.According to the established offline Bayesian network model,inputting the working condition category,target and the initial state of operation as evidence,we can reason the operation adjustment strategy under different working conditions for the copper cleaning process.(4)The prediction about the success probability of the copper concentrate grade which is adjusted from the unqualified to the qualified is proposed.The working condition category,the initial state of operation and the obtained operation adjustment strategy by Bayesian network reasoning are inputted into the Bayesian network as evidence.Then we can predict the success probability of the qualified copper concentrate grade.Therefore,we can help operators to make adjustment decisions based on the success probability.Finally,the validity of the operation adjustment strategy based on Bayesian network reasoning is verified by test data.