Research On Bayesian Network And Its Application In The Design Decision Method Of Beneficiation Principle Process

With the rise and development of artificial intelligence,it has created conditions for the automation of complex industrial process design to become an important part of knowledge automation and intelligent manufacturing.Bayesian networks have great advantages in combining empirical knowledge with data to build networks to make decisions.The beneficiation process is a typical complex industrial process,its design has been carried out through trial and error,and previous empirical experience and actual design data have not been used effectively.Therefore,it is of great significance to study the Bayesian network decision method and apply it to the design of the beneficiation process to realize design decision automation.This article first introduces the domestic and international development present condition of Bayesian network and beneficiation process design,and conducts in-depth research on the theoretical basis and modeling process of Bayesian network.At the same time,it also introduces the composition process and basic characteristics of the beneficiation process and clarifies the design content of beneficiation principle process.Aiming at the design decision of the grinding process,a Bayesian network-based grinding principle process design decision making method is proposed,and the grinding fineness decision making process is taken as an example to introduce the Bayesian network structure construction,parameter learning,and inference process in detail.The specific modeling process is to select relevant variables as Bayesian network nodes according to the design mechanism and expert knowledge of the beneficiation process,and determine the classification of the nodes and the causal relationship between the nodes,and complete the construction of the Bayesian network structure.Then,the case database of actual beneficiation process design was sorted out to build the data set,and the Bayesian network parameters were learned to obtain the conditional probability table of Bayesian network,thereby the Bayesian network decision model of each part is established to realize the design decision of the fineness of grinding,the number of grinding sections,the number of crushing sections and the type of crushing equipment,and the actual process design data verifies the validity of the decision model.On this basis,a Bayesian network decision making method based on mutual information and Bagging algorithm was proposed for the characteristics of the selection principle process with many influencing factors and a small data set,and applied to the selection principle process design decision.Among them,the Bayesian network decision making method based on mutual information is applied in the selection method decision making stage.In the decision making phase of the selection segments and selection times,in order to verify that the decision is more accurate when the data set is larger,the Bayesian network based on Bagging algorithm is applied to realize the decision of the number of selected segments and selected times,and the effectiveness of the method is verified by the actual process design data.Finally,through the in-depth analysis of the mechanism of the pharmaceutical system,the design content of which is determined.In view of extremely complex influencing factors of pharmaceutical addition,the data set is a mixed type and the data set is small,this paper proposes a Bayesian network decision making method based on neighborhood rough set and Bagging algorithm,and its application to the design decision of the pharmaceutical system.Among them,the Bayesian network decision making method based on neighborhood rough set and Bagging algorithm is applied in the decision making phase of the collector type decision stage,on this basis,in order to determine the specific type of collector,a production rule method is introduced,In the decision making stage of inhibitor dosage,the Bayesian network decision making method based on Bagging algorithm is applied.The validity of the method is verified by actual process design data.