Research On Safe Operation Control Method Of Dense Medium Coal Preparation Process Based On Bayesian Network

Coal is the important foundation of economic development and industrial production,and it occupies an important position in China’s national economic system.Dense medium coal preparation is an important technology of industrial coal preparation,which is widely used by various coal enterprises because of its high separation efficiency and accuracy.However,the coal preparation site is prone to abnormal working conditions due to the complex and changeable production environment and frequent changes in raw coal properties.The occurrence of abnormal working conditions affects the comprehensive economic benefits of enterprises,and even threatens the personal safety of operators.Thus,it owns an important theoretical significance and application value to make a reliable safety control decision according to the characteristics of abnormal working conditions to ensure the safe and stable operation of coal preparation process.Based on the in-depth understanding of dense medium coal preparation process technology and common abnormal working conditions in the coal preparation process mechanism,this thesis researches the safe operation control method of dense medium coal preparation process based on Bayesian network(BN).The main research content is as follows.(1)Due to the long process flow,complex mechanism and low degree of equipment automation,the working conditions of dense medium coal preparation process fluctuate violently and the coal separation process is occasionally prone to abnormal operating conditions.Currently,the coal preparation process mainly relies on manual adjustment of control parameters to eliminate abnormal working conditions,which makes it difficult to timely control the cleaned coal ash and other key indicators within the normal range.For the safety control problem of common abnormal working conditions in dense medium coal preparation process,the safe operation control method and online application strategy based on Bayesian network are proposed.Firstly,the causes,phenomena and corresponding control measures of abnormal working conditions are analyzed,and the expert knowledge related to abnormal working conditions are collected.Secondly,the safety control Bayesian network model integrating expert knowledge and abnormal working condition data information is established.Then,the online relevant data are collected and input into the Bayesian network model as evidence information,and the posterior probabilities of different control variables are obtained by Bayesian inference algorithm.The corresponding control decisions are made and implemented according to the principle of maximum posterior probability,so as to restore the production process to the normal and stable operation state.Finally,the effectiveness of the proposed method is verified by testing on the simulation platform.(2)In the process of dense medium coal preparation,there is a certain amount of expert knowledge that can be used to determine the causal relationship between Bayesian network nodes.However,when the number of Bayesian network nodes is large,the network structure given by different experts may not be unique.Thus,the method of obtaining Bayesian network structure by using expert knowledge has strong uncertainty.To address this issue,the safe operation control model modeling method based on active learning and Bayesian network is proposed.The safety control Bayesian network model is constructed by using the observation data and experimental data information.Firstly,the skeleton graph is obtained by conditional independence test,and then each edge in the skeleton graph is oriented.In order to reduce the amount of data of structure learning,the specific variables selected by active learning are used for intervention experiments,and the Bayesian network structure is established by observing the change of probability distribution of other variables,so as to improve the efficiency of Bayesian network structure learning.Finally,the effectiveness of the proposed method is verified by Asia network and dense medium coal preparation process data.(3)The production environment of dense medium coal preparation is complex and changeable,and new abnormal conditions are inevitable.The performance of the original model is degraded due to the lack of ability to adapt to the changes of abnormal working conditions.Thus,a safety control Bayesian network model updating strategy is proposed to ensure the decision-making performance of the model under new abnormal conditions.Firstly,expert knowledge is used to judge whether the control decision provided by the existing security control model is reasonable.If the control decision can’t eliminate abnormal working conditions,the Bayesian network structure and parameters are updated according to different conditions to ensure the updated performance of the Bayesian network model.Considering that some structures in the original model can still describe the probability distribution of corresponding variables in the new data,a Bayesian network structure update learning method based on incremental learning is proposed.This method can retain the part of the old model that can adapt to the changes of the environment,and update the remaining structure,so as to improve the efficiency of model updating.Finally,the effectiveness of the proposed method is tested by Asia network and dense medium coal preparation process data.The thesis includes 25 figures,22 tables and 88 references.