Fault Diagnosis Method For Thickening Process Based On Dynamic Causality Diagram

With the rapid development of social economy,the demand for gold and other rare metals is increasing year by year,and the reserves of high-grade mineral resources decrease rapidly.Hydrometallurgy is widely used for its unique advantages in the treatment of low grade and complex ores.With the scale of hydrometallurgical production becoming larger and larger,the complexity and danger of the production process are also increasing.The Continuous production and long-term operation of equipment all can lead to the increase of the probability of failure in hydrometallurgical process.Once a failure occurs,it may cause great losses.Therefore,it is of great significance to diagnose faults in hydrometallurgy process.Due to the coexistence of qualitative and quantitative information and the confusion of deterministic and uncertain information in hydrometallurgical production process,the effect of fault diagnosis only based on data or mechanism model is often not ideal.Dynamic Causality Diagram(DCD)as a representative of expert knowledge-based fault diagnosis technology,to a certain extent,can make up for the shortcomings of fault diagnosis methods that rely entirely on process data or system model,and better avoid the loss of information,thus attracted wide attention of experts and scholars.Based on the hydrometallurgical dense process as the research background,this paper establishes the Dynamic Causality Diagram fault diagnosis model of dense process and calculates the conditional probabilities of various possible fault causes to obtain the deep-seated process fault causes in dense process.Two key steps of applying Dynamic Causality Diagram to fault diagnosis is to establish a fault diagnosis model and carry out the fault diagnosis reasoning.The paper establishes the fault diagnosis model.Firstly,analyze the process mechanism and common faults of hydrometallurgical dense process,and establish the structure of initial dynamic causality diagram by mechanism knowledge.Establish the Dynamic Causality Diagram of dense process fault diagnosis by the improved structure learning algorithm,and obtain the model parameters by the parameter learning method of causality diagram.The improved structure learning method mainly fuses the multi-expert knowledge to obtain more complete prior information.This method introduces the fuzzy number to express the expert knowledge;and the prior knowledge obtained is encoded and compressed into the scoring criterion to improve the minimum description length(MDL)score criterion,which make it more suitable for causality diagram structure learning;apply structure learning algorithm to search the optimal causality graph structure.The simulation results show that the improved structure learning method can obtain more accurate Dynamic Causality Diagram structure.On the basis of obtaining the fault diagnosis model,this paper continues to study the diagnostic reasoning process of Dynamic Causality Diagram.The conventional causal graph diagnosis algorithm does not take into account the existing trend and the degree of failure before the failure occurs.Fuzzy idea is introduced into the improved reasoning diagnosis mechanism,and a fuzzy interval is used to replace the point value,which reduces the loss of information in the process of transforming quantitative information into stereotyped information.Finally,apply the proposed method to the thickening process of hydrometallurgy,and compare results of conventional method with those of improved method,and prove the effectiveness and advancement of the improved method.