Research On Operating Performance Optimal Degree Assessment For Industrial Processes With The Coexistence Of Quantitative And Qualitative Information

As the development of the automation techniques,the demand on product quality and production benefit increases accordingly for industrial processes.Hence,it is significant to judge the optimal degree of process operating performance,under normal working conditions.Process operating performance optimality assessment considers the product quality,material cost,energy cost,economic benefit etc.The the optimal degree is assessed in real-time.The variables responsible for the non-optimal performance is identified,which offers adjustment guidance for the operators.The operating performance optimality usually reflects on the comprehensive economic benefit.The larger comprehensive economic benefit is,the better operating performance is.A factory usually evaluates the comprehensive economic benefit within a certain period,which is regarded as the production inspecting standard.But the evaluation period is often too long to be utilized for performance online assessment and production guidance.Therefore,the research on process operating performance optimality online assessment is of great significance both in theoretical study and practical application.However,due to the poor production circumstance,high measurement cost,simple basic automation facilities etc.,the quantitative and qualitative information coexist in the actual complex industrial processes,which limits the application of the traditional performance optimality assessment techniques.In this dissertation,the quantitative information indicates the variable information in the form of quantitative value.The qualitative information implies the variable information in the form of qualitative status,which is usually described by semantics.After deep study on the characteristics of the actual complicated industrial processes,the operating performance optimality assessment approaches with both the quantitative and qualitative information are researched to solve the actual assessment problem.(1)To solve the operating performance optimality assessment problem for industrial processes with more quantitative variables and less qualitative variables under a single working mode,a GMM and Bayesian theory based approach is proposed in this dissertation.The joint probability distribution of both the quantitative and qualitative variables is learned.The posterior probability of each performance grade is computed as the base of performance assessment.Under non-optimal operating performances,cause identification techniques are respectively proposed for qualitative and quantitative variables.(2)To solve the operating performance optimality assessment problem for industrial processes with more qualitative variables and less quantitative variables under a single working mode,the traditional fuzzy probabilistic rough set is modified,which takes full advantage of the hybrid types of information.Hence,the information loss is decreased and the assessment precision is increased.Additionally,the assessment result relies less on the selection of parameter.The modified technique is explored in the process operating performance optimality assessment.The online assessment strategy and non-optimal cause identification method are established to guide the production,based on the probability of each performance grade.(3)As to the industrial processes with hybrid types of information and sufficient causal knowledge under a single working mode,the conventional dynamic causal diagram is modified.The fuzzy theory and information fusion theory are respectively introduced in dynamic causal diagram to solve its information loss and fusion problems.The information loss problem originates from hard division of variable status.The information fusion problem indicates that the traditional dynamic causal diagram could not balance both the measuring and reasoning information.In addition,to reduce the computation time and storage space,the forward reasoning way is changed into hierarchical reasoning way for performance assessment.Under the non-optimal performances,an event contribution rate based cause identification method is proposed based on the dynamic causal diagram,which clearly shows the propagation way of the non-optimality.It has strong interpretation.(4)For the large-scaled plant-wide industrial processes with a large number of variables under a single working mode,a two-level multi-block hybrid model is proposed to solve the operating performance optimality assessment problem.A plant-wide process is divided into multiple sub-blocks according to its physical characteristics.Then two assessment levels emerge,namely the sub-block level and the global level.In the sub-block level,a quantitative or qualitative modeling approach is selected according to the information character of each sub-block.Besides,due to the proposed modeling data processing technique,the global performance grade is directly detennined by the worst performance grade in the sub-block level.Under a non-optimal performance,the responsible cause is identified only in the non-optimal sub-blocks.(5)In terms of the large-scaled plant-wide industrial processes with a large number of variables under multiple working modes,based on the two-level multi-block structure,the mode identification methods of both the sub-block level and global level are developed at first.Then the operating performance optimality assessment methods are respectively proposed for each stable working mode and between-mode transition.The proposed method considers three challenges at the same time,namely the coexistence of quatitative and qualitative information,the large-scaled plant-wide characteristics,and the multi-mode features.The novel method clearly shows the procedure of a between-mode transition.It assesses the operating performance under different modes,with strong interpretability and high precision.The proposed method to solve the operating performance optimality assessment problem for industrial processes with the coexistence of quantitative and qualitative information is applied to a simulated gold hydrometallurgy process.The feasibility and validity of the proposed approaches is illustrated.The proposed methods overcome the difficulty in practical application to some extent and combine the theoretical research with the practical application.