Data And Knowledge-driven Based Identification And Self-healing Control Of Abnormal Conditions In Thickening-dewatering Process

Thickening dehydration process is widely used in hydrometallurgical processes,flotation and other industrial processes to realize the solid-liquid separation of slurry.With the constant improvement of the basic automation level of the thickening dehydration process,the selfhealing control under abnormal conditions has attracted more and more attention in its development.The environment of production field in the thickening dehydration process is relatively abominable,although the basic automation level is up to the standard,the optimization control level is poor,meanwhile,due to the poor stability of the upstream process,the large fluctuation of the feeding slurry characteristics and the improper field operation strategy,the abnormalities in the thickening dehydration process often occur.The occurrence of abnormalities can cause failure and damage of devices,and severely to cause serious economic loss due to production stagnation.At present,the abnormalities are basically decided by the operators according to their own experience.Different operators with different professional skills give different decision-making results,and most of the decisions given by operators are qualitati ve,unable to conduct quantitative and accurate analysis.Some operators may conduct analysis based on field data,but there is a large amount of uncertainty information in the actual production process,as well as the influence of external environment and production conditions,resulting in relatively little effective information about abnormal working conditions that can be collected.In addition,because it is difficult for operators to treat multi-source information at the same time,it is often difficult to give an effective decision scheme.Therefore,according to the characteristics of abnormal conditions,it is of great theoretical significance and practical value to develop an effective self-healing control scheme to ensure the safe and stable operation of the thickening dehydration process.There is a lot of expert knowledge and operation experience that can be used in the process of identification of actual abnormal conditions and decision-making of self-healing control.Through the reasonable integration of expert knowledge and information of abnormalities data,an accurate abnormal conditions identification model can be established,and an effective self-healing control scheme can be developed on this basis.Bayesian Network is an effective tool for integration of expert knowledge expression and data information.Therefore,from the perspective of solving practical problems,this paper studies a Bayesian Network-based identification method for abnormal conditions of thickening dehydration process and a data-driven self-healing control method for abnormal conditions of feeding process.(1)Based on the information provided by pressure sensors inside the thickener,a datadriven dynamic model structure was established to predict the future state of the thickener.KPRM model was used to describe the concentration of the underflow concentration,and dynamic ARX model was used to predict the future changes of the pressure sensor.(2)According to the identification problem of abnormal conditions in the thickening dehydration process,this paper proposes a Bayesian Network-based modeling method for abnormal conditions identification and online application strategy.On the basis of summarizing the common abnormalities in the thickening dehydration process,collecting the process variables related to the abnormalities,as well as clarifying the causes and corresponding phenomena of the abnormalities,the abnormal conditions identification model of the thickening dehydration process was established.By using the online evidence information of abnormal conditions,the real-time decision scheme of abnormal conditions identification is obtained.(3)Based on the identification modeling of abnormal conditions,aiming at some abnormal conditions that can be finely optimized,this paper proposes data driven-based self-healing control scheme,the proposed multilayer optimization control strategy allows the sorting of several competing objective function on the basis of prioritization,and optimal control decision is obtained by solving function step by step,previous tier will provide the reference information for the new tier.The simulation application experiment was carried out through the optimization control system of thickening dehydration process and the laboratory simulation platform.This paper proposes Bayesian network-based identification model on abnormal conditions of thickening dehydration process,based on this,we put forward the self-healing control scheme of the abnormal feeding conditions.Through the simulation results,we confirm the validity of the proposed method,both can guarantee the safe operation of the thickening dehydration process,and reduce the energy consumption as well as provide convenience for the operators,it provides a new train of thought,being of great significance,for solving the safe operation control problem of the abnormal conditions of thickening dehydration process.