Abnormal Condition Identification Of Fused Magnesium Furnace Smelting Process Based On Knowledge And Data

Fused magnesia is a widely used refractory material.Its main production equipment is fused magnesium furnace.In the production process of fused magnesia,the current setpoint is unreasonable due to frequent changes in raw material impurity components and particle length.When the fused magnesium furnace is operated for a long time under an unreasonable current setpoint,an abnormal condition will occur.The occurrence of abnormal conditions will result in a decline in product quality and even a serious production accident.The fused magnesium furnace smelting process is a typical complex industrial process and it is impossible to establish accurate mathematical models.However,the operators on site can use their own experience and knowledge to combine real-time current data,furnace conditions and sound signals to detect anomalies in time.This thesis is the research work of using the empirical knowledge and production data to identify abnormal conditions.Firstly,the development status of the anomaly identification method is reviewed,and three key conditions in the smelting process of fused magnesium furnace are introduced.In addition,the four typical abnormal conditions in the smelting process are analyzed,and the characteristics and causes of the abnormal conditions are obtained.Secondly,this thesis establishes a heating and melting abnormal condition identification model based on rule-based reasoning,analyzes the characteristics of the image in abnormal heating and melting,filters the image,segments the target area and extracts the relevant image features.Besides,it proposes the short-time energy of current change rate and uses it as the current characteristic in heating and melting anomaly identification together with the tracking error of current.The data fusion of basic data,image information and sound information is realized,and enables sequence synchronization.Moreover,this thesis also verifies the accuracy of the model through simulations.Finally,this thesis establishes an abnormal exhausting identification model based on case-based reasoning.For the characteristics of abnormal exhausting,the characteristics of the fusion image,current and sound are combined as a representation of the case.It also proposes a similarity calculation method based on the combination of weighted Euclidean distance and weighted cosine function.According to historical data,genetic algorithm and introspection learning are used to determine the weight of case features.The case retrieval method of group decision making is used to fully exploit the potential knowledge of data.In addition,this thesis designs case reuse and case storage strategies and verifies the rationality and effectiveness of the model through simulation experiments.