Research On Status Diagnosis And Superheat Dgree Identification Of Aluminum Reduction Cell Based On Fire Eye Image

With the development of large-scale,comprehensive and complex aluminum electrolysis industry,it is more and more difficult to model the aluminum electrolysis industry process,which involves not only the inherent complex mechanism problems such as nonlinearity,uncertainty,large time delay,parameter distribution and time-variability,but also the objective environment and human factors.As the main equipment for producing electrolytic aluminum,the electrolytic cell’s production status directly affects the output and quality of aluminum.Therefore,it is necessary to monitor and diagnose the electrolytic cell’s status in real time in the process of aluminum electrolysis.During the normal production of electrolytic cells,keeping the aluminum electrolysis production in low superheat degree can improve the current efficiency,and thus play a role in improving economic benefits,saving energy and protecting the environment.In this thesis,based on the actual production of an aluminum electrolysis plant,the diagnosis of the thermal equilibrium state of aluminum electrolysis cells based on the flame image is studied.Aiming at the problem of energy saving and consumption reduction in aluminum electrolysis production process,superheat degree recognition research based on flame image was carried out,and effective experimental results were obtained.The main contents of this thesis are summarized as follows:First of all,by investigating the research status of this subject,taking the aluminum electrolysis production process as the research background,the method of diagnosis of electrolytic cell’s status and superheat degree identification based on flame image is studied,which can promote the stable production of aluminum electrolysis process and improve current efficiency,and achieve the purpose of energy saving and consumption reduction.Secondly,in the aluminum electrolysis plant,the industrial camera is used to obtain the video stream of the fire hole,and the fire hole video is decomposed into many fire hole images.Aiming at the fact that the fire hole images contain a large amount of redundant data,a fire hole image segmentation method is proposed.At the same time,in order to increase the number of samples,the data of the fire eye image is expanded.Then,the shallow features and deep features of the fire eye images are extracted to complete the data preprocessing,which provides conditions for the diagnosis of electrolytic cell’s status and the identification of superheat degree.Then,aiming at the characteristics of small labeled samples and large unlabeled samples of flame images with different cell’s status,a semi-supervised kernel extreme learning machine(SSKELM)algorithm is proposed by making full use of the useful information contained in unlabeled samples,and establish the aluminum electrolytic cell’s status diagnosis model to realize the real-time diagnosis of the cell’s status.The experimental results show that this method improves the operation efficiency and diagnosis accuracy in the process of electrolytic cell’s diagnosis.Suggestions on parameter adjustment are given when the production state of aluminum reduction cell is abnormal.Finally,in order to achieve low superheat degree production of aluminum reduction cells,thereby improving the current efficiency,it is necessary to identify the superheat degree in real time under the normal production status of electrolytic cell.Therefore,this thesis proposes a superheat degree recognition model,which uses convolution neural network to extract the deep features of the fire eye image,at the same time,the shallow features of fire eye image are extracted,and then fuses the deep features and the shallow features to form the input feature matrix.The fused feature matrix is input into the improved least squares twin support vector machine optimized by chaotic grey wolf optimization algorithm as a classifier for superheat degree recognition.The experimental results show that the superheat degree recognition model proposed in this thesis has good generalization ability and high recognition accuracy.When the identification result of superheat degree is high,this thesis uses the Projection Pursuit(PP)algorithm to analyze the correlation degree of production parameters,and gives suggestions on parameter optimization and adjustment.