Short-Circuit Detection And Operation Performance Assessment For Copper Electrorefining Process Based On Infrared Image

In copper electrorefining process,short circuit of electrode plates in the cell will cause abnormal temperature of electrode and electrolytic reaction stop,result in lower electrical efficiency,higher energy consumption,seriously reduce the quantity and quality of the electrolytic product(cathode copper),and produce bad influence on the electrolytic process.We utilize infrared thermal imaging technology to obtain the thermal distribution picture of the cell surfaces.Through image analysing to detect the short-circuit electrode and combining process mechanism and data to realize real-time state evaluation of electrolytic process are of great significance to improve production efficiency and reduce electricity consumption.However,on the one hand,the special structure of electrolytic cell,the complex heat conduction in the cell,the short-circuit target in a variety of forms,coupled with occlusion and random environmental interference,all the above factors bring difficulties for short-circuit targets detection.On the one hand,the long cycle leads the state evaluation lags,the large scale of electrolysis production results complex correlation ship among the process variables of subsystems,The complex electrolysis reaction mechanism cause multiple evaluation dimensions exist simultaneously.All these characteristics brings challenges to the real-time operation performance assessment of electrolytic production.In this paper,combined with the characteristics of copper electrolysis process,short-circuit identification and real-time operation performance assessment method of copper electrolysis process are studied based on infrared thermal image.The main research work and innovation of this thesis are as follows(1)Aiming at the problems of complex thermal distribution characteristics of different functional areas and blurry discontinuous boundary between functional areas,that causes each function area accurate segmentation difficult.A hybrid superpixel merging method was proposed to segment functional regions.The Simple Linear Iterative Clusting(SLIC)method is used to over-segment the cell images to overcome the complex gray distribution in the region.Then,the region adjacency graph based combination and classification method are used to complete the superpixel fusion by using the superpixel features in different scales.Under the constraint of the spatial association of the features among functional regions,a threshold of Moran’s I is setted as the stopping condition of the region iterative merging to avoid over merging among regions.The experimental results show that this method can effectively ensure the integrity of functional regions segmentation and provide foundation for the follow-up works.(2)In order to solve the problem that the detection accuracy of short circuit in canvas area is affected by the occlusion and various forms and scales of short circuit,a short circuit identification method based on synthetic samples and improved Faster R-CNN is proposed But the network training lacks of sufficient number of labeled images and manually annotation is a tedious work.An automatic synthetic method of labeled image is proposed.On the basis of the analysis and establishment of the key variables leading to cell surface image diversity,simulation methods of the key variables are designed to maintain the samples diversity.Thus any number of samples can be labeled automatically without manual labeling.Attention module,introduced in the network architecture,through integrating the semantic segmentation result of small scale targets by U-Net into the synthetic infrared data concentration,to increase the network’s perception of small scale targets.Then combine strategies of anchor parameters tuning and transfer learning,the network is training on the synthetic data.The improved identification network can effectively improve the short circuit detection accuracy of canvas area.(3)In order to cope with the problem of missing detection caused by the unevenn occlusion which makes the grayscale change of the cooresponding area not obvious,a two-stage short-circuit detection method based on local contrast is designed.Firstly,based on the spatial continuity of gray distribution between short circuit and the highlighted background,the short circuits are separated from the normal working ones as a part of the background by removing the background.Then,the difference of Gaussian filter(Do G)is used to locate the normal electrode.After the texture period estimation,through comparing the distance between two adjacent located electrodes and the texture period to identify the suspicious short circuit area.Experiment results show that the proposed method can effectively reduce the miss detection rate of short circuit.(4)An online assessment model based on two-layer multi task collaborative clustering and Fisher discriminant analysis is proposed to evaluate the operation performance of copper electrolysis.According to the characteristics of the physical structure of the copper electrolysis system,the system is divided into two layers: the core layer and the peripheral layer,and the evaluation problem is decomposed into two subsystems evaluation.For the core layer,the multi task clustering learning of infrared thermal image is used to realize the online evaluation of electrolysis reaction state.On this basis,the evaluation results and the process variables of the peripheral layer together constitute the feature space.Through Fisher discriminant analysis of these off-line data,the characteristic attributes of the operation state are extracted,and the evaluation model is established.By comparing the similarity between the real-time data with each state level,the online evaluation of the operation state of the whole copper electrolysis process is realized.On the basis of theoretical method research,according to the actual process characteristics,a copper electrolysis cell surface temperature acquisition platform based on infrared thermal imager is designed,and an intelligent monitoring system is developed.The automatic detection of short circuit is realized,and the detection time of the whole workshop is less than8 minutes,which greatly improves the detection efficiency,and the detection accuracy meets the needs of actual production.At the same time,the online assessment of copper electrolysis operation performance is realized,which provides a good foundation for reducing the power consumption and improving the quality of cathode copper.