This paper presents the results of monitoring data analysis for experimental areas producing primary aluminum based on the RA-300 and Soderberg technologies. The authors considered the events of technological disorders, such as anode effects and formations on anode face, carried out the statistical analysis of technological parameters and process disruptions in relation to the extreme values of the parameters. The results of analysis allowed us to obtain new knowledge about technological disorders and features of abnormal working states of aluminum reduction cells, detect the technological patterns, conditions and causes for the occurrence of technological disorders.
Quality control of the aluminum smelting process is based on monitoring the technological parameters of aluminum reduction cells [1]. Controlled parameters include measurements of the chemical composition of the melt, physical characteristics of the melt, energy balance, and other working variables. The most common technological disorders in aluminum production include anode effects and distortions of the anode geometry formed during electrolysis [1][2]. Anode effect is a phenomenon characterized by decreasing in the alumina dissolution and a significant increase in the cell voltage. Distortions of the anode geometry are classified into formations on the anode face ("spikes", "laggings", "chunks"), and anode destruction ("corner shedding"). "Spike" is a formation of a cylindrical or conical shape [3]. "Lagging" is the formation of a rectangular crosssection or unevenness, occupying up to 50-60% of the anode bottom [3]. "Chunk" is a special term, adopted at Bratsk aluminum smelter, for formations of any form weighing up to several hundred kilograms on the Soderberg type of carbon anodes.
"Corner shedding" is a physical loss of large carbon pieces from the anode surface [4]. Statistical analysis of the monitoring data allows us to obtain new knowledge about technological disorders and features of abnormal working states of aluminum reduction cells.
This paper presents the results of the monitoring data analysis in the context of extreme values of technological parameters in order to explore technological patterns and detect the technical conditions and causes for the occurrence of technological disorders.
The analysis was carried out on monitoring data for three experimental areas: Khakas aluminum smelter (KhAZ), Boguchansky aluminum smelter (BoAZ), that use the RUSAL's proprietary RA-300 technology, and Bratsk aluminum smelter (BrAZ) with Soderberg technology. Technology determines the design features of aluminum reduction cells and plants and, as a result, the scope of controlled technological parameters and types of process disruptions [5,6]. In the experimental area of KhAZ, there were 6784 cases of "spikes" and 8123 anode effects over the observation period of 2014-2019 years. In the experimental area of BoAZ, there were 3872 cases of "laggings", 349 cases of "spikes" and 4689 anode effects over the 2019 year. In the experimental area of BrAZ, there were 30989 cases of "corner shedding", 2986 cases of "chunks" and 73412 anode effects over the observation period of 2015-2019 years.
To determine the extreme parameter values, the variability of a set of parameter values was estimated for each cell, and the ranges of reliable values were found as μ ± σ, where μ is the mean value of the data series, σ is the standard deviation. Parameter values outside this range are considered as extreme values. The result of the analysis of the statistical characteristics showed that the cells, in the context of the same parameters, differ significantly both in the average values of the parameters and in their deviation intervals. For each experimental area, the parameters with the largest percent of anomalies over entire observation period were determined, among them: for KhAZ -Back EMF, Dose of alumina, Coefficient of anode-cathode distance; for BoAZ -Cryolite ratio, Coefficient of anode-cathode distance, CaF2 concentration, for BrAZ -AlF3 dose, Minimum distance, MgF2 concentration.
In order to explore the events of technological disorders in relation to the extreme values of the parameters, we considered the dates of disorders detection (in cases of anode effects) and the five-day period preceding the registration of the disorder (in cases of any formations). As a result, for each type of disorder, the distribution of cases accompanied by extreme values of the parameters was obtained (Tables 1, 2). It can be seen from the diagrams that during the period of occurrence of disorders, as a rule, outliers of values are observed for several parameters, but there are cases when technological disorders occurred when the values of the parameters were within the statistical norm. At the same time, the number and composition of parameters are different for different types of disorders. Moreover, regardless of the aluminum production technology, in the case of anode effects, the number of parameters with extreme values is less than in the case of formations on the anodes. So, KhAZ is characterized by extreme values of 2-4 parameters for the anode effect and 6-8 parameters for "spikes"; BoAZ is characterized by extreme values of 1-2 parameters for the anode effect, 5-7 parameters for "lagging" and 4-6 parameters for "spikes"; BrAZ is characterized by extreme values of 4-6 parameters in case of anode effects, 9-11 parameters in cases of "corner shedding" and "chunks".
Also, this research included the correlation analysis of parameters in part of extreme values. The result demonstrated a quite strong relationship between the following parameters: for KhAZ: Duration of pouring and Velocity ratio with correlation coefficient -0.69, Amperage and Bus voltage with correlation coefficient The analysis of detailed data made it possible to detect the dependence of the number of technological disorders on the location of the anodes in the cells (Figures 4, 5). For instance, at KhAZ we can observe significantly more "spikes" on the anodes of the front side of the cells -in its central part (anodes No. 9 and No. 10) and along the edges (anodes No. 1 and No. 18). At BoAZ, in contrast, we can observe more "spikes" and "laggings" on the anodes of the backside of the cells -mostly around the edges (anodes No. 22, 23 and 33, 34). Thus, as a result of a detailed analysis of the monitoring data in the context of extreme values of the controlled technological parameters, the authors determined the characteristic dependencies and features of the functioning of individual units of the aluminum production complex in atypical operating modes.
In this paper, the authors carried out the analysis of the aluminum production process disruptions in terms of the extreme values of the technological parameters for three experimental areas: KhAZ, BoAZ and BrAZ. The analysis of the statistical characteristics showed that the cells, in the context of the same parameters, differ significantly both in the average values of the parameters and in their deviation intervals. Within the study, for each experimental area, the parameters with the largest percent of anomalies were determined. The distribution of disorders events (in the cases of anode effects and formations on the anode face) by the number of parameters with extreme values was obtained. The results showed that in most cases (75%-95%) a large number of disorders are accompanied by a large number of extreme values of technological parameters. The analysis of detailed data revealed the dependence of the number of technological disorders on the location of the anodes in the cells.
Thus, the results of research allowed us to obtain new knowledge about technological disorders and features of abnormal working states of aluminum reduction cells, detect the technological patterns, conditions and causes for the occurrence of technological disorders.