Search

[Display omitted] • Anti-seepage mechanism of DB and the phase composition of SDB are analyzed in detail. • Previous studies have confirmed that SDB cannot be treated by VD or VSR process. • The detailed reduction sequence and mechanism of each component in SDB were obtained. • The recovery Na and F in this process can reached 95.35% and 99.68%, respectively. • The reduction slag obtained meets the standard of making new refractory materials. After the failure of an aluminum battery, a large amount of aluminum cell spent pot lining (SPL) material is produced. The second cutting material of the SPL is mainly composed of spent dry barrier (SDB) material, which is a kind of hazardous solid waste. In the present study, a vacuum aluminothermic reduction process was used to separate and recycle the SDB. The feasibility of separation of each component in the SDB under vacuum and high-temperature conditions was evaluated through thermodynamic analysis. Separate reaction experiments of aluminum and each component in the SDB were carried out, and the reaction mechanism in the aluminothermic reduction process of the SDB was examined. The optimum process conditions of the vacuum aluminothermic reduction SDB process were studied. The results showed that the recovery of sodium and fluoride in the SDB reached 95.35% and 99.68%, respectively, using a reduction temperature of 1150°C, reduction time of 2 h, and aluminum powder content of 10 wt%. The main phases in the slag after reduction were SiO 2 and Al 2 O 3 , and sodium and fluorine concentrations were below 0.1 and 0.05 wt%, respectively, which could be returned to use as raw materials for the dry barrier of aluminum cells or refractory materials. This method can realize the separation, recovery, and reuse of SDB, maximize the value-added utilization of the resource without producing waste such as slag, water, or gas, and achieve cleaner production. [ABSTRACT FROM AUTHOR]

Anode effect (AE) is influenced by adaptive spatial and temporal factors in the aluminum electrolysis process. Anode current signals (ACS) are the only online distributed signals which provide spatial temporal characteristics of AE. How to extract interpretability adaptive spatial and temporal characteristics of ACS for AE prediction is a challenging problem. In this paper, a multi-generator adversarial dynamic spatial–temporal shapelet network is proposed to capture the explainable spatial–temporal characteristics. Multi-generator adversarial training added into shapelet learning is used to reinforce the interpretability of shapelets. A dynamic distance calculation strategy is proposed to extract dynamic shapelets of multi-variable time series whose dynamic discriminative subsequence is not restricted to dimension. Based on the proposed self-regulative spatial network and diversity regularization, the extracted shapelets will have adaptive spatial networked correlations and diversity. Moreover, the proposed method can guide the domain technicians to locate the abnormal conditions. The results of experiment using real-world aluminum electrolysis plant data show that the proposed method is feasible and quite effective to detect anode effect earlier and enhance the interpretability of shapelets. [ABSTRACT FROM AUTHOR]

• An enhanced LSTM architecture is proposed that the target output is extracted from the learning process. • A self-supervised model is constructed with long-short-term memory network for PT prediction. • Comprehensive experimental studies are conducted to evaluate the performance and generalization. This article studies the influence of pot temperature or electrolyte temperature in the aluminum reduction production. Specifically, these indexes reflect the distribution of the physical and energy field of the reduction cell, the current efficiency, and the lifespan of the aluminum reduction cell. Therefore, the pot temperature detection and identification are two critical and significant issues in the whole production process of aluminum electrolysis. However, due to the low measurement accuracy and high maintenance costs with the thermocouple sensor in the practical production process, the real-time measurement of pot temperature index is still a major challenge, which motivate us to develop a self-supervised soft sensor method based on deep long-short term memory (LSTM). Under the constraint of the limited samples, the proposed method achieves a competitive performance. First, the input variables are selected according to the expert experience. Then, a deep self-supervised model is built. Finally, the proposed self-supervised LSTM model is applied to real-time detection in an industrial electrolysis production case. The performance in the experiment outperforms other existing methods in terms of both accuracy and robustness aspects. [ABSTRACT FROM AUTHOR]

Aluminum electrolysis cells are characterized by harsh environments where several measurements have to be done manually. Due to the operational costs related to manual sampling, the sampling rates of these measurements are low. Therefore, the information in the data can be limited, making it challenging to develop robust data-driven methods for aluminum electrolysis process. A broad array of physics-based models have been developed throughout the years to provide excellent system knowledge about the dynamics in the aluminum electrolysis cells. However, due to highly complex and interrelated sub-processes, the state-of-the-art physics-based models are insufficient to accurately express the dynamics in the cell. The combination of inadequate prediction models and low sampling rates makes estimating process variables in the aluminum electrolysis process less accurate than what is desired for optimal and safe operation of the cells. In this paper, a novel hybrid modeling approach that addresses insufficient prediction models and low sampling rates is suggested. The novel hybrid modeling approach involves manipulating a measured signal with a first principle model estimate. This manipulation, which consists of subtracting the first principle model estimate from the measurements of the signal, produces a residual that represents the unmodeled dynamics in the signal. Since the unmodeled dynamics of the measured signal is much sparser than the measured signal itself, this manipulation enables utilizing a powerful technique for estimating sparse signals from only a few measurements. The technique is called compressed sensing. The manipulated data is used as input data in a compressed sensing algorithm which produces a high fidelity estimate of the unmodeled dynamics in the original signal. Thus, the novelty in this article is two-fold. First, compressed sensing is introduced to the field of aluminum electrolysis. Second, the novel technique of sparsifying a measured signal with a first principle model in order to utilize compressed sensing on the measured data is introduced. The signal estimate of the unmodeled dynamics is integrated into an Extended Kalman filter as a pseudo measurement to improve the estimation of the system states. The novel method applies to signals with stationary periodical unmodeled dynamics. The case study in this article is conducted on simulated data of a sub-process describing the mass balance in an aluminum electrolysis cell. • A new modelling paradigm called the hybrid analysis and modeling (HAM) is introduced • Application of compressed sensing to seemingly non-sparse data using HAM • HAM is shown to improve estimation with Kalman filter by using unmodeled dynamics [ABSTRACT FROM AUTHOR]

In the aluminum electrolysis process, the accurate identification of anode effect (AE) can improve production efficiency. However, the existing methods fail to effectively capture the features of the anode current signal (ACS) due to its complex dynamic characteristics and temporal–spatial dependence. To address this issue, we propose a Process Knowledge-guided Deep Temporal–spatial Feature Learning Network (PKG-DTSFLN). We believe that knowledge and production data are complementary. Knowledge has potential to deduce beyond observational conditions. Data can be used to detect unexpected patterns. The combination of data and knowledge is potential to improve the performance. Specifically, knowledge is utilized to construct the adjacency matrix to represent the spatial structure of ACS. Then, a deep learning model is constructed by integrating the 1D-CNN and GAT, which is used to capture the temporal–spatial features of ACS. The experimental results on ACS dataset show that the accuracy is more than 99% with low computational cost. [ABSTRACT FROM AUTHOR]

Developing advanced population-based multi-objective optimization algorithms to explore optimal operating parameters is a crucial means of achieving energy saving and consumption reduction in aluminum electrolysis process (AEP). Many techniques, such as NSGA-II/NSGA-III, rely on dominance relationships among population members to prioritize evolving individuals. However, this mechanism assumes equal relationships among selected candidates, overlooking the potential of preponderant individuals to guide rapid population evolution. To tackle this issue, we propose a multi-objective optimization algorithm driven by preponderant individuals and symmetric sampling (MOPISS). This algorithm first establishes an individual selection strategy driven by preponderant individuals to expedite population evolution. Subsequently, it employs the count of preponderant individuals to calculate the interaction learning probability (ILP) for adaptively adjusting the genetic intensity of population. ILP is then integrated into the evolutionary process to create a novel genetic mechanism based on symmetric sampling, thereby enhancing the randomness of the solution space and genetic efficacy. Finally, MOPISS is validated across 17 benchmark functions and operational parameter optimization cases of AEP. The results indicate that the proposed method exhibits a significant competitive advantage over other mainstream approaches in terms of evolutionary speed and optimization performance. Specifically, it can increase current efficiency by approximately 3% and reduce energy consumption by 276 kW ⋅ h/t-Al for AEP. [ABSTRACT FROM AUTHOR]

Semi-dry desulfurization is an efficient means of SO 2 removal from the effluent gases from electrolysis cells in aluminum smelters. These gases are at low temperature and contain low concentrations of SO 2 , as opposed to thermal power plants. The removal is carried out by injecting powdered alkaline sorbent, hydrated lime (solid particles), into the SO 2 -containing gas (gas phase) in the presence of humidity. The reaction is controlled by the adsorption of SO 2 onto the surface of lime. This study involves the mathematical modelling of a lab-scale scrubber using a Lagrangian-Eulerian approach in order to analyze the desulfurization efficiency. The model was validated based on experimental data. A parametric study was carried out to investigate the effects of particle size, sorbent amount, and relative humidity (RH) on the desulfurization efficiency. The results show that the particle size is the most important parameter; as the particle size decreases, the desulfurization efficiency increases. However, using finer particles may increase the process cost. The loss in SO 2 capture efficiency due to the use of coarser particle size could be compensated by increasing the relative humidity (RH) of the gas, another key parameter of the process. [Display omitted] • A new unit is needed after the GTC to separate SO 2 from the electrolysis gas. • The humidity enhances the desulfurization efficiency. • The sorbent size is slightly more effective than relative humidity in SO 2 removal. • The particle phase effect on continuous phase is determined by two-way coupling with DPM. • Increasing humidity can compensate the negative effects of increasing particle size. [ABSTRACT FROM AUTHOR]

Unknown abnormal working condition discovery is the key of refinement industrial production. Clustering industrial time series is an effective way to discover unknown working condition types. However, it is challenge for existing time series cluster methods to discover unknown abnormal working condition from industrial time series. In this study, a novel prior knowledge-augmented unsupervised shapelet learning method is proposed to discover abnormal and meaningful working condition through interpretable subsequences. A prior feature extracting module is proposed to change prior knowledge into a recognizable form for the data model. The prior knowledge contains abnormal working condition information. The knowledge-augmented clustering module can learn informative shapelets which stand for abnormal working condition by combining prior features with data features. Furthermore, the preference of prior knowledge and data are self-adjusted in the learning phase. Numerical test results on the real-world aluminum electrolysis process, simulated Tennessee Eastman process, and continuous stirred tank heater process verify the superior performances of the proposed method. The proposed method provides a new perspective for the fusion of prior knowledge and data model. It also provides a new way to solve the problem of abnormal unknown working condition discovery in industrial process. [ABSTRACT FROM AUTHOR]

Superheat degree identification of aluminum electrolysis cell is a typical knowledge-intensive work which is the precondition of efficient production. However, existing methods have application limitations, and the large cognitive differences among experts are ignored in the existing methods. To address these issues, we propose the consensus-based probabilistic hesitant intuitionistic linguistic Petri nets (CPHILPNs). The truth values of places in CPHILPNs are represented by probabilistic hesitant intuitionistic linguistic term sets (PHILTSs), which can enrich the experts' expression and improve the efficiency of dealing with uncertainties. The proposed probabilistic hesitant intuitionistic linguistic consensus optimization model can reduce the deviation of individual preferences to reach a consensus. Then the obtained experts' weights and decision assessments will be more accurate and comprehensive with objective evaluations. The order weighted averaging PHILTSs concurrent reasoning algorithm is proposed to enhance the inference efficiency. Finally, the usefulness and validity of the proposed CPHILPNs are verified by conducting comparisons and actual experiments in a real-world plant. [ABSTRACT FROM AUTHOR]

• Anodic reactions on alloys was in-situ characterized by linear sweep voltammetry. • All of the Fe-Ni alloys can be passivated by oxygen except nickel and 20Fe-80Ni. • A three-layer passivation film was inferred to be formed on the Fe-rich alloys. • NiFe 2 O 4 in the passivation film greatly promoted the corrosion resistance. The corrosion behavior of Fe-Ni alloys in molten 45KF-50AlF 3 -5Al 2 O 3 (wt.%) at 700 °C was characterized in this study. After oxygen evolution, all of the Fe-Ni alloys formed a passivation film consisting of an outer Fe 2 O 3 layer, an intermediate Fe 3 O 4 layer, and an inner FeO layer. Because of the outward diffusions of nickel ions, a NiO layer formed in the passivation film when the 43Fe-57Ni alloy was passivated by O 2. A solid-state reaction between Fe 2 O 3 and NiO led to the formation of NiFe 2 O 4 in the passivation film, which greatly promoted its corrosion resistance to the molten fluoride salts. [ABSTRACT FROM AUTHOR]

Highlights • Geometries of both anode cover material (ACM) and anode crust are improved. • The melting of anode crust is included in the heat transfer modeling. • Thermal conductivity and emissivity of both ACM and anode crust are evaluated. • Increasing the ACM thickness reduces the top heat losses and increases the side ones. • Sludge increases resistive heating, enlarges the cavity and reduces the side ledge. Abstract A thermal-electric model is developed using finite elements and thoroughly validated against thermal measurements performed in industrial aluminum electrolysis cells (AEC). Knowledge about the geometries of anode cover material (ACM) and anode crust is improved by measuring their profiles in industrial cells. Moreover, the shape of the cavity formed by the melting of the anode crust is predicted with the numerical model, using a radiosity module combined to an iterative method. The thermal conductivity and the emissivity of both ACM and anode crust are also evaluated based on experimental measurements. The thermal-electric model accurately predicts the measurements obtained from heat flux sensors and thermocouples installed on industrial anodes. Modeling results show that increasing the ACM thickness reduces the top heat losses and increases the heat dissipation from the side, while the bottom losses remain constant. With thicker top insulation, the side ledge shrinks and the anode crust melts. The impact of a film of sludge under the liquid aluminum is quantified with the model. Accordingly, the sludge increases the cathode voltage drop (CVD), enlarges the cavity, reduces the side ledge thickness and amplifies the side heat dissipation. The thermal-electric modeling provides insights to improve the design and operation of AEC in order to reach higher efficiency. [ABSTRACT FROM AUTHOR]

Highlights • Ancillary service market participation posed as a mixed-integer nonlinear program. • Regression model for forecasting of upcoming maximum profit on balancing markets. • Optimal bidding strategy for an aluminum electrolysis at primary balancing market. • Savings compared to pure exploitation of spot market price spreads enabled. Graphical abstract Abstract Energy-intense enterprises that flexibilize their electricity consumption can market this either at electricity spot markets or by offering ancillary services on demand, such as balancing power. We formulate optimization of the balancing power bidding strategy as a mixed-integer nonlinear program considering both price forecasts for the ancillary service market and hourly varying spot market prices. We solve this two-stage approach by decomposition into a nonlinear bidding problem and a mixed-integer linear scheduling problem. We consider aluminum electrolysis participating in the German primary balancing market. We show savings in weekly production costs of 5–20% compared to stationary operation. The savings due to the optimal bidding strategy are up to twice the savings from pure exploitation of electricity spot market price spreads. We thus demonstrate that energy-intense processes can systematically take advantage of highly profitable demand-side management measures beyond a spot market price adjusted production. [ABSTRACT FROM AUTHOR]

Highlights • A new method is developed to measure the thermal diffusivity of solid-liquid salts. • Thermal conductivity of anode crust decreases with temperature up to 950 °C. • The crystalline phases of anode crust are quantified with the Rietveld method. • Anode crust shows concentration gradients of Al 2 O 3 , Na 3 AlF 6 and Na 2 Ca 3 Al 2 F 14. • A cavity grows from the melting and collapsing of industrial anode crust. Abstract The anode cover material (ACM) of aluminum electrolysis cells reacts with the liquid bath and vapor penetrating inside this powdered material to form a denser material, which is called the anode crust. The thermal property of the anode crust affects the top heat losses of aluminum electrolysis cells. The operation and design of electrolysis cells can be controlled through a better understanding of the thermochemical evolution of the anode crust. The typical anode crust is found either in the solid state up to 696 °C or in the solid-liquid state at higher temperatures. Its thermal conductivity is determined from 100 °C up to 950 °C on the samples extracted from an industrial electrolysis cell. For example, the thermal conductivity of the lower part of the anode crust decreases from 1.38 W m−1 °C−1 at 100 °C to 0.89 W m−1 °C−1 at 950 °C. The specific heat capacity, enthalpy, bulk density and thermal diffusivity are also determined from experimental measurements. The crystalline phases composing the anode crust are quantified with a Rietveld method performed on X-ray diffraction patterns. The thermochemical behavior of the industrial anode crust is investigated with differential scanning calorimetry and thermodynamic equilibrium calculations. The Na 5 Al 3 F 14 , α-Al 2 O 3 and Na 3 AlF 6 are the main phases composing the anode crust. The chemical analysis also demonstrates a significant amount of amorphous alumina and a low amount of calcium cryolite compounds. From a temperature of 717 °C, the α-Al 2 O 3(s) , Na 3 AlF 6(s) solution and liquid bath solution compose the anode crust according to the thermodynamic equilibrium results. Furthermore, the thermal property of each sample is given with the corresponding chemical composition. [ABSTRACT FROM AUTHOR]

In process industry, it is a habitual phenomenon that the production efficiency is improved via replacing the obsolete devices with the advanced ones. To achieve an optimal performance, these devices are always required for heavily empirical adjustments, which is very time-consuming and inefficient. Though outdated, invaluable experiences for adjusting devices towards a more efficient industrial production have been accumulated by those replaced facilities. Thereby, new devices are expected to adapt to perform the industrial task without many modulations if the aforementioned experiences can be appropriately utilized. Inspired by the fact that evolutionary multitasking is capable of exploiting latent similarities and commonalities among multiple optimization tasks so as to improve the overall convergence of multi-task optimization, in this paper, we propose a novel framework to automatically search for the optimal settings for new devices based on the knowledge accumulated by the old. The framework, dubbed Piggybacking on Past Problem for Faster Optimization (PPPFO), is able to piggyback on the past optimization problem for a faster convergence of the targeted. By means of automatically transferring search experiences (i.e. genetic and cultural characteristics) from source task to the target, PPPFO can assist an engineering optimizer to improve its search exercises. PPPFO has been tested with a number of widely used benchmark functions and has been successfully adopted to an important real application, i.e., aluminum electrolysis process design. The remarkable results verify the efficacy and efficiency of the proposed framework. • An efficient model for device renewal is designed for utilizing existed experiences. • A random-keys-inspired mapping link is proposed to convert the task domains. • A novel framework of optimization dubbed PPPFO is presented. • PPPFO is verified by benchmark functions and aluminum electrolysis process design. • The results show PPPFO can improve the overall performance of optimization problems. [ABSTRACT FROM AUTHOR]

Modeling complex physical processes such as the extraction of aluminum is mainly done using pure physics-based models derived from first principles. However, the accuracy of these models can often suffer due to a partial understanding of the process, uncertainty in the input parameters, and numerous modeling assumptions. More recently, with the ever-increasing availability of data, there has been an explosion of interest in applying modern machine learning methods because of their ability to learn complex mappings directly from data. Unfortunately, these models tend to be black boxes, require an enormous amount of data, and do not utilize existing domain knowledge. In this work, we develop a novel approach combining physics-based and data-driven modeling approaches while eliminating some weaknesses. We use a data-driven model to correct a misspecified physics-based model of the Hall–Héroult process in an aluminum electrolysis cell using a corrective source term added to the set of governing ordinary differential equations. Our approach ensures that the existing knowledge is utilized to the maximum extent possible while relying on the data-driven models only to model those aspects which the physics-based model does not represent well. We compare this approach with an end-to-end learning approach and an ablated physics-based model, showing that the proposed hybrid method is more accurate, consistent, and stable for long-term predictions. [ABSTRACT FROM AUTHOR]

Individual transport numbers provide a measure of the charge ratio which is carried by each ionic species of an electrolyte. However, unlike the electrical conductivity or diffusion coefficients, this quantity lacks a clear definition from the statistical mechanics point of view. In practice, it is measured via complex experimental setups, and most of the interpretation of the available data is made by using the Nernst–Einstein approximation. Here we show that this approach is not suitable in molten salts due to the large contributions of the cross-terms due to the ionic interactions in the total ionic conductivity. We propose a partition scheme that allows to attribute these cross-terms to the various ions, allowing for an estimate of the transport numbers in a series of molten salt formed by mixing NaF and AlF 3 at various compositions. The results are interpreted using the speciation of the melt, which is characterized by the formation of various AlF x clusters. At large AlF 3 , the Na + ions are shown to contribute almost totally to the ionic conductivity. [ABSTRACT FROM AUTHOR]

A thermodynamic evaluation and optimization of the CALPHAD type of the Cu-O-F system, together with phase diagram measurements, are considered in the present work as the first step towards a complete evaluation of Cu in the multicomponent system Cu-Fe-Ni-Na-Al-Ca-O-F. Differential Scanning Calorimetry experiments were conducted using a closed crucible to identify the phase transitions in the CuF 2 -Cu and CuF 2 -Pt pseudo-binary systems in the presence of oxide impurities. The thermograms and the diffractograms, as measured by X-Ray Diffraction, showed that reactions occurred in both systems, where the CuO impurity from the CuF 2 reactant led to the formation of Cu 2 O after three or four runs. All new model parameters of the fluoride-rich region of the Cu-O-F system were optimized based on this study’s experimental results. Measured and calculated results are in good agreement and provide a new interpretation of the phase transitions reported in the literature. [ABSTRACT FROM AUTHOR]

The industrial aluminum reduction cell is an electrochemistry reactor that operates under high temperatures and corrosive conditions. Monitoring the molten aluminum and electrolyte components is very important for controlling the chemical reaction process. Due to the lack of fast methods to monitor the components, controlling aluminum reduction cells is difficult. In this work, laser-induced breakdown spectroscopy (LIBS) was applied to aluminum electrolysis. A new method for calculating the molecular ratio, which is an important control parameter that represents the acidity of the electrolyte, was proposed. Experiments were first performed on solid electrolyte samples to test the performance of the proposed method. Using this method, the average relative standard deviation (RSD) of the molecular ratio measurement was 0.39%, and the average root mean square error (RMSE) was 0.0236. These results prove that LIBS can accurately measure the molecular ratio. Then, in situ measurements of the molten aluminum and electrolyte were performed in industrial aluminum induction cells using the developed LIBS equipment. The spectra of the molten electrolyte were successfully obtained and were consistent with the spectra of the solid electrolyte. [ABSTRACT FROM AUTHOR]

• First use of plasma alternative to solid material as an inert anode in aluminum electrolysis; • The anode product of electrolysis is oxygen; • Plasma anodes are available easily and without complicated operations; • Theoretically, the plasma anode is completely inert in the aluminum cell; The aluminum industry is one of the major causes of CO 2 emissions into the atmosphere, mainly caused by carbon anode consumption and the emission of perfluorocarbon gases. Carbon anodes are consumed by anodic reactions in the standard technology. The environmental constraints and the costs associated with the utilization of carbon anodes have, for many decades, led to the look for non-consumable anodes, called "inert anodes," considered for years to be the future of aluminum production. However, the current research on inert anodes is limited to metal, ceramic, cermet, and gas anodes and remains impotent in solving the difficulties of anode consumption, CO 2 , and other greenhouse gas emissions. Here, we propose the utilization of argon plasma as an inert anode for aluminum electrolysis. The results from emission spectroscopy analysis and Density Functional Theory (DFT) calculations depict a production of positively charged argon ion (Ar+) at the anode region, which infuses into the electrolyte and reacts electrochemically with the oxy-fluoro aluminate complexes (Al 2 OF 6)2-. For a current ≤ 0.4 A, the oxygen evolution occurs anodically from 2Al 2 OF 6 2- + 4Ar+ → 4AlF 3 + O 2 + 4Ar, with no argon consumption. Furthermore, the aluminum is reduced cathodically, and the decomposition reaction is 2Al 2 O 3 → 4Al + 3O 2. This innovative approach enables carbon-free aluminum electrolysis, a large-scale reduction in greenhouse gas (GHG) emissions, which can be extended to similar electrolytic industries. [ABSTRACT FROM AUTHOR]

This paper studies the crystal transition of graphite intercalation compounds during aluminum electrolysis. Transmission electron microscopy (TEM) analysis demonstrates that the transition induces the structural degradation of carbon cathodes. Buckling of basal planes in the hexagonal compounds contributes to the crystal transition of intercalation compounds from hexagonal to rhombohedral modification. First-principle techniques are used to calculate the structural and electronic properties of graphite intercalation compounds by density-functional theory using the plane wave Pseudo-potential method. The results have demonstrated a close relationship between lattice mismatch and intercalation compounds. The electronic band structure close to the Fermi level has been investigated. The main effect of intercalation compounds is to transfer charge from alkali metal s electron to the carbon p z orbitals, rendering the system metallic. With a decrease of band gap, the crystal transition induces an increase in the interlayer spacing and C–C bond lengths and deteriorates the structural stability of carbon cathodes. The overwhelming majority of contribution to the bands neighboring the Fermi level is from C pz orbital in both intercalation compounds. These findings provide new insights into the mechanisms of carbon cathodes deformation, damage and fracture behavior. • The crystal transition deteriorates the structural stability of carbon cathodes. • The intercalation compounds contribute to the metallic behavior of carbon cathodes. • The crystal defects facilitate the crystal transition of intercalation compounds. • The carbon p z orbitals play a role in the density of states at the Fermi energy. [ABSTRACT FROM AUTHOR]

With continuous scale expansion and higher safety requirements in modern aluminum electrolysis, it is more and more necessary to realize an intelligent decision-making for shutting-down and running of aluminum reduction cells (SRARC). However, this realization needs to solve a special single-objective constrained integer optimization problem (SCIOP), where the key challenge is a high requirement for the constraint-handling ability. In this paper, based on a high-scalability intelligent optimization algorithm called discrete state transition algorithm (DSTA), a population-based DSTA with decomposition and knowledge guidance (PDSTA/D-S) is proposed. This PDSTA/D-S improves the constraint-handling ability of DSTA from three aspects. Firstly, a hybrid framework combining DSTA with genetic algorithm is proposed. Secondly, a decomposition-based multi-objective optimization for constrained problems with uniformly-angled weighted sum vector is proposed. Thirdly, the manual decision-making is transferred to a knowledge-based transformation operator of DSTA. Therefore, a high-level performance of decision-making for SRARC can be obtained. The related experiments on a SRARC which is built from the practical production have demonstrated that the proposed PDSTA/D-S not only makes an effective improvement of DSTA from three aspects, but also has a more advanced performance compared with other existing high-performance intelligent optimization algorithms. • A large-scale and multi-constrained integer optimization problem is established for shutting-down and running of aluminum reduction cells. • A hybrid framework combining discrete state transition with genetic algorithm is proposed. • A decomposition-based multi-objective optimization for constrained problems with uniformly-angled weighted sum vector is proposed. • The manual decision-making is transferred to a knowledge-based transformation operator of discrete state transition algorithm. • The experiments on a practical production show the proposed framework can easily find a higher-quality feasible solution. [ABSTRACT FROM AUTHOR]

Deep neural networks have become very popular in modeling complex nonlinear processes due to their extraordinary ability to fit arbitrary nonlinear functions from data with minimal expert intervention. However, they are almost always overparameterized and challenging to interpret due to their internal complexity. Furthermore, the optimization process to find the learned model parameters can be unstable due to the process getting stuck in local minima. In this work, we demonstrate the value of sparse regularization techniques to significantly reduce the model complexity. We demonstrate this for the case of an aluminum extraction process, which is highly nonlinear system with many interrelated subprocesses. We trained a densely connected deep neural network to model the process and then compared the effects of sparsity promoting ℓ 1 regularization on generalizability, interpretability, and training stability. We found that the regularization significantly reduces model complexity compared to a corresponding dense neural network. We argue that this makes the model more interpretable, and show that training an ensemble of sparse neural networks with different parameter initializations often converges to similar model structures with similar learned input features. Furthermore, the empirical study shows that the resulting sparse models generalize better from small training sets than their dense counterparts. • Using sparse neural networks to model set of nonlinear ODEs representing an aluminum electrolysis cell. • Increased interpretability with sparse neural networks. • Higher generalizability for sparse neural networks than dense neural networks given small datasets. • Sparse neural networks show more stable convergence in training than their dense counterpart. [ABSTRACT FROM AUTHOR]

The working condition evaluation of the aluminum reduction cell is the basis of formulating operation strategy, ensuring production safety and realizing stable and optimized operation of the aluminum electrolysis process. In this study, we deal with the cell working condition evaluation problem from a new perspective, and an improved Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) based multi-criteria decision-making approach is proposed. This approach starts from the process data and working condition standard. In order to obtain the appropriate weights to the process indicators, Criteria Importance Through Inter-criteria Correlation (CRITIC) approach is used in the proposed approach. Then the TOPSIS is used to calculate the relative closeness of the cell working condition to the working condition standard. Finally, a crisp level determination approach is used to provide a delicate level associated with the working condition standard. Two real cases demonstrated that the proposed approach is useful and effective in the working condition evaluation of the aluminum reduction cell. [ABSTRACT FROM AUTHOR]

Highlights: [•] Cu–Ni–Fe alloy was produced at semi-pilot scale by ball milling. [•] The Cu–Ni–Fe powder was consolidated by spark plasma sintering. [•] Its nanocrystalline and monophased structure was retained during the sintering. [•] It was evaluated as inert anode for Al electrolysis in low-temperature electrolyte. [•] Aluminum with a purity of 99.4% was produced. [ABSTRACT FROM AUTHOR]

Abstract: Electrochemical intercalation of potassium into graphite in molten potassium fluoride at 1163K was investigated by means of cyclic voltammetry, galvanostatic electrolysis and open-circuit potential measurements. It was found that potassium intercalated into graphite solely between graphite layers. In addition, the intercalation compound formed in graphite bulk in molten KF was quite unstable and decomposed very fast. X-ray diffraction measurements indicate that a very dilute potassium–graphite intercalation compound was formed in graphite matrix in the fluoride melt. Analysis with scanning electron microscope and transmission electron microscope shows that graphite was exfoliated to sheets and tubes due to lattice expansion caused by intercalation of potassium in molten KF. [Copyright &y& Elsevier]

The multivariate time series (MTS) classification is one of the major tasks of time series data mining. Many methods have been proposed to investigate the MTS classification. Among them, the method based on feature representation is the most popular and widely used one. However, there exist some shortcomings for this method, such as unsatisfactory accuracy, being sensitive to noise and not able to fully make use of time series data attributes. In order to overcome these disadvantages, we propose a new method called functional deep echo state network (FDESN) for MTS classification that utilizes two special operators: temporal aggregation and spatial aggregation. In general, the parameters of the FDESN are determined by random selection, human experience or trial and error. This may increase the complexity of the FDESN or reduce the accuracy of the FDESN. In this study, a novel bi-level optimization approach is proposed to optimize the parameters of the FDESN. The parameter selection problem in the FDESN is transformed into the bi-level optimization problem. The state transition algorithm (STA) is used to solve the bi-level optimization problem. Finally, the experimental results show that the proposed method is superior to other methods. In addition, the proposed method is successfully applied to anode condition identification in aluminum electrolysis. For the aluminum electrolysis datasets, the proposed method improved the average classification accuracy by about 3.5% compared with the other methods. For a specific aluminum electrolysis dataset ACS2504, the classification accuracy significantly increased from 77.92% to 82.69% by using the proposed method. • A novel DESN method called functional deep echo state network (FDESN) is proposed for MTS classification. • A bi-level optimization model is established to optimize the parameters of the FDESN. • The STA is used to solve the bi-level optimization problem. • The proposed method is successfully applied to anode condition recognition in aluminum electrolysis. [ABSTRACT FROM AUTHOR]

An oxide film containing NiFe 2 O 4 was prepared by oxidizing Fe–Ni alloy in hot air, and the performance of the protective oxide film in molten KF–AlF 3 –Al 2 O 3 salts at 700 °C was characterized by electrochemical methods, including electrochemical impedance spectroscopy, capacitance measurements, linear sweep voltammetry, and chronoamperometry to clarify the breakdown and growth process of the oxide film. For the pre-oxidized 57Fe–43Ni, 50Fe–50Ni, and 43Fe–57Ni alloys, the oxide films, with a thickness of 5 μm, are composed of Fe 2 O 3 and NiFe 2 O 4 , according to X-ray diffraction and scanning electron microscopy analysis. The transfer resistance of ions in the oxide film increases with increasing nickel content, and reaches its maximum value of 91.39 Ω⋅cm2 for the pre-oxidized 43Fe–57Ni alloy based on electrochemical impedance spectroscopy measurements. Moreover, the carrier density in the oxide film determined from Mott–Schottky plots decreases as the nickel content increases. Under anodic polarization, the breakdown of oxide film follows the generation of metal cation vacancies and condensation of excess cation vacancies. Concurrently, a new oxide layer forms along the substrate owing to the annihilation of oxygen vacancies and diffusion of oxygen anion in the oxide film. After oxygen evolution, a new oxide film layer containing NiFe 2 O 4 forms on the alloy substrate, leading to the passivation observed in the anodic polarization curves. The maximum polarization resistance in the passivation region is 4.86 Ω⋅cm−2 for the pre-oxidized 43Fe–57Ni alloy, indicating that the pre-oxidized 43Fe–57Ni alloy possesses better corrosion resistance to molten fluorides than the other two pre-oxidized Fe–Ni alloys. Chronoamperometry measurements show that the oxide film on the 43Fe–57Ni alloy is in a dynamically stable state during oxygen evolution. • An oxide film containing NiFe 2 O 4 was prepared by oxidizing Fe–Ni alloy in hot air. • NiFe 2 O 4 greatly improved the transfer resistance of ions in the oxide film. • The anodic behaviors of the pre-oxidized Fe–Ni alloy were in situ characterized. • The breakdown and growth processes of oxide films on Fe–Ni alloys were clarified. • After oxygen evolution, a multilayered oxide film containing NiFe 2 O 4 layer formed. [ABSTRACT FROM AUTHOR]