Aluminum Electrolysis Process Research Articles

Multimode Process Monitoring and Mode Identification Based on Multiple Dictionary Learning

In modern industrial systems, the processes often operate under different modes. Potential fault diagnosis and mode identification are extremely vital to maintain the system safe and reliable. Recently, many methods have been proposed to address these problems separately. Moreover, many of them make an assumption that the data from industrial site only contain the Gaussian noise. However, this assumption is not held in practice, which further reduces their performances. Considering the complicated noise feature of industrial data, we came up with an improved dictionary learning method to settle these problems simultaneously. First, the measurement data were decomposed into three parts: clean data, mode-based noise, and dense Gaussian noise. Then, the dictionary learning method was proposed to characterize each part separately. Inspired by the framework of label-consistent K-SVD, the mode information was incorporated into the dictionary learning method, and we developed a solution to settle the multiple dictionary learning optimization problem. Finally, when new samples arrive, we reconstruct them under the learned dictionary so that each sample's mode and abnormal data and can be determined. The experiments on two different types of simulated process and aluminum electrolysis process show the strength and reliability of our method, which indicates the engineering application value of the proposed method.

Исследование МГД-стабильности алюминиевого электролизера при различных условиях проведения технологического процесса

Mathematical simulation of industrial aluminium electrolytic cell operation allows us to predict and indicate the causes of magnetohydrodynamic (MHD) instability and bath level skewing, as well as investigate other features of the aluminium electrolysis process. In order to analyse the MHD stability of the electrolytic cell, we adapted a three-dimensional mathematical model that uses a multi-phase approach to describing the media (aluminium, electrolyte and gas) and treats the hydrodynamic, electromagnetic, thermal and electrochemical processes in the bath as interrelated. Our test calculations confirmed that the model is adequate and that the numerical solution proposed converges with sufficient accuracy. The paper describes our numerical investigation results concerning MHD stability of a multi-anode electrolytic cell when its thermal conditions and working space shape configuration change; our simulation included the metal-electrolyte phase interfaces and took into account the MHD instability developing when replacing burnt-out anodes. We estimated how various initial crust configurations affect the MHD stability. We investigated how the process parameters affect the working space shape in the bath, which is a dynamic object, same as the metal-electrolyte interface and the reverse oxidation zone surface. We specifically studied the way changes in potential affect the MHD stable shape of the working space in the bath. We show that varying the potential between any given pair of anodes can change the shape of the working space, that is, crust melts as potential increases, while lowering potential leads to further accretion. As this happens, we note that there is an increase in the vibration magnitudes of the liquid metal and the lower reverse oxidation zone boundary, but these variations are still within the range acceptable in terms of MHD stability of the electrolysis process

Properties of Bio-pitch and Its Wettability on Coke

Bio-pitch, driven from biomass, is a potential green alternative of coal-tar-pitch in the production of carbon anodes for the aluminum electrolysis process. Information on the wetting capacity of b...

A geometry constrained dictionary learning method for industrial process monitoring

Data-driven process monitoring methods have attracted great attention due to the case that it can provide an efficient way to cope with the industrial process without the need of first-principle models. The local information that results from the reaction process often influences the process monitoring result. Unfortunately, this local information is usually ignored in the data-driven process monitoring method. In this paper, a geometry constrained dictionary learning (GCDL) method is proposed to address the above problem. By exploiting the underlying characteristics and simultaneously holding the local geometrical information, the GCDL method leads to a balance between the reconstructive item and the discriminative item. Inspired by the manifold method, the GCDL method can provide discriminative sparse coding, which means that samples that belong to the same class will share similar sparse coding. It is in accordance with the fact that the samples of different categories often reside on different subspaces embedded in a high-dimensional feature space. The effectiveness of the proposed method is evaluated through two numerical simulation cases, a continuous stirred tank reactor (CSTR) case and a real industrial aluminum electrolysis process. Extensive experimental results reveal that the proposed method shows better performance by taking the local geometrical information into consideration.

An improved cell situation identification approach with convolutional neural network and wavelet extreme learning machine

In the aluminum reduction process, the flame hole is an influential index of the whole production situation, which reflects the distribution of the physical field of the reduction cell, the current efficiency and the lifespan of the aluminum reduction cell. Therefore, flame hole situation detection and identification are critical and significant in the whole process of aluminum electrolysis. However, in the practical industrial production, flame hole identification result is coming from the experimental operation workers, with the loss of workers and different experimental levels, the real-time measurement of the flame hole index is still a challenge beyond solution. This article develops a flame hole image classification method based on wavelet extreme learning machine. First, a deep feature set is extracted from the original images with a convolutional neural network. Then, a classification model based on an extreme learning machine with wavelet activation function is developed. Finally, the proposed convolutional neural network–weighted extreme learning machine model is applied to superheat degree real-time detection in the industrial electrolysis cell. The proposed method is evaluated on aluminum production which outperforms existing other superheat degree methods in accuracy and robustness.

A mechanism knowledge-driven method for identifying the pseudo dissolution hysteresis coefficient in the industrial aluminium electrolysis process

To overcome the difficulties that are associated with the online recognition of the alumina dissolution properties in industrial aluminium electrolysis cells, this paper proposes a method driven by process mechanism knowledge for the identification of a pseudo dissolution hysteresis coefficient (PDHC). The method explicitly represents the process semantemes that are implied in the normalized cell voltage (NCV) and the feed state using the proposed PDHC via process mechanism analysis and process semantic embedding. The PDHC quantifies the online dissolution performance of alumina and, thus, can overcome the inability to express the alumina online dissolution performance in industrial cells. Compared with slope-based methods, the PDHC-based method can not only realize the online recognition of the bath temperature but also detect an abnormal alumina concentration with a lead time of a few integrated feed periods (IFPs), thereby providing a new online basis for the temperature control and feed control of industrial cells. The PDHC identification is an application case of automatic knowledge acquisition in industrial aluminium electrolysis production.

Sparse stacked autoencoder network for complex system monitoring with industrial applications

• A fault classification and isolation method based on SAE is proposed for complex system. • Deep learning model formed by AE with sparse and regular constraints can well extract features. • The fault isolation is conducted by calculating the contribution rate of each variable and choosing the largest one. • High fault classification and isolation accuracy of the proposed method is validated by some typical processes. Modern industrial systems are increasingly complex with the development of technology and equipment. Due to the variation of process loads and other factors, many processes often switch their operating modes, which results in the problem of multimode process monitoring. In addition, owing to the harsh industrial conditions, monitoring of multimode processes is difficult and challenging. In this paper, a fault classification and isolation method were proposed based on sparse stacked autoencoder network. In detail, a single autoencoder is trained one by one in an unsupervised way. Then, the hidden layer of each trained autoencoder is cascade connected to form a deep structure. Finally, the stacked autoencoder network is followed by a Softmax layer to realize the fault classification task. Since the deep structure can well learn and fit the nonlinear relationship in the process and perform feature extraction more effectively compare with other traditional methods, it can classify the faults accurately. After an in-depth exploration of the deep neural network , a novel fault isolation scheme is proposed. By calculating the contribution rate of each variable, the one with the largest contribution was identified as the fault location. In order to evaluate the efficiency of the proposed method, some numerical experiments were conducted on the continuous stirred tank heater (CSTH) benchmark process as well as two industrial cases including the wind turbines monitoring and the aluminum electrolysis process monitoring in comparison with several traditional methods. The experimental results show that the accuracy of the proposed method in the fault classification process is better than other methods and the fault isolation scheme can well locate the variable that caused the fault.

The combined algorithm of aluminum-electrolytic cell voltage collection based on Kalman filter

This paper proposes a combined cell voltage acquisition algorithm based on Kalman filtering, which is used to solve the problem of insufficient cell voltage acquisition and high delay in the current aluminum electrolysis process. This algorithm is based on Kalman filtering, and uses the mean square error of the data to represent the Gaussian white noise power of the collected and predicted values. It has strong tracking performance in the state of stable voltage. The voltage acquisition data out of the interval is comply to first-order inertial filtering, to replace the voltage sampling value, and then using the Kalman filtering algorithm to calculate the filtering gain, to ensure that the combined algorithm can filter out the influence of noise when the cell voltage fluctuates greatly, and in the stable process of electrolysis, the convergence of battery voltage can be tracked quickly. The simulation results show that compared with the first-order inertial filter, the improved Kalman filter reduces the root mean square error by 50% when the voltage is stable, and reduces the convergence time by 50% after the electrolytic cell regenerates needle vibration and swing.

Structure Dictionary Learning-Based Multimode Process Monitoring and its Application to Aluminum Electrolysis Process

Most industrial systems frequently switch their operation modes due to various factors, such as the changing of raw materials, static parameter setpoints, and market demands. To guarantee stable and reliable operation of complex industrial processes under different operation modes, the monitoring strategy has to adapt different operation modes. In addition, different operation modes usually have some common patterns. To address these needs, this article proposes a structure dictionary learning-based method for multimode process monitoring. In order to validate the proposed approach, extensive experiments were conducted on a numerical simulation case, a continuous stirred tank heater (CSTH) process, and an industrial aluminum electrolysis process, in comparison with several state-of-the-art methods. The results show that the proposed method performs better than other conventional methods. Compared with conventional methods, the proposed approach overcomes the assumption that each operation mode of industrial processes should be modeled separately. Therefore, it can effectively detect faulty states. It is worth to mention that the proposed method can not only detect the faulty of the data but also classify the modes of normal data to obtain the operation conditions so as to adopt an appropriate control strategy. Note to Practitioners —Motivated by the fact that the industrial process often has different modes and they may have common patterns, this article proposes a structure dictionary learning method for multimode process monitoring. First, the structure dictionary learning method was proposed to extract the common pattern and mode-specific pattern of each mode. After two different patterns are extracted, the control limit for process monitoring can be obtained from the training data. When new data arrive, the monitoring process can be carried out. Intensive experimental results show that the proposed method performs better than other conventional methods. Compared to conventional methods, the proposed approach overcomes the assumption that each operation mode of the industrial process should be modeled separately. Therefore, it can effectively detect faulty states. Above all, it is suitable for monitoring of real industrial systems.

Distributed dictionary learning for high-dimensional process monitoring

In order to conduct efficient process monitoring of modern industrial system featured with complexity, distributed and high-dimensional, a distributed dictionary learning is proposed for fault detection and fault isolation task. Firstly, it can reduce the computational complexity by decomposing the whole high-dimensional industrial system into several low-dimensional modules, and some prior process knowledge is integrated into the data-driven model to ensure the reliability during the decomposition stage. Secondly, since the small failure is easy to hide in high-dimensional data, it is more conducive to detecting the process data by using the sub-modules. Based on this, a Bayesian inference method is presented to fuse the distributed results for global industrial process monitoring. For the fault samples which have been detected successfully, a count time based method is introduced to determine the fault location on the block level. Then, a sparse contribution plot method is used to locate the failure source of the system on the variable level further. In the end, the performance of the proposed method is verified on a numerical simulation case, the Tennessee Eastman (TE) benchmark and an aluminum electrolysis process.

Simulation of the Aluminum Electrolysis Process in a High-Current Electrolytic Cell in Modern Software

A three-dimensional nonstationary model of a 600 kA electrolytic cell is described. The mathematical model describes the electrolysis processes occurring in the cell. The assumptions underlying the model are justified. The dependence of the process on raw materials is shown and taken into account in the model.

Mechanism of aluminum carbide formation in aluminum electrolysis cells

The formation and dissolution of aluminum carbide is considered the primary factor affecting the life of aluminum electrolysis cells. Herein, the characteristics of sodium-graphite intercalation compounds (Na-GICs) were measured and the formation mechanism of Al4C3 during the aluminum electrolysis process was experimentally studied. The Na-GIC characteristics and the products of aluminum and Na-GIC reactions were investigated by Raman spectroscopy, X-ray diffraction, X-ray photoelectron spectroscopy, and scanning electron microscopy. The results showed that graphite can react with the sodium metal to form Na-GICs, which were detectable by Raman spectroscopy. Sodium inserted into the graphite layered structure acted as an intercalation agent to change the original graphite layered structure and increase the volume and specific surface area of graphite. Further, Al4C3 was produced by using sodium-graphite intercalation compounds and aluminum as materials. Thus, the presence of sodium plays an important role in the formation process of Al4C3 in aluminum electrolysis cells.

Surface characterisation and wettability of titanium diboride by aluminium at low temperature

ABSTRACTTitanium diboride (TiB2) represents a perspective material for innovations in aluminium technology. This refractory ceramic material has superior thermal and electrical conductivity and considerable chemical stability at high temperatures (>1000°C). Furthermore, excellent wettability by molten aluminium at operating temperatures is required for the application of TiB2 as an inert cathode in aluminium electrolysis process. The wettability of various TiB2-based materials has been studied in literature at different conditions in temperature ranges above to 1000°C. The present paper surveys the current state of knowledge on aluminium wettability of TiB2 at high temperatures and examines the wettability of pure hot-press sintered TiB2 material with different values of surface roughness at low temperatures, i.e. below 900°C.

Thermo-Chemo-Poromechanical Modeling of the Anode Mixture During the Baking Process: Constitutive Laws and Governing Equations

Abstract Aluminum is reduced from alumina by the Hall–Héroult electrolysis process in which the anode is utilized as the positive electrode. The quality of the prebaked anode plays a crucial rule in the efficiency of the aluminum electrolysis process. To produce high-quality anodes in the aluminum industry, the anode baking process calls for a deep understanding of mechanisms that govern the evolution of the anode mixture properties under the high-temperature condition. Therefore, the aim of this paper is to establish a thermo-chemo-poromechanical model for the baking anode by using the theory of reactive porous media based on the theory of mixtures within the thermodynamic framework. For this purpose, an internal state variable called “shrinking index” is defined to characterize the chemical progress of the pitch pyrolysis in the anode skeleton, and the Clausius–Duhem inequality is developed according to the Lagrangian formalism. By introducing a reduced Green–Lagrange strain tensor, a Lagrangian free energy is formulated to found a set of state equations. Then, the thermodynamic dissipation for this pyrolyzing solid–gas mixture is derived, and a constitutive model linking the chemical pyrolysis with the mechanical behavior is achieved. A dissipation potential is consistently defined to ensure the non-negativeness of the thermodynamic dissipation and to obtain the constitutive laws for viscous behaviors. Field equations governing the volatile diffusion and the heat transfer through the draining mixture body are derived from the entropy balance.

A modified neighborhood mutual information and light gradient boosting machine-based long-term prediction approach for anode effect

The anode effect (AE) often occurs in the aluminum electrolysis process, which seriously affects production efficiency and causes large energy consumption. Therefore, predicting AE in advance has become very important. However, traditional cell resistance-based methods have the disadvantages of low accuracy and short predicting time, and the data-driven methods have not fully considered the relevance between other features and AE, causing the low accuracy for long-term prediction. In this paper, a long-term prediction approach for AE based on a modified neighborhood mutual information and light gradient boosting machine (MNMI-LGBM) is proposed. The MNMI is used to measure the relevance between features and AE for feature selection, and a derivative feature of the cell resistance is fed into the LGBM as a key feature to predict AE. The predicting time in the range of 0–50 min is considered. The results show that the proposed approach has an accuracy above 89% over the entire 0–50 min predicting time range. Specifically, the accuracy achieves 93.6% when predicting time is 50 min. The proposed approach only requires six features as the input of LGBM and can be applied for real-time AE prediction.

Nonlinear process monitoring using kernel dictionary learning with application to aluminum electrolysis process

In practice, because of complex mechanism processes, such as heating process, volume heterogeneity, and various chemical reaction characteristics, there is a nonlinear relationship among variables in industrial systems. The nonlinearity brings some difficulties to process monitoring. In order to ensure that the process monitoring system can work normally in nonlinear production processes, the nonlinear relationship between variables ought to be considered. In this work, a new fault detection and isolation method based on kernel dictionary learning is presented. In detail, the linearly inseparable data is mapped to a high-dimensional space. Then, a new nonlinear dictionary learning method based on kernel method was proposed to learn the dictionary. After obtaining the dictionary, the control limit can be calculated from the training data according to the kernel density estimation (KDE) method. When new data arrive, they can be represented by the well-learned dictionary, and the kernel reconstruction error can be used as a classifier for process monitoring. As for the fault data, the iterative reconstruction based method is proposed for fault isolation. In order to evaluate the effectiveness of the proposed process monitoring method, some extensive experiments on a numerical simulation, the continuous stirred tank heater (CSTH) process, and a real industrial aluminum electrolysis process are conducted. The proposed method is compared with several state-of-the-art process monitoring methods and the experimental results show that the proposed method can provide satisfactory monitoring results, especially for some small faults, thus it is suitable for process monitoring of nonlinear industrial processes.

Multiscale Simulation of Bubble Behavior in Aluminum Reduction Cell Using a Combined Discrete-Bubble-Model–Volume-of-Fluid–Magnetohydrodynamical Method

The physics of aluminum electrolysis process involves many spatial scales, generating a wide variety of bubbles of different sizes in the magnetohydrodynamical (MHD) flow. To capture the dynamics o...

A Hybrid Regularization Semi-Supervised Extreme Learning Machine Method and Its Application

Semi-supervised extreme learning machine (SS-ELM) has been applied to many classification and regression assignments with high performance, in which both the labeled and unlabeled data are exploited to enhance accuracy and computation efficiency. The Laplacian manifold regularization method has been incorporated to explore the geometry of the underlying manifold structure. However, the Laplacian manifold regularization lacks the extrapolating ability and biases the solution to a real constant function. In this paper, we propose a novel algorithm, the Laplacian-Hessian regularization SS-ELM (LHRSS-ELM), to enhance the performance of conventional SS-ELM. The main advantages of LHRSS-ELM are as follows: 1) LHRSS-ELM exhibits the learning capability and computational efficiency of traditional SS-ELMs; 2) LHRSS-ELM algorithm combines both Laplacian and Hessian term to enhance the extrapolating power, accuracy, and robustness and also show significant performance in multiclass classification tasks; and 3) for the purpose of pursuing the best pair of hyperparameters to establish a comparable model, we dynamically update them from sequences. The proposed algorithm is evaluated on publicly available data sets and further applied for the state classification of superheating degree in the aluminum electrolysis process. The experimental results demonstrate that the proposed mechanism is superior to the existing state-of-the-art semi-supervised learning algorithms in the matter of accuracy and robustness.

Multimode process monitoring based on robust dictionary learning with application to aluminium electrolysis process

In modern process industries, many parameters or states can be acquired with sensors, and these parameters or states often have a close relationship with operation conditions. Unfortunately, the process often operates under different modes, and labels thereof are often unknown. In practice, labeling for sampled data is expensive and time-consuming, so identifying the operation conditions of the industrial process is difficult. In addition, sampled data from the industrial system are always contaminated by outliers or noise. Therefore, a robust process monitoring method for the multimode process is particularly important and challenging. In this paper, a robust dictionary learning method is proposed for processes with multiple unknown modes. Firstly, by taking the sparsity of outliers into account, a robust dictionary learning method is proposed to identify and remove the outliers and noise in the sampled training data. Secondly, an iterative minimization algorithm is designed for solving the dictionary learning optimization program. Thirdly, based on the sparsity of the sparse code, we partition the sparse code into different clusters via spectral clustering method, and then the dictionary is divided into some sub-dictionaries according to the cluster results of sparse code. Lastly, when a new sample is generated, we reconstruct it under different sub-dictionaries, and the smallest dictionary reconstruction error is calculated as a classifier for process monitoring and fault detection. To evaluate the validity and effectiveness of the proposed monitoring approach, we conduct extensive experiments on a numerical simulation, the continuous stirred tank heater (CSTH) process, and an industrial aluminum electrolysis process, in comparison with several state-of-the-art methods. The experimental results demonstrate that the proposed method is able to provide satisfying monitoring results, and it is also robust to outliers in the sampled training data. It is worth mentioning that the proposed method is an unsupervised learning method, therefore, it is more suitable for the process monitoring of real industrial systems.

Anode effect prediction based on a singular value thresholding and extreme gradient boosting approach

An anode effect often occurs during the process of aluminum electrolysis that will cause large energy consumption and low efficiency in aluminum production, thus how to identify the anode effect in advance has become an important issue. However, traditional approaches ignore the common incomplete information problem existing in the acquired data, and only consider a single predicting time, resulting in an unreliable result in anode effect prediction. In this paper, a hybrid prediction approach based on a singular value thresholding and extreme gradient boosting (SVT-XGBoost) approach is proposed to identify the anode effect in the aluminum electrolysis process. The SVT is used for data filling by the whole-features transformation, and the XGBoost is utilized for classification of the anode effect. The predicting time is set to 10 min by the comparison. The experimental results show that the proposed approach has an effective ability for anode effect classification using the SVT-XGBoost compared to the previous approaches. Here, the effect of the training sample number is also investigated. The proposed approach could be applied in real-time anode effect prediction in the future.