Search

In the aluminum electrolysis, the amount of carbon consumed per unit of metal produced is always higher than the theoretical one. Raw material variability, anode properties, and reduction cell operation and performance can influence this discrepancy. Linking these factors to each individual anode could help close this gap thus improving the environmental footprint and reducing production cost. In today’s industrial environment, large volume of data, both structured and unstructured, are collected, otherwise known as big data. These capabilities could enable predicting the consumption for each anode as opposed to estimating it at the plant level as performed so far. This can be achieved using a tracking system to link the pot operation to individual baked and green anode properties. The performance of R&D Carbon Net Carbon Consumption formula was compared to a multivariate statistical approach that considers process data variables gathered from the aforementioned, linked, databases. Given the process data at hand, the prediction power of the net carbon consumption for individual anodes is significantly improved as opposed to using the existing model.

A machine vision sensor was developed for predicting deviations from the optimum amount of pitch in anode formulations using paste texture analysis. It could help operators mitigate the impact of the increasing variability of anode raw materials (coke and pitch). Paste samples were formulated in the laboratory using dry aggregate mixes obtained using two cokes having different properties and various amounts of pitch. These were imaged, formed into small cylindrical anodes, and baked to measure their density. A combination of image texture methods was used for extracting relevant paste textural features. The latter were then used as inputs of partial least squares regression models to predict deviations from the maximum baked density. Good prediction results were obtained. Furthermore, the sensor was able to detect when the paste was at the optimal amount of pitch for both cokes and to measure deviations from it.

The hundreds of aluminum reduction cells within an aluminum smelter are typically operated in an individual fashion. It is known that reduction cells from a single potline can have different behavior and performance depending on their different lining design, historical performance and/or setpoints for bath chemistry, temperature and electrical resistance. The pot operational state is generally characterized by a limited set of pseudo real time indicators and different variables, such as bath chemistry or temperature, manually sampled at low frequency, where consecutive samples can be separated by a few days. Smelting operators have designed process control systems and reaction procedures to cope with process variations, while not always acknowledging the cyclic variations characterizing the pot state. This paper sheds some light on cyclic operation parameters variations and how it can impact negatively control actions taken by the control system or the pot operators.

This paper reports on the application of an acousto-ultrasonic (AU) scheme for the inspection of industrial-size carbon anode blocks used in the production of primary aluminium by the Hall-Héroult process. A frequency-modulated wave is used to excite the anode blocks at multiple points. The collected attenuated AU signals are decomposed using the Discrete Wavelet Transform (DTW) after which vectors of features are calculated. Principal Component Analysis (PCA) is utilized to cluster the AU responses of the anodes. The approach allows locating cracks in the blocks and the AU features were found sensitive to crack severity. The results are validated using images collected after cutting some anodes.

Potrooms are elongated buildings in which hundreds of aluminum smelting pots are housed, and have the particularity, among others, to be naturally ventilated. A CFD model simulating a potroom is de...

Anodic incidents occur when an aluminum electrolysis cell short circuits at an anode position. They reduce the current efficiency of the affected cell, which makes it economically advantageous to detect and remove them as early as possible. This study evaluates the performance of three anodic incident detection systems of varying complexity. Individual anode current signals from 6 different cells were monitored with a Principal Component Analysis (PCA) model to detect anodic incidents using the Squared Prediction Error (SPE) statistic. The results suggest that a simple detection system using a common PCA model and a common detection threshold for all cells may be preferable. A novel approach to estimating the economic benefit of a detection system was also developed and used to optimize the systems’ detection threshold. The anodic incident detection systems were found to deliver positive profits relative to the standard detection technique.

A rapid and nondestructive machine vision sensor was developed for predicting the vibrated bulk density (VBD) of petroleum coke particles based on image texture analysis. It could be used for making corrective adjustments to a paste plant operation to reduce green anode variability (e.g., changes in binder demand). Wavelet texture analysis (WTA) and gray level co-occurrence matrix (GLCM) algorithms were used jointly for extracting the surface textural features of coke aggregates from images. These were correlated with the VBD using partial least-squares (PLS) regression. Coke samples of several sizes and from different sources were used to test the sensor. Variations in the coke surface texture introduced by coke size and source allowed for making good predictions of the VBD of individual coke samples and mixtures of them (blends involving two sources and different sizes). Promising results were also obtained for coke blends collected from an industrial-baked carbon anode manufacturer.

Anodic incidents occur when an aluminum electrolysis cell short circuits at a specific anode position as a result of a spike or other deformation forming on the bottom surface of the anode. These undesirable process events reduce the current efficiency of the affected cell, which makes early detection critical for minimizing losses. This paper describes the development of an anodic incident detection system based on monitoring individual anode current signals with a multivariate Principal Component Analysis (PCA) model. The model and monitoring system were developed and tested using 2422 individual anode current signals collected from an aluminum smelter. The results indicate that anode current signals may be useful for detecting anodic incidents in advance relative to regular detection methods based on typical process signals. Anodic incidents appear to alter the correlation structure of individual anode current signals, which could make detection possible by monitoring the Squared Prediction Error (SPE) statistic.

Reduction of pot draft in aluminum smelting cells is a promising approach for energy saving and waste heat recovery. The control of fugitive emissions (hazardous by-product gases escaping from cells to potroom) is investigated in a smelting cell with the reduced pot draft, down to half of the current level based on CFD simulations. Physical models with different length scales are created in order to directly resolve the very narrow escaping gaps on the pot shell in the physical model with smallest scale. A systematic analysis of the pot tightness is presented by considering various factors in the simulations, e.g., pot draft and hood placement. The results have shown that current pot structure, even within ideal operating conditions, fails to maintain 100% hooding efficiency under a 50% reduced pot draft. Two design modifications are proposed to enhance the pot tightness. An efficient sealing is observed when covering the lower half of the gaps between hoods. An estimation of the leaking hydrogen fluoride is made under different scenarios in order to quantitatively verify the modifications.

Aluminium industry depends today on only the coal-tar-pitch as a binder to produce carbon anodes. However, it becomes difficult to secure a high-quality binder for the future needs of the aluminum industry as coal-tar-pitch is a non-renewable fossil binder. In this work bio-pitch, as a renewable and environmentally friendly binder, was produced by the heat treatment of a bio-oil. The range of heat treatment temperature and heating rate was selected to achieve the desired physical properties of the produced bio-pitch. This bio-pitch showed a softening point and a viscosity similar to that of the coal-tar-pitch. In addition, it was able to cover and spread over the coke particles and further, penetrate into their open pores. According to the results of this work, the bio-pitch is a promising candidate to replace the coal-tar-pitch to produce a homogenous and mechanically stable green carbon anode.

Always striving to increase molten aluminum annual production, since its commissioning, Alcoa Aluminerie de Deschambault has used its booster section to test new pot designs and technologies at higher amperage. Learnings from this section are later transferred to the whole potline, which has enabled continuous amperage increase from the original 300 kA nominal capacity. However, the smelter recently set a new goal for high amperage operation, requiring major modification and extensive testing in its booster section. Higher booster capacity was required; a magnetic compensation loop was also installed, and several anode plant improvements were made. After careful planning and execution of the capital project, operation of different pot designs was demonstrated at high amperage for a five month period. This paper presents results from some of the special measurements taken during this first step of high amperage operation, as well as key pot performance indicators.

Startup fuses are used to protect cathode busbars from arcing damage when putting a cell in circuit. While an ideal fuse would divert all electrical current flowing through a given short-circuiting station when removing its last wedge, this is never the case in an operating cell. Fuses are part of a parallel network of conductors, therefore, the amount of current effectively diverted depends not only on their dimensions, but also on pot internal resistance at the preheating phase and on the interfaces electrical contact resistance. This article discusses the physics of startup fuses and the on-duty behavior of both bolted-on-bus and clamped specimens. Prototypes were designed using numerical analyses and then field-trialed at different amperages. Apart from validating numerical predictions, experimental data indicated that fusing time may be directly estimated by pot voltage readings, if sampled raw and at high enough frequency.

Aluminium reduction cells consume about 1.9 kg of alumina for each kilogram of aluminum they produce. Control of the alumina concentration within the electrolyte is of prime importance to achieve high metallurgical and energy efficiency and to reduce environmental footprint related to PFC and CO2 emissions. However, modern high amperage reduction cells, oftentimes operated at much higher amperage than their nominal design, are facing challenging conditions to ensure a homogeneous alumina concentration over time and across the pot at a given time. This has negative impact on the cell efficiency. This paper presents findings from different sampling campaigns performed at different amperage and pot condition for a given cell technology. It is shown that large gradients of alumina concentration are present over time and spatially.

This work aims to demonstrate how the use of latent variable methods can detect and diagnose the onset of an abnormal situation in aluminium reduction cells. Using recent data from Alcoa Fjardaal, Principal Component Analysis (PCA) was used to model typical variations occurring in 336 different pots. A drift in process operation was correctly identified sooner than with traditional statistical control technique. Additionally, concentration in low trace elements in the metal also corresponded to a drift in process operation. Such an early warning could help mitigate the impact of abnormal events.

The new Sequential Multi-block PLS algorithm is applied for establishing multivariate specification regions jointly on multiple types of raw materials. SMB-PLS distinguishes between process variations associated with raw materials from other orthogonal sources of variations such as operating policies. Combinations of raw material properties not compensated by the control schemes are clearly identified. Multivariate specifications are required for these combinations to avoid negative impact on process performance and product quality. The method was applied to an industrial aluminum smelter. Bad combinations of raw material properties were identified and validated against process knowledge.

Baked anodes quality control is still mostly based on core sampling and characterization. Only a small proportion of the production can be tested by this method due to the costly, time consuming, and destructive nature of the analytical techniques. Furthermore, the core properties are not necessarily representative of those of the whole block. A rapid and non-destructive method for anode quality control based on vibration modal analysis is proposed. A number of baked anodes produced under different conditions at the Alcoa Deschambault smelter were selected. These were excited mechanically at different locations and the vibration signals of the blocks were measured by accelerometers. Principal Component Analysis (PCA) was used to cluster the vibration modes of the anodes. It is shown that the proposed approach allows detecting and distinguishing different types of anode internal defects. Some of the tested anodes were cut and imaged in order to confirm the results visually.

High quality baked carbon anodes contribute to the optimal performance of aluminum reduction cells. However, the currently decreasing quality and increasing variability of anode raw materials (coke and pitch) make it challenging to manufacture the anodes with consistent overall quality. Intercepting faulty anodes (e.g., presence of cracks and pores) before they are set in reduction cells and deteriorate their performance is therefore important. This is a difficult task, even in modern and well-instrumented anode plants, because lab testing using core samples can only characterize a small proportion of the anode production due to the costly, time-consuming, and destructive nature of the analytical methods. In addition, these results are not necessarily representative of the whole anode block. The objective of this work is to develop a rapid and non-destructive method for quality control of baked anodes using acousto-ultrasonic (AU) techniques. The acoustic responses of anode samples (sliced sections) were analyzed using a combination of temporal features computed from AU signals and principal component analysis (PCA). The AU signals were found sensitive to pores and cracks and were able to discriminate the two types of defects. The results were validated qualitatively by submitting the samples to X-ray Computed Tomography (CT scan).

The baked anode quality control strategy currently used by most carbon plants based on testing anode core samples in the laboratory is inadequate for facing increased raw material variability. The low core sampling rate limited by lab capacity and the common practice of reporting averaged properties based on some anode population mask a significant amount of individual anode variability. In addition, lab results are typically available a few weeks after production and the anodes are often already set in the reduction cells preventing early remedial actions when necessary. A database approach is proposed in this work to develop a soft-sensor for predicting individual baked anode properties at the end of baking cycle. A large historical database including raw material properties, process operating parameters and anode core data was collected from a modern Alcoa plant. A multivariate latent variable PLS regression method was used for analyzing the large database and building the soft-sensor model. It is shown that the general low frequency trends in most anode physical and mechanical properties driven by raw material changes are very well captured by the model. Improvements in the data infrastructure (instrumentation, sampling frequency and location) will be necessary for predicting higher frequency variations in individual baked anode properties. This paper also demonstrates how multivariate latent variable models can be interpreted against process knowledge and used for real-time process monitoring of carbon plants, and detection of faults and abnormal operation.

A multiblock partial least squares (PLS) modeling approach is proposed in this Article for multivariate analysis and monitoring of aluminum reduction smelters and other electrochemical processes. These industries commonly operate from hundreds to a thousand of metallurgical reactors in parallel, which makes is difficult to build and maintain separate models for each unit. To cope with this problem, the proposed approach is based on the assumption that reactors sharing the same design (i.e., technology) and fed with the same lots of raw materials should also share a similar latent variable space. This allows reducing the number of latent variable models to build for process monitoring. The approach is illustrated on the basis of data collected from 31 reactors used for aluminum reduction. It was shown that well-defined regions in the raw material property and process operation spaces were associated with higher smelter performance. These could be used to establish joint multivariate specification regions for raw materials as well as for plant-wide process monitoring.

A metallurgical process generally processes some raw materials or ores into concentrates or primary metals using different technologies. However, the quality of the incoming raw materials has an effect on the process performance. Hence, engineers have to apply different set points in order to account or compensate for raw material quality variations. This is frequently performed using a set of univariate specifications. In this paper, a methodology to build multivariate specifications is presented and illustrated using a case study from an aluminum reduction smelter.

In this paper, the use of different image analysis techniques is investigated for predicting alumina content of anode cover materials used in primary aluminum smelters. This approach is proposed in to order to allow on-line estimation of alumina content for feedback control purposes, which is not currently possible due to the long time delays and limited number of samples that can be analyzed in the laboratory. Both color and textural features of various anode cover materials are extracted from digital RGB images, and partial least squares (PLS) regression models are developed for predicting alumina content from these features. Most alumina content prediction errors are within ± 2 σ of the X-ray fluorescence laboratory measurements, when larger variations are required by operators to make control decisions. Some challenges arising from the use of image analysis techniques for process control are also discussed.

A soft-sensor model developed from historical carbon plant data and multivariate statistical methods was proposed in past work to obtain quick predictions of individual anode properties right after baking for quality control purposes. It could only be used for anodes baked at the coldest and hottest positions within the furnace due to the core sampling method used at the partner plant and the way the data is collected. To complement the soft-sensor, this work proposes a strategy to account for the thermal history of anodes baked at any position. It is shown that combining categorical variables for pit and baking positions and routinely available firing equipment data is sufficient to predict the temperature history of anodes baked in different positions, which is typically not measured during normal operation. The model will be integrated to the soft-sensor in the future to account for the impact of baking position on anode properties.

The baked anode quality control scheme used in most carbon plants consists of lab testing of anode core samples and monitoring weekly averaged properties. Both the low anode sampling rate and the averaging hide a significant amount of variability in the anode populations. Additional consideration depending on the sampling procedure needs to be taken into account while analyzing the core sample properties. In previous work, a multivariate latent variable PLS model was developed for predicting individual anode properties at the end of the baking cycle. All the data available at the Alcoa Deschambault smelter were used to build the model. This work investigates how to use this model to learn from data and, in particular, to help diagnose the root cause of variations in the electrical resistivity and LC. Changes in raw material suppliers and non-uniform temperature distribution within the baking furnace were found to contribute to the drifts.

Variations in run-of-mine ore properties such as size, composition, and grindability strongly affect AG and SAG mills performance. In the past, most efforts to track and control these variations have focused on developing on-line size analysis using vision systems to reduce power consumption and avoid mill overload. To improve ore sorting/blending strategies and to account for variations in grindability, on-line run-of-mine composition analysis would be very helpful, but has received much less attention from the research community. This paper describes a general machine vision approach for on-line estimation of rock mixture composition, and is illustrated on a very challenging nickel mineral system: very heterogeneous minerals, similar coloration, and rock fragments can be dry or wet. The proposed mineral type recognition method involves: (1) dividing into sub-images; (2) extracting color and textural features using principal components analysis (PCA) and wavelet texture analysis (WTA), respectively; (3) reducing feature space dimensionality and removing dry/wet systematic variations using discriminant partial least squares (PLS-DA); and (4) establishing class boundaries using support vector machines (SVM). Through a pilot plant conveyor belt application, very good results were obtained for dry minerals. For wet rock mixtures, further investigation is required, but results are very promising.

Carbon anodes used in the production of primary aluminium have to be covered with a layer of particulate materials, a mixture of alumina and crushed electrolytic bath. Alumina composition analyses of this material are performed daily in laboratory on a very limited number of samples. However, analyses results from this small number of samples do not capture all the composition variations occurring on a daily basis. In this paper, alumina composition estimation of anode cover material, based on the use of image analysis and regression models is presented.

This paper summarizes research into the study of the control of flotation processes based on froth images. A review of the froth flotation operation background and the relevance and role of the froth phase are presented. The use of numerical image analysis techniques for the extraction of significant patterns from industrial flotation froths is discussed. First, the froth was analysed by the use of the Multivariate Image Analysis (MIA) method to extract the spectral variation contained in a RBG image. The Partial Least Squares (PLS) method was applied to develop an empirical model for the prediction of froth grade. The results of this application are illustrated using an industrial case study. Second, the froth was also analysed by Grey-Level Co-occurrence Matrix (GLCM) and Wavelet Transform Analysis (WTA) to extract features related more to morphological aspects. Since there is useful information hidden in froth texture, these methods are used as texture descriptor to classify different types of froths. They also provide qualitative information on the changes in the visual appearance of the froth. The pros and the cons of those methods are highlighted. Third, a control strategy based on a froth structure indicator was developed and implemented on an industrial zinc flotation circuit. The feedback controller uses the reagent flow-rate in order to maintain the froth structure at its set point. Process improvements obtained from the implemented control strategy are highlighted. Finally, the feasibility of implementing a hybrid control strategy consisting in a froth structure supervision controller based on the froth concentrate grade estimated by MIA and recovery inferred by image analysis is discussed.

The objective of this work is to develop a rapid and nondestructive machine vision sensor to predict Vibrated Bulk Density (VBD) of coke aggregate samples based on their surface textural characteristics obtained by imaging. This information could be useful for ultimately making real-time process adjustments to reduce green anode variability. Coke samples from different sources and sieved in a number of size classes were investigated individually and in blends. Wavelet Texture Analysis (WTA) was used to extract textural features of coke samples, and these were related to their VBD using Partial Least Squares (PLS) regression. It is shown that WTA captures variations in coke size and source and leads to good VBD predictions.

An important issue for the anode manufacturers is the increased variability of the raw materials and its negative impact on baked anode quality. This variability could be reduced by timely corrective actions applied on selected manipulated variables at the green mill (i.e. process control). However, the lack of rapid sensors for measuring anode quality is an important limitation to implementing feedback/feedforward control. Weekly averages of lab measurement are available too late to be used for feedback control. This work investigates the use of machine vision for sensing paste properties. The effects of changes in pitch demand of the aggregate and its particle size distribution, and mixing temperature on the visual appearance of laboratory paste image were studied. The 2D discrete wavelet transform (2D-DWT) was used to extract textural features for each image, followed by their analysis using Partial Least-Squares regression (PLS). Preliminary results show that the imaging sensor is sensitive to variations in pitch demand and aggregate size.

Aluminum reduction cells use about 1.9 kg of alumina in order to produce 1 kg of aluminum. That is, for modern reduction cells operating in the 350 to 400 kA range, 5000 to 6000 kg of alumina is fed daily. Considering that 5000 to 10000 kg of molten bath is available to dissolve the alumina, the dissolution rate is an important factor in order to avoid muck and enable alumina feed control system to operate within the 2 to 5% alumina concentration. However, on top of cell status, alumina properties have an impact on alumina dissolution rate. Hence, supplier changes and/or segregation of alumina within the delivery system may have negative impact on alumina dissolution rate leading to muck and/or anode effects. This paper discusses modification to an alumina feeder pipe promoting the dissolution rate. Promising results obtained during trial in a pilot plant section are presented and discussed.

Aluminum reduction cells typically use about 1.9 kg of alumina in order to produce 1 kg of aluminum. Hence, for modern reduction cells operating in the 350 to 400 kA range, 5000 to 6000 kg of alumina is fed to reduction cells on a daily basis. However, no information is available in an on-line fashion about the alumina properties fed to the pot. Alumina feeding control systems assume that alumina properties are constant for all pots within a potroom and also over time. Therefore, these control systems aim at controlling alumina concentration dissolved in the bath without accounting for the time varying effects of alumina properties and/or pot condition on alumina dissolution. Based on sampling campaigns, this paper presents evidences of time varying alumina properties impacting its dissolution rate and also proposes a novel approach in order to measure on-line, at the pot, parameters that are related to alumina dissolution.

Prebaked anode quality is available through a weekly average of core sample laboratory measurements. Unfortunately, long delays between production and results (approximately 4-6 weeks) can lead to poor abnormal operation and faulty anode detection, and difficult process control. Extensive raw material and process data are available at Alcoa Deschambault smelter’s carbon plant. Using projection to latent structures, a multivariate statistical method, it was possible to correlate raw material and process conditions to weekly lab results. The effect of different petroleum coke and coal tar pitch was analyzed and instant weekly prediction of anode properties was achieved.

Aluminum reduction cells performance are affected by many factors. In order to efficiently understand possible causes of performance upsets, all major sources of variations have to be monitored. This implies monitoring all anode and alumina properties, as well as pot state and manipulated variables, while also taking into account pot design or integrity after start-up. Considering the high number of variables involved in such a task, this is practically impossible using typical statistical process control tools. The problem is even worst when applied on a pot basis. This paper proposes the use of multiblock PLS (MBPLS) to build a monitoring scheme on a pot basis, simultaneously taking into account the influence of alumina and anode properties, of pot state and manipulated variables, as well as the pot state following its start-up. Derived from a regression model, the monitoring policy ensures that only variations relevant to pot performance variations are highlighted.