The impact of anode-related process dynamics on cell behaviour during aluminium electrolysis

Abstract

The HaII-Heroult process for the production of aluminium is a three phase, high temperature operation, influenced by magnetohydrodynamics with many inter-related variables and associated dynamics. Despite significant advances in mathematical modelling of cell designs, process control strategies, and the study of alternative processes, energy efficiency remains around 40%, with scope for improvement. The objective of this thesis was to investigate the impact of anode-related process dynamics on cell behaviour, since this aspect is strongly linked with electrochemical inefficiency. This was achieved by analysing the electrical character of anodic cell signals, which required a data collection and digital signal processing system to be developed. The signal parameters which correlated best with cell behaviour in this study were the signal magnitude, amplitude and frequency components. These parameters were extracted using digital signal processing techniques, and referred to as the DC Offset, AC Component, AC/DC Ratio, and the location of the Dominant Frequency. Data presented in this thesis highlighted the pulsing nature of current flow through an individual anode, which changed in both intensity and frequency with different electrochemical conditions, allowing correlations to be established. Signal analysis techniques were used to transduce the gas evolution process, allowing the following aspects of the anode process to be studied: Normal Cell Operations The overall anode current cycle was reduced to three stages: current-rise, current-peak, and current-decay. Both the current-rise and current-decay stages were fitted to first order exponential functions for all cell anodes, which were then grouped and modelled as a function of spatial arrangement in the cell. The results showed centrally located anodes reached peak current fastest after 0.78 days, whereas corner anodes required 1.63 days. These values were principally controlled by the thickness of freeze under the anode following replacement. This behaviour is related to the anode setting profile, which resulted in centre anodes being set too low with respect to the metal pad curvature, as evidenced by the rapid pick up of current and uneven current distribution. From the anode current model, the total charge transfer for the cell anodes over the cycle was determined, showing predicted carbon consumption rates of 1.80 cm/day and 1.68 cm/day for centre and corner zones respectively. These trends were consistent with actual carbon measurements. Cell current distribution at anode change was not uniform across the cell. Anodic signal data showed anodes adjacent and directly opposite the anode pair removed, acquired a disproportionately larger amount of current load. Typically, the load increased by around twice the theoretical cell average of 10% extra load per remaining anode. A plausible explanation for the observed superimposed effects centred on favourable magnetohydrodynamics profiles and bath chemistry, which facilitated current flow in localised areas of the cell. Over an anode cycle, the intensity of individual anode current fluctuations reduced dramatically, in the order of 50%, while the frequency of gas release increased by around 25 %. The changes in signal character were interpreted as a direct result of reduced bubble coalescence and faster release of gases from the rounded, inclined electrode surface. Anode Carbon Consumption Anode carbon measurements showed electrochemical consumption increased with higher average current density, but varied considerably with cell zone. This observation reflected the uneven current distribution produced by the initial setting profile. A modified setting profile which better matched the contour of the metal pad resulted in a distinct smoothing in the variation of electrochemical consumption across the cell. The standard deviation of butt measurements decreased by around 40 %, indicating a more even current distribution. Abnormal Cell Operations Signal analysis of the individual anode current of a spiked anode revealed a high frequency/large amplitude fluctuating waveform. This was due to its associated higher current density and intermittent shorting with the metal pad. The intense electrical activity of an individual spiked anode was reflected in the cell voltage signal as a series of higher frequency “fine structures” superimposed on the normal voltage waveshape. Individual anode current signals recorded for moderate anode cathode distance (ACD) variation showed a near linear increase in current draw with ACD reduction. Increasing the current density will increase the rate of gas formation in a near proportional manner. This will produce an average higher degree of surface coverage of the electrode by the gas. Larger current oscillations in the individual anode currents, due to the fluctuating gas resistance layer, will occur. This effect, combined with the larger current magnitude, caused the AC component to steadily intensify with reduced ACD. At the same time, a proportional increase in the frequency of bubble release was observed, reflecting the increased gas production conditions at lower ACD values. At very low ACD levels (4.5 cm below reference), a direct short-circuit occurred. At this point, the electrochemical voltage was eliminated, with the anode current increasing rapidly, without gas evolution. The frequency of gas release was observed to decrease significantly, consistent with the short-circuit condition. The onset of an anode effect (AE) was investigated by analysing the electrical behaviour of anodic signals during this period. Data analysis around 1 - 2 minutes before each AB, showed a particular individual anode abruptly dropped in current density by - 50%. This was thought to be due to localised build up of AE gases, through poor alumina feeding being maintained in that cell zone. At the same time, the anode riser current in that half of the cell reduced in magnitude, reflecting the localised resistance build up. The opposite anode riser increased in current draw proportionately, as the load was redistributed. Ultimately, the cell voltage increased to - 40 V. marking the start of a full constant current AE. The study showed the onset of an AE may be localised and detected using individual anode and riser current measurements. Drained Cathode Cell Operations The impact of electrode geometry on gas release and electrolyte dynamics was investigated in an industrial drained cathode cell. Current flow through a flat anode early in Life was characterised by a series of low frequency harmonics around 1.1 Hz, with relatively high amplitude pulsations, registering ~145% of cell average. The current flow through a 4° V-shaped transversal sloped anode of similar age, however, showed a significantly reduced AC intensity, registering ~ 45% of cell average. The power spectrum indicated the overall spectral energy content was weak, with a dominant frequency peak located around 2 Hz. This was considered a direct result of the increased angle of inclination from the horizontal of the sloped anode, which reduced bubble coalescence and facilitated faster gas release, leading to lower current fluctuations. As these studies were conducted on a metal-free cell, the observed oscillations in the individual anode current signal are clearly due to bubble fluctuations in the electrolyte, which cause localised screening to the current, due to the resistive nature of the gas. The onset of an anode effect (AE) was investigated by analysing the electrical behaviour of anodic signals during this period. Data analysis around 1 - 2 minutes before each AE, showed a particular individual anode abruptly dropped in current density by ~50%. This was thought to be due to localised build up of AE gases, through poor alumina feeding being maintained in that cell zone. At the same time, the anode riser current in that half of the cell reduced in magnitude, reflecting the localised resistance build up. The opposite anode riser increased in current draw proportionately, as the load was redistributed. Ultimately, the cell voltage increased to ~ 40 V. marking the start of a full constant current AE. The study showed the onset of an AE may be localised and detected using individual anode and riser current measurements. Drained Cathode Cell Operations The impact of electrode geometry on gas release and electrolyte dynamics was investigated in an industrial drained cathode cell. Current flow through a flat anode early in life was characterised by a series of low frequency harmonics around 1.1 Hz, with relatively high amplitude pulsations, registering ~ 145%of cell average. The current flow through a 4° V-shaped transversal sloped anode of similar age, however, showed a significantly reduced AC intensity, registering ~45% of cell average. The power spectrum indicated the overall spectral energy content was weak, with a dominant frequency peak located around 2 Hz. This was considered a direct result of the increased angle of inclination from the horizontal of the sloped anode, which reduced bubble coalescence and facilitated faster gas release, leading to lower current fluctuations. As these studies were conducted on a metal-free cell, the observed oscillations in the individual anode current signal are clearly due to bubble fluctuations in the electrolyte, which cause localised screening to the current, due to the resistive nature of the gas. Analysis of individual anode current signals from several other industrial cell designs, including the drained cathode cell, concluded the anode bubbling mechanism was profoundly affected by anode width (area), at similar current density. Small conventional cathode cells showed an average AC fluctuation of ±74 A. Using the Bruggemann equation, these current fluctuations corresponded to a change in average gas volume fraction of ±8.9%. The calculated variation in gas fraction increased in a near linear fashion with anode area to ±29.2% for large conventional cells, consistent with the larger signal fluctuations which averaged ±527 A. This was interpreted as a larger anode area promoting increased bubble coalescence, which increased the average gas volume fraction, giving rise to higher AC oscillations. Sloped electrodes in drained cathode cells, however, exhibited a significantly lower average AC content of ±33 A, which corresponded to a ±3.7% change in gas volume fraction, even at similar sizes to their flat anode counterparts. This is consistent with the rapid evacuation of bubbles through increased anode tilt. It is suggested sloped anodes promote higher frequency chain-type bubbling, however, the accumulation of gases associated with flat anodes produce a change in the gas release mechanism, to one of lower frequency severe pulsing-type character. Overall, the results of the thesis confirmed the importance of using individual anode current measurements during cell operations, by supplying more detailed information about the electrochemical process. Digital signal processing of anodic cell currents represents a noninvasive method for auditing smelting cell behaviour, providing new and useful information to assist in the diagnosis and classification of process disturbances. Through the use of appropriate software tools and suitably robust hardware for the industrial environment, anode-related process dynamics can be more effectively monitored and controlled; an aspect of the smelting process which has not been studied extensively in the past.