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

https://doi.org/10.1016/j.conengprac.2020.104533Get rights and content

Highlights

  • A process mechanism knowledge-driven method is proposed and establishes a bridge between process mechanism research and industrial online application. The empirical knowledge, process mechanism knowledge and data knowledge are explicitly represented by a parameter that can be automatically obtained online.

  • The pseudo dissolution hysteresis coefficient (PDHC) is proposed. The PDHC can overcome the inability to express the online dissolution performance of alumina in industrial cells.

  • The relationship between the PDHC and the online bath temperature is identified. The bath temperature recognition based on the PDHC can alleviate the inability to measure the temperature online.

  • The online alumina concentration anomaly detection base on the PDHC can not only detect concentration anomalies, but also detect the anomalies approximately two hours earlier than the slope-based method. Moreover, the proposed method is simpler than the slope-based method.

  • A PDHC-based feed optimization control framework is developed. In this framework, the feed feedback compensation control loop and the real-time correction of parameters based on online temperature information are incorporated, which are difficult to implement in the slope-based method.

Abstract

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.

Introduction

Aluminium electrolysis is a process in which alumina is used to obtain molten aluminium (metal) on a cathode by passing direct current through a carbon anode in a molten cryolite (bath) in an electrolysis cell, with carbon dioxide and carbon monoxide being emitted at the anode. Due to the complex composition, high temperature and strong corrosion of the bath, it is highly difficult to obtain the alumina dissolution rate, the bath temperature and the alumina concentration online. With the current high-efficiency energy-saving technology in modern aluminium electrolysis and the large-scale trend of aluminium electrolysis cells, it is difficult to effectively dissolve more alumina with a smaller bath volume, a lower bath temperature and a lower electrolysis voltage.

The dissolution process of alumina in the bath is highly complicated (Jain et al., 1983, Lavoie et al., 2016), and the alumina that cannot be dissolved in time will form a sludge and crust, which may directly affect the stability of the cell. Many scholars have conducted experiments and studies on the reaction mechanism and influencing factors of the alumina dissolution process in the Lindsay, 2014, Welch and Kuschel, 2007 and Yang et al. (2015), and many dissolution models have been proposed in the Bardet, et al., 2016, Haverkamp and Welch, 1998, Vasyunina et al., 2009 and Zhan et al. (2014). Several studies (Henry and Lafky, 1956, Meng et al., 2010, Zhou et al., 1998) conducted chemical experiments at a specified temperature and bath composition environment, which showed that the dissolution rate of alumina was very high in the first 5 to 10 min after the alumina was added to the melt and subsequently became very low. In an industrial cell, the influencing factors are constantly changing, and the main influencing factors also differ from those of chemical experiments. One study (Kuschel & Welch, 1990) found that in an industrial cell, the bath temperature, the bath agitation and the feed mode have substantial impacts on the speed of the alumina dissolution. However, little research has been conducted on the dissolution hysteresis of alumina in industrial cells. The authors of Liu and Li (2008) carried out experiments on 160-kA prebaked anode cells with a cryolite ratio of 2.6–2.7, a bath temperature of 955 C and 5% calcium fluoride, and the results demonstrated that the dissolution hysteresis time was approximately 2 min and the time that was required for dissolving 63.2% of the alumina was approximately 10 min. Most of the research on alumina dissolution performance has remained in the experimental stage. Very few reports have been published on the online dissolution performance of alumina in industrial cells.

The bath temperature is the target of the temperature control, which has a substantial effect on the current efficiency. In the industrial production process, each bath composition corresponds to the most suitable operating temperature (Utigard, 1999), which is termed the normal bath temperature. For example, if the traditional bath composition (3%–7% excess AlF3, 3%–7% CaF2, 4%–5% Al2O3) is used, the corresponding most suitable bath temperature range is 965 C–975 C (Qiu, 2005). If the bath temperature exceeds this normal range, abnormal working conditions related to the bath temperature are likely to occur. If the temperature continues to be higher than this normal range, the cell will enter a hot stroke, and the side ledge and crust will melt, which in turn will increase the molecular ratio, thus reducing the current efficiency. If the bath temperature continues to fall below this normal temperature, the cell will gradually enter a cold stroke where the bath is viscous and the fluidity is poor. This is not only detrimental to the evolution of the anode gas but also reduces the solubility and dissolution rate of the alumina, which can cause abnormal working conditions such as increased cell bottom deposition and frequent anode effects. This seriously affects the stability of the cell and eventually causes the “secondary reaction” of the metal, thereby reducing the current efficiency. Since a sheath of thermocouples that can be used for a long time has not been found, online temperature measurement remains highly difficult. At industrial sites, people typically use consumption intermittent methods or rely on field experience. The alumina concentration is the target of the feed control. Since the online continuous measurement of the alumina concentration does not satisfy the feed control requirements of the industrial cells, the most commonly used method is the cell voltage (or resistance) slope method. A concentration that is too low can easily induce an anode effect (Vogt, 2013, Vogt and Thonstad, 2002). As the cell voltage rises sharply, the energy and material balances of the cell may be destroyed. The commonly used predictions of the anode effect are based on the methods of cell voltage (or resistance) slope tracking (Blatch et al., 1992, Meyer and Earley, 1986) and employing the slope as the main basis and a material balance estimate as the auxiliary basis (Li et al., 2001, Zhou, et al., 2015). Herein, a method of this type is referred to as a slope-based method. In contrast, a concentration of alumina that is too high will result in deposition on the cathode, which may impair the cell stability and reduce the current efficiency. Unfortunately, few reports have been published on the online detection of a high alumina concentration in industrial cells. The typical methods for detecting an abnormal concentration include model-based methods (Gao et al., 2015a, Yurkov et al., 2004) and data-based methods (Chandola et al., 2009, Gao et al., 2015b, Hestetun and Hovd, 2006). The data-driven methods include expert systems (Berezin et al., 2005, Cao et al., 2011, Rolland, et al., 1991), multivariate statistical methods (Majid et al., 2011, Stam et al., 2008) and intelligent algorithms (Li et al., 2006, Yi et al., 2016).

Most research results on the alumina dissolution performance, the bath temperature and the alumina concentration are based on a mechanism model or an experimental cell for specified process conditions, and many key parameters are fixed. In industrial cells, these parameters change in real time and cannot be measured online. Therefore, it is difficult to establish an online mechanism model that adapts to the frequent changes in the working conditions, which limits the applicability and accuracy of mechanism-model-based methods. Due to the absence of process mechanism knowledge, pure data-driven methods result in complex algorithms, high computational cost, and insensitivity to abnormal working conditions. Signal-based methods typically extract the main statistical characteristics of the signal as input to data-driven methods, which have the same problems as pure data-based methods. Therefore, the alumina dissolution properties are still not widely used as control basis in the industry. The temperature control relies mainly on intermittent measurements, and the concentration anomaly detection is realized mainly via a slope-based method.

The aluminium electrolysis process is an electrochemical reaction process that involves complex mechanism knowledge. The relationship among the alumina dissolution performance, bath temperature, and alumina concentration is complex and interactive. The cell control system collects and stores a large amount of historical production data. These data consist mainly of simple time-series data but contain a wealth of online information. In this paper, a method utilizes process mechanism knowledge to perform mechanism analysis and explicitly represents the empirical knowledge, process mechanism knowledge and data knowledge as a parameter — pseudo dissolution hysteresis coefficient (PDHC) that has semantic knowledge and can be automatically obtained online. The proposed PDHC quantifies the alumina dissolution performance of industrial cells and can characterize the alumina dissolution performance online. Thus, the PDHC can not only recognize the bath temperature online but also advance the detection time of an abnormal concentration online. By using the parameters that have semantic knowledge and can be automatically obtained online, the shortcomings of the methods that are based on the mechanism model or driven by pure data can be overcome.

The rest parts of the paper are organized as follows. In the second section, according to the process mechanism analysis, the process mechanism knowledge M1–M11 among the alumina dissolution properties, bath temperature, alumina concentration, melt motion and cell voltage is obtained, which is the basis for identifying the PDHC. In the third section, the PDHC with normal concentration is obtained by a qualitative analysis and a quantitative representation. In the fourth section, the process semantemes of the five common cases are analysed separately, and the PDHC that corresponds to the abnormal concentration is obtained by the process semantic embedding. The PDHC identification algorithm is presented in Section 5. Section 6 describes the application of the PDHC to bath temperature online recognition, alumina concentration anomaly detection and feed control. Section 7 presents the conclusions of this paper.

Section snippets

Mechanism analysis

The modern aluminium electrolysis industry uses a cryolite-alumina molten salt electrolysis method to dissolve added alumina powder in a molten electrolyte. The total reaction formula of the electrolysis process is 12Al2O3+34γC=Al+34(21γ)CO2+32(1γ1)COwhere γ is the current efficiency.

In feed control (alumina concentration control), the on-demand feed control technology, of which the theoretical foundation is the U-curve in Fig. 1, is typically adopted. According to Fig. 1, under the stable

Explicit representation the online dissolution property for a normal alumina concentration

In this section, based on a qualitative analysis of the alumina dissolution hysteresis phenomenon, a quantitative representation is realized in which the pseudo dissolution hysteresis coefficient (PDHC) and the pseudo dissolution hysteresis time (PDHT) are obtained. The PDHC and PDHT can overcome the present inability to quantitatively represent the online dissolution properties of alumina in industrial cells.

Explicit representation the process semanteme for an abnormal alumina concentration

According to the above analysis, the pseudo alumina concentration that is extracted from the NCV has the same tendency as the NCV. Therefore, this section continues to analyse the working conditions of abnormal aluminium concentrations in terms of the relationship between the pseudo alumina concentration and the feed state. According to numerous production process investigations and process mechanism analyses, a breakdown of the “under-rise and over-fall” relationship can be observed when

Identification of the PDHC

The PAC PUj can represent the online alumina concentration in an industrial cell, and the extreme point of PUj can represent the transition point of the change in the alumina concentration. Therefore, this section will use the relationship among the position, the number of the extremal points of PUj and the feed state FUj as the main basis for the design of the PDHC identification algorithm.

PDHC application in online bath temperature recognition

In an aluminium electrolysis plant, workers use consumable thermocouples to measure the bath temperature every day, which has disadvantages such as high cost, inability to conduct online measurement, and only the local bath temperature can be obtained. Another approach is to use an infrared temperature measuring device to measure the peripheral temperature of the cell, which is substantially affected by the outside conditions, thereby leading to a large error. The bath temperature in an

Conclusions

This study is an application case of knowledge automation (Gui et al., 2016, Gui et al., 2018a, Gui et al., 2018b) in the industrial aluminium electrolysis process. In this paper, an identification method for the pseudo dissolution hysteresis coefficient that is driven by process mechanism knowledge for the industrial aluminium electrolysis process is proposed. The time complexity is O(nifp); hence, it is suitable for the frequently changing conditions of industrial aluminium electrolysis

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

References (50)

  • Bonny, P., Gerphagnon, J.-L., Laboure, G., Keinborg, M., Homsi, P., & Langon, B. (1984). Process and apparatus for...
  • ChandolaV. et al.

    Anomaly detection: A survey

    ACM Computing Surveys (CSUR)

    (2009)
  • GaoZ. et al.

    A survey of fault diagnosis and fault-tolerant techniques-Part I: Fault diagnosis with model-based and signal-based approaches

    IEEE Transactions on Industrial Electronics

    (2015)
  • GaoZ. et al.

    A survey of fault diagnosis and fault-tolerant techniques-part II: Fault diagnosis with knowledge-based and hybrid/active approaches

    IEEE Transactions on Industrial Electronics

    (2015)
  • GerlachJ. et al.

    The dissolution of aluminum oxide in cryolite melts

    Metallurgical Transactions

    (1975)
  • GrjotheimK. et al.

    On convection in aluminum reduction cells

    Metallurgical Transactions

    (1970)
  • GuiW.H. et al.

    Knowledge automation and its industrial application

    Science China: Scientia Sinica Informationis

    (2016)
  • GuiW.H. et al.

    Process industry knowledge automation and its applications in aluminum electrolysis production

    Control Theory and Applications

    (2018)
  • GuiW.H. et al.

    A review of intelligent optimal manufacturing for aluminum reduction production

    Acta Automatica Sinica

    (2018)
  • HaurpinW.

    Polarization in an aluminum reduction cell

    Journal of The Electrochemical Society

    (1956)
  • HenryJ.L. et al.

    Solubility data for aluminum reduction systems

    Industrial and Engineering Chemistry

    (1956)
  • JainR. et al.

    Interaction of aluminas with aluminum smelting electrolytes

  • Kuschel, G. I., & Welch, B. J. (1990). Effect of alumina properties and operation of smelting cells on the dissolution...
  • KvandeH. et al.

    Cell voltage in aluminum electrolysis: A practical approach

    JOM

    (2000)
  • LavoieP. et al.

    A review of alumina feeding and dissolution factors in aluminum reduction cells

    Metallurgical and Materials Transactions B

    (2016)
  • Cited by (14)

    • An improved TOPSIS-based multi-criteria decision-making approach for evaluating the working condition of the aluminum reduction cell

      2023, Engineering Applications of Artificial Intelligence
      Citation Excerpt :

      The working condition evaluation of the cell is the basis of formulating operation strategy, ensuring production safety and realizing stable and optimized operation of the aluminum electrolysis process (Shi et al., 2022; Yue et al., 2020a). In practice, the production management of the cell is mainly carried out through two steps: (1) professional technicians evaluate the working condition of the cell by analyzing the process data every day (Zeng et al., 2020; Li et al., 2022; Zeng et al., 2021); (2) workers operate and maintain the cell according to the working condition evaluation result of the cell (Yi et al., 2016, 2018; Wang et al., 2022). If the working condition evaluation of the cell is not accurate, the operation and maintenance of the cell will be problematic, leading to the cell in bad working condition (Yue et al., 2017; Huang et al., 2020a; Yu et al., 2019).

    • A Cell Condition-Sensitive Frequency Segmentation Method Based on the Sub-Band Instantaneous Energy Spectrum of Aluminum Electrolysis Cell Voltage

      2021, Engineering
      Citation Excerpt :

      This cell condition information of different frequencies is often superimposed together and eventually appears as various complex oscillation forms in the cell voltage. The low-frequency component of the cell voltage corresponding to the alumina concentration is used to calculate the voltage slope [2,13,14]; the low-frequency noise of the cell voltage related to the metal (molten aluminum) movement is used to calculate the voltage swing; and the high-frequency noise of the cell voltage that is linked to anode problems is used to calculate the voltage vibration [15,16]. Cell voltage frequency segmentation is the basis for determining the passband of digital filters that are used to separate the cell voltage components relating to these pieces of cell condition information [17].

    View all citing articles on Scopus

    This work was supported in part by the National Natural Science Foundation of China [grant numbers 61751312, 61773405, 61533020].

    View full text