Increasing Potlife of Hall–Héroult Reduction Cells Through Multivariate On-Line Monitoring of Preheating, Start-Up, and Early Operation

Aluminum is produced inside metallurgical reactors known as pots that are replaced at the end of their service life. New pots are preheated, started, and then enter a period known as early operation in which different control strategies are used before entering regular operation. It is known that how preheating, start-up, and early operation are performed can damage a well-designed pot and lead to a shorter service life. However, the impact of these phases with respect to potlife is not well documented quantitatively. In this article, multivariate statistical analysis techniques are used to investigate the impact of pot-to-pot variations during the three phases. A partial least squares regression model is first proposed for predicting potlife, within an error of 90 days, using process data gathered until the end of early operation. This model is also used to identify those variables having the greatest influence on potlife. Finally, multivariate statistical process control charts are proposed to monitor the three steps efficiently. These charts have a low false-alarm rate and can help find the root cause of abnormal operation occurring during the early phases. A few examples are used to illustrate how operators and engineers could use the charts to maintain consistent early operation and help improve mean potlife. Nomenclature: In this article, bold characters are used to identify vectors (bold lowercase), matrices (bold capital), and three-dimensional arrays (bold, underlined capital). Lowercase italics letters are used to define indices. Al—Aluminium; Al2O3—Alumina; C—Carbon; CO2—Carbon dioxide; kA—kilo-Amperes; Na—Sodium; Na3AlF6—Cryolite; V—Volt.