Rusal has implemented neural networks to improve alumina quality

At the decomposition site of the Rusal alumina plant in Kamensk-Uralsky, AI technology has been introduced to control the size of aluminum hydroxide.

According to the company's press service, "the elements of the technology are a neural digital twin of the decomposition process, trained on 15 years of retrospective data, and a mathematical algorithm that performs multiparametric optimization of the process."

Decomposition is considered a key step in a multi-stage method for producing alumina from bauxite ore. The quality of alumina and aluminum smelted from it directly depends on the quality of the intermediate product of decomposition, aluminum hydroxide. In the decomposer battery, the aluminate-alkaline solution obtained from bauxite decomposes with precipitation of aluminum hydroxide. Alumina (alumina oxide) is produced from it at the final stage of the cycle.

"A machine-trained neural network accumulates data on the fractional composition of aluminum hydroxide and the physical parameters of decomposition over the past 120 days and provides a forecast of the fractional composition for 90 days ahead. Based on the forecast, the optimization algorithm suggests parameter changes that can improve the fractional composition of aluminum hydroxide," said Viktor Mann, Technical Director of Rusal.

According to him, forecasts contain less than 5% of errors, which allows the technologist to manage the process more accurately than was previously possible. "Thanks to the introduction of technology, the plant has reduced the content of fine fractions in products by 4.4%. The consolidation of alumina improves the performance of dry gas purification of aluminum plants, reduces the consumption of alumina, anodes and electricity per ton of aluminum, thereby improving the performance of electrolysis," explained Mann.

The director of the alumina division of the company, Yakov Itskov, added that thanks to the use of neural networks, for the first time the possibility of predictive advisory decomposition management was obtained. This decision was ahead of similar projects by Western providers, he stressed.

All important news is on the Izvestia channel in the MAX messenger.