Abstract
The MONSOON project is a European H2020 innovation project dedicated to optimize process industry through resources and energy efficiency. The consortium is composed of 11 partners from 7 European countries. Rio Tinto and Aluminium Dunkerque (AD) are among the industrial partners, while ProbaYes acts as Data Science experts. The MONSOON project has built a two-components platform dedicated to both development and deployment of data analytics functions, employed for AD’s Paste Plant process optimization. The carbon anodes are a key component to the electrolysis reaction. The quality of the anodes (density, composition…) directly impacts the quantity and quality of the produced aluminum. A method, based on machine learning techniques, has been developed for monitoring the quality of the produced anodes and understanding the root causes of non-quality, using real-time Paste Plant data. This article presents the approach proposed in this context, the designed tools, and the first results obtained so far.