Machine learning to optimise chemical synthesis
Our client mandated MP DATA to use an innovating approach to study potential non-quality root causes and develop a process parameter recommendation engine to maximize output quality depending on the production context.
To further an initial analysis, we used an analytical approach based on 100% explainable prediction models capable of predicting non-quality risks from the very first synthesis phases as well as the corrective actions to take to minimise them. Model explainability not only made it easier for the solution to be adopted, but also to complete the root cause analyses through targeted process studies that confirmed the Machine Learning approach conclusions.
The solution was then directly included on specific screens in the control room.
compared to the worst situations the site had encountered
>250k€ of savings
per year at least