Enhancing resource efficiency

Data science for enhancing resource efficiency in manufacturing processes

Started

January 1, 2023

Status

Completed

Share this project

About 96% of the 16.5 million tons of malt made in the world is used as the main ingredient for beer production. Bühler estimates that about 75-80 % of the world’s barley malt is produced using its malting equipment. The MontBlanc solution was developed to address the carbon emissions associated with the high energy demand of malting processes.

People

Collaborators

SDSC Team:

Clément Lefebvre

Senior Data Scientist

Clément became part of the SDSC team in January 2019, assuming the role of a Data Scientist with an emphasis on industry-oriented projects. He holds a Bachelor of Science degree in Physics, acquired in 2016 from the École Polytechnique Fédérale de Lausanne (EPFL) in Switzerland. Following this, he earned a Master of Science in Computational Science and Engineering in 2018, also from EPFL. Throughout his academic journey, Clément concentrated on leveraging Data Science and Machine Learning techniques to enhance efficiency in industrial processes. Subsequently, he developed a specialized interest in Generative AI, with a particular focus on Natural Language Processing (NLP), especially in the application of Large Language Models for innovation.

Clément Lefebvre

PI | Partners:

Buehler Group

Dr. Matthias Graeber, Head of Data Science

description

Objectives

‍MontBlanc aims at reducing the energy consumption, and therefore greenhouse gas (GHG) emissions, of the malting process by optimizing a key component of the procedure. In a typical plant, the kilning process (i.e., the drying of the malt) follows a strict, pre-defined schedule with defined process parameters. However, depending on batch characteristics (such as initial moisture level) and weather conditions, the process parameters do not always fit the pre-defined schedule. MontBlanc leverages a large set of variables (weather, plant environment, grain characteristics) to forecast the evolution of the kilning process for each batch of product. It estimates the waiting time expected at the end of the next batch and then it computes the optimal fan speed necessary to reduce this expected waiting time and thus optimizing the electrical power consumed.

Benefits

After deploying the MontBlanc service in an industrial trial at a pilot customer for a period of 11 months, about 16% reduction in electrical power consumption with respect to standard operations has been observed. The food production domain is vast and complex, with thousands of processes and machines in use. MontBlanc shows a tangible example of data science potential in the manufacturing sector.

*The different steps of the malting process (Source: German Maltsters’ Association)*

Notes

‍Special thanks to the Bühler GQ Unit, and particularly to Matthias Graeber, Head of Data Science at Bühler Group, for their support and exchanges during the MontBlanc project.

‍