Croatia, Czech Republic, Germany, Greece, Spain, France
It is currently impossible to estimate cement strength with any certainty until its 2-day, 7-day and 28-day strengths are measured through physical tests. Plants tend to compensate for this by using a higher amount of costly, high-quality clinker and additives to ensure a high-performing product. Alternatively, they may grind the product more extensively, as the fineness of the cement is another factor in determining its final strength. These approaches offer no way of accounting for the multitude of other production process variables that can affect cement quality.
Predictive quality models use machine learning algorithms to correlate the quality of each production batch with the relevant production parameters while minimizing clinker factor.
This enables forecasting cement quality accurately in real-time at any point during the production process, reducing typical overspending of up to 15 percent in efforts to meet quality targets as well as achieving CO2 reductions and energy savings.
Interested in more information on this initiative? Then follow our Manufacturing Maqer Linkedin page for more insights!
Model predictions are in real time and 24/7, they are run and maintained centrally
Prediction of 2d, 7d and 28d strength, after mill and/or dispatch depending on market and plant requirements
Minimum Viable Product (MVP) was developed jointly by Digital, IT and the business with support of R&D
MVPs run in several plants to confirm business value
Teams that have contributed to this Proof of Concept include: Croatia, Czech Republic, Germany, Greece, Spain, France, Cement Excellence-Manufacturing (CE-M), Digital, Group IT, Innovation Center Lyon