Challenge
CCB Cementir, a key player in the cement industry, continuously seeks to improve energy efficiency and production throughput to remain competitive in the global market. With cement grinding being a particularly energy-intensive process, CCB aims to set industry benchmarks for sustainability and operational excellence. However, achieving this vision is hindered by their inability to maintain consistent control over key mill parameters, directly impacting operational efficiency and profitability.
Underlying factors causing this challenge include:
- High energy consumption: grinding mills consume a significant amount of energy, which increases the production costs, and hence reducing energy consumption was a priority
- Limited throughput efficiency: maximising material flow through the mill without compromising product quality was a complex challenge, constraining overall productivity
- Inconsistent cement fineness: maintaining consistent fineness while adapting to fluctuating production conditions proved to be a tedious task, compromising quality assurance and customer satisfaction
To address these challenges, CCB needed an AI-driven solution that could optimise energy consumption and throughput, while keeping in mind process variability.
Solution
Faktion developed an AI-powered recommendation system that analyses cement mill sensor data and provides real-time process adjustments to operators. Central to the solution was the use of reinforcement learning (RL) - an advanced AI approach that optimises complex decision-making tasks through interaction and feedback loops.
The reinforcement learning component involved training an AI agent using a digital twin of the cement mill production process. This AI agent dynamically determined optimal operational setpoints (such as separator RPM and feeder rates) by continuously learning from interactions within a simulated environment. By experimenting with various configurations virtually, the RL agent efficiently identified the best setpoints to achieve both desired cement fineness and optimal mill filling level.
The solution focused on:
- Delivering real-time recommendations directly integrated into existing SCADA systems, enabling immediate, actionable interventions to reduce energy consumption
- Optimising throughput efficiency by leveraging historical and real-time data analysis, ensuring maximal productivity without sacrificing quality
- Reducing process variability through continuous AI-driven monitoring, achieving stable cement fineness under varying production conditions
Approach
Faktion's comprehensive, collaborative approach ensured the solution was practical, scalable, and directly aligned with CCB’s operational requirements:
- Data analysis & process understanding: the team analysed historical sensor data from the cement mill to identify trends, inefficiencies, and improvement opportunities.
- Digital Twin & Reinforcement Learning: a digital twin model of the mill provided a realistic environment for training the RL agent, significantly accelerating optimization compared to traditional methods.
- Operator feedback & iteration: the system was tested in a real-world production environment, allowing operators to validate recommendations and provide feedback for ensuring maximum relevance, usability, and accuracy of AI recommendations.
Outcome
The PoC successfully provided valuable process insights and operational guidance to CCBs production teams. Key outcomes included:
- Improved operator decision-making, enabling better understanding and precise adjustments of key mill parameters.
- Enhanced process stability, reducing variability in cement fineness and material flow consistency.
- Established foundation for broader AI integration, demonstrating clear potential for future scalability and full-scale AI optimization.
By leveraging AI-powered process control, CCB is taking a data-driven approach to optimising mill operations, ensuring more efficient and sustainable cement production in the future.