Case Study

Our client is a large sugar mill looking to improve energy efficiency and production output. We provided a solution that combines a Digital Twin of the mill with a Reinforcement learning optimization algorithm to dynamically control the milling process.

​

The Digital Twin is a model that simulates the operation of the factory based on the mill settings and input cane data. It calculates the amount of juice extracted at each mill in the milling train based on its settings and given RPM and cane parameters. This model can be used for experiments with new settings and training of the optimization model.

​

The optimization model reduces the energy used for the extraction of sucrose juice from prepared sugarcane by using a milling train. Depending on the type of prime mover (electric motor/steam turbine), the system saves energy either in the form of electrical power or steam while improving extraction. The algorithm dynamically sets the prime mover to the lowest possible RPM value needed to produce the crushing force, ensuring optimal extraction while saving steam. The optimal RPM is calculated based on the properties of the incoming cane and the settings of the mill (work openings, motor size, roller groove depth, type, etc.). The values are calculated automatically by a two-stage system.

​

The algorithm improves the extraction rate by 3% and reduces energy consumption by 7%. This translates to significant cost savings for the client.

Key concepts