In the aluminum recycling plants, the melting process in rotary furnaces was organized in non-standardized casts, depending on the subjective criteria of the people who operated them. This lack of uniformity prevented the automation of the process, hindered energy and operational efficiency, and generated a strong dependence on the tacit knowledge of the personnel.
Developed a neural network-based predictive model trained with historical furnace data – highly sensed with per-second data – to anticipate the optimal time to recharge the furnace, using “torque” as a key variable.
The system identifies the instant when the furnace has reached an optimum state of metal fluidity, allowing automated reloading without the continuous supervision of the operator. This allows:
- Automate the casting process.
- Increase operational efficiency.
- Reduce reliance on undocumented knowledge.
- Improve productivity and reduce energy consumption.
We apply neural networks to predict the optimal reloading time in aluminum recycling furnaces, automating casting and improving efficiency, productivity and energy consumption.
- Neural networks trained with two years of operational data.
- 80% prediction accuracy within ±6 minutes.
- Real-time automation of critical processes.
- Transformation of expert knowledge into replicable logic.
- Increased casting and reduced consumption: more efficiency, fewer resources.
