Speaker
Description
The application of machine learning techniques to the development of regression models aimed at minimizing electrical energy consumption in industrial heating processes has gained significant relevance in recent years. This approach enables accurate prediction of process behavior under specific operating conditions in a fast and cost-effective manner. One particularly relevant case study is the heating of steel billets using induction furnaces, where energy efficiency is a critical factor due to the high electrical power required to achieve the desired temperature profiles. However, induction heating involves complex interactions among electromagnetic fields, induced currents, and heat transfer mechanisms, which are difficult to analyze using purely physical models. These traditional modeling approaches often face limitations in accurately representing these dynamics under variable operating conditions. In contrast, machine learning–based regression models provide a flexible, data-driven framework capable of capturing such complexities. Moreover, the proposed linear models demonstrate how current intensity and energy consumption are influenced by factors such as frequency, heating time, and material properties. Methods such as multiple linear regression and other linear models incorporating regularization terms can be trained using experimental or historical data to predict the temperatures achieved as a function of current intensity and other relevant parameters. Once calibrated, these models enable rapid evaluation of different operating scenarios, facilitating the identification of optimal current intensity levels that minimize electrical consumption while ensuring compliance with constraints such as target temperature and heating uniformity. Additionally, these models support sensitivity analysis, offering clear and valuable insights into the relative influence of current intensity. Overall, this approach establishes a solid foundation for improving the efficiency and sustainability of induction heating processes.
This work received financial support from the Basque Government through projects KK-2025/00041 and KK-2023/00020 under the ELKARTEK Research Program.
| Speaker Company/University | University |
|---|