Bootstrapping Regression Support Tool for Energy Management in Induction Melting Processes: a Multi-Plant Case Study

Abstract

Efficient energy management is a basis of operational quality in the foundry industry, where induction melting accounts for 70–80% of total electricity costs. This study proposes a decision-support model for predicting energy consumption, developed using data from 11 industrial plants. Facing the challenge of a small dataset, a bootstrap regression approach was employed to ensure robust estimation of Energy Performance Indicators (EnPIs). The resulting model shows high explanatory power and yields statistically significant coefficients for all variables. The model identifies key managerial and technical drivers, such as charge contamination, packing density, and organisational heat loss (furnace temperature). Unlike purely technical models, this approach utilises semi-annual aggregated data, making it a strategic tool for energy budgeting and benchmarking. The model’s predictive power was externally validated against a 12th foundry, yielding a precision of 98,4%, thus proving its reliability for managerial forecasting and carbon footprint auditing in the framework of ISO 50001 standards.