Circular Design Through Machine Learning: A Preliminary Study About Disassembly Time Evaluation of Rusted Screws

This study presents a method for estimating product disassembly times using machine learning techniques. It demonstrates how to integrate the machine learning model with economic evaluations, circularity indices, and environmental impacts. The goal is to support sustainable design and optimize product end-of-life processes, enabling more informed decisions within the circular economy. The predictive model will be trained on experimental data to accurately estimate disassembly times, considering variables such as the condition of mechanical joints. A preliminary exploratory case study validated the characteristics and experimental dataset required for data collection through a screw rusting test. The results show a significant correlation between joint characteristics, their end-of-life conditions, disassembly configuration, and total disassembly time. Further developments are planned to expand the experimental database and refine the model, extending its application to different products and industrial contexts.