Towards Responsible AI: Evaluating Intelligent Models for Sensor Fault Detection Through the Lens of Sustainability and Performance Optimization

Abstract

This article introduces a methodological framework for evaluating intelligent models for fault detection by considering both technical performance and sustainability aspects. The impact of model design selection, such as model complexity and hyperparameter optimization, was examined in terms of accuracy, robustness, and sustainability. Through a manipulator robot, an analytical comparison is conducted of several configurations of Machine Learning (ML) models, including K-Nearest Neighbours (KNN), Random Forest Regression (RFR), and Support Vector Regression (SVR), and their performance to design parameters. The results highlight that the KNN model effectively balances technical performance and sustainability. This work contributes to the development of AI systems that align with the principles of sustainable design and reliable performance in smart manufacturing.