Machine learning applications on IoT data in manufacturing operations and their interpretability implications: A systematic literature review

Abstract

Industry 4.0 has transformed manufacturing with real-time plant data collection across operations and effective analysis is crucial to unlock the full potential of Internet-of-Things (IoT) sensor data, integrating IoT with Artificial Intelligence (AI) techniques, such as Machine Learning (ML) and Deep Learning (DL). They can provide powerful predictions but anticipating issues is not enough. Manufacturing companies must prioritize avoiding inefficiencies, thereby developing improvement strategies from an Operational Excellence perspective. Here, the interpretability dimension of AI-based models could support a complete understanding of the reasons behind the outcomes, making ML and DL models transparent, and allowing the identification of the causal linkages between the inputs and outputs of the system. Within this context, this study aims first to deliver a comprehensive overview of the existing applications of Advanced Analytics techniques leveraging IoT data in manufacturing environments to then analyze their interpretability implications, referring to the interpretability as the description of the link between the independent and dependent variables in a way that is understandable to humans. Different gaps in terms of lack of full data enhancement are highlighted, providing directions for future research.