2025

Terziyan, Vagan; Vitko, Oleksandra

This paper addresses a critical gap in the artificial intelligence (AI) and machine learning (ML) capabilities used in smart manufacturing by integrating an adaptive generalized mean function and a generalized social distance metric. Current AI/ML approaches often struggle with the dynamic, heterogeneous nature of manufacturing environments. Our proposed framework offers a flexible, context-aware solution for tasks such as clustering and data aggregation in the Industry 4.0 and 5.0 landscapes. Generalized mean functions offer flexibility in data aggregation, while the social distance metric provides new insights into data relationships, including factors like social asymmetry. We presented novel generalized mean function on the basis of Lehmer mean and enhanced by the sigmoid-logit pair. Such a mean can be controlled by two trainable parameters and has several useful properties for various ML and AI tasks. We presented also two options for the generalized social distance metric, each utilizing the suggested generalized mean function in different ways. The first one is based on social asymmetry of the neighborhoods in data distribution. The second one measures the distance between the centroids of corresponding social context areas. Our approach offers practical benefits for a variety of smart manufacturing tasks, effectively addressing the limitations of existing methods. Find presentation slides here: https://ai.it.jyu.fi/ISM-2024-MEAN-DISTANCE.pptx