Enhancing Decision Quality in Smart Manufacturing: Uncertainty-Aware Evaluation of Edge–Cloud Architectures with T-Spherical Hesitant Fuzzy Rough Sets

Turgay S., Başar E. E., Çalışan M., Geçkil M. F., Baydaş M., Stević Ž.

INTERNATIONAL JOURNAL OF COMPUTATIONAL METHODS AND EXPERIMENTAL MEASUREMENTS, cilt.13, 2025 (Scopus)

Özet

In today's digitalized production environments, AI-supported systems not only transform production processes, but also complicate the nature of decisions taken in these processes. Especially in smart production scenarios where edge and cloud computing infrastructures are used together, decision processes must be managed with both low-latency local data and large-volume centralized analyses. This bidirectional data flow brings about multi-criteria decision problems that cannot be easily solved with classical algorithms due to the presence of incomplete, uncertain and unstable information. This study proposes a new decision support model for such multi-criteria and uncertain decision problems that arise in computer-aided production environments. Unlike classical data analytics methods, our model is designed based on the T-Spherical Hesitant Fuzzy Rough Set (T-SHFR) theory. While T-SHFR evaluates decision alternatives in the triangle of truth, falsehood and uncertainty, it can also systematically process incomplete or contradictory data with hesitant membership and rough set logic. In this respect, the model goes beyond the artificial intelligence applications frequently found in the literature and offers a structure where uncertainty is directly modeled. In the study, this method was integrated with edge and cloud computing architectures and the multi-criteria performance of Edge-only, Cloud-only and Hybrid approaches was evaluated; scenario-based analyses were conducted on basic parameters such as production efficiency, downtime, cost and resource usage. The findings show that the T-SHFR-based model significantly increases decision quality especially in hybrid architectures and offers higher stability and flexibility in stuations where classical methods are difficult. Thus, the proposed approach offers a holistic framework that strengthens decision making under uncertainty in computer-driven production systems.