A novel reliability-aware performance assessment for photonic ML analyzers using Frank norm

The reliability and consistent performance of photonic machine-learning form analyzers have proven to be a major challenge, because credibility-dependent measurements, optical noise, dynamic signal variation, and performance evaluation by experts have become critical issues. This research paper introduces a reliability-conscious multi-criteria decision model conducted in an Intuitionistic Fuzzy Frank Z-Number ( \(\textrm{IFFZN}\) ) environment, in which each evaluation is represented as \(\Omega = \{(\mu , r_\mu ), (\nu , r_\nu )\}\) , an independent encoding of truth-membership, falsity-membership, hesitation, and their associated reliability degrees. Compared with traditional fuzzy and intuitionistic systems, the semantic representation of performance behavior and reliability is two-fold in IFFZN , avoiding information distortion during aggregation. The main innovation is that reliability is decoupled from the main evaluation data without losing its interactive effect via Frank operational laws, allowing effective decision-making under incomplete, imprecise, and credibility-variant observations intrinsic to intelligent photonic analytical systems. Six Frank norm-based aggregation operators are developed: \(\textrm{IFFZNWA}\) , \(\textrm{IFFZNWG}\) , \(\textrm{IFFZNOWA}\) , \(\textrm{IFFZNOWG}\) , \(\textrm{IFFZNHWA}\) , and \(\textrm{IFFZNHWG}\) , with mathematical properties of idempotency, monotonicity, and boundedness formally established. The framework is applied to a synthetic case study evaluating four photonic ML analyzer alternatives across four reliability-oriented attributes, assessed by three domain experts using synthetic sensor data calibrated to photonic ML scenarios. This results in a more realistic representation of the cognitive behavior of domain experts and produces rankings that remain consistent under confidence perturbations and heterogeneous data conditions. A comparative investigation with classical WASPAS shows deterministic schemes cannot encode credibility-aware uncertainty and inter-criteria coupling essential for high-precision photonic evaluation. The outcomes provide a scalable, analytically robust reliability assessment protocol for next-generation photonic ML form analyzers and a transferable uncertainty-conscious decision infrastructure for other smart optical and AI-based diagnostic environments requiring reliable performance measurement. This work establishes a new research direction in confidence- preserving computational intelligence, advancing the theoretical basis of \(\textrm{IFFZN}\) -based multi-criteria decision-making by combining reliability semantics and advanced aggregation theory. The numerical validation is conducted on synthetic data; hardware validation with real photonic ML systems is identified as essential future work.