Evidence-Grounded Hybrid Framework for ISO/IEC 17025 Laboratory Operations Audits: A Neuro-Symbolic Decision-Support Approach

Authors

DOI:

https://doi.org/10.31098/lomr.v5i1.4258

Abstract

The accreditation of testing and calibration laboratories under ISO/IEC 17025:2017 requires systematic mapping of operational evidence to regulatory clauses, a labor-intensive process that depends on subjective auditor interpretation. As critical quality infrastructure nodes in industrial supply chains, accredited laboratories directly influence product safety, environmental monitoring, and service reliability. Although Retrieval-Augmented Generation (RAG) technologies offer partial automation, conventional RAG systems produce extrinsic hallucinations at rates ranging from 17 to 33 percent, creating unacceptable liability risks for compliance governance. This study proposes an Evidence-Grounded Hybrid Intelligence framework that enforces verifiable compliance mapping through a neuro-symbolic architecture. Using a Design Science Research methodology, the framework integrates dual retrieval methods (TF-IDF and SBERT), a verifiable RAG pipeline based on Qwen3-32B with a temperature setting of 0.1, mandatory evidence span extraction and character offset verification, a deterministic rule-based validation engine, and RDF/Turtle provenance export. Benchmarked against a conventional RAG baseline across 847 query-clause pairs derived from 50 purposively sampled Indonesian audit reports, the framework eliminated propagated hallucinations, reducing them from 14 percent to 0 percent (McNemar chi-square = 118.6, p less than 0.001). It increased evidence-grounding precision to 0.94 with an F1 score of 0.91, reduced reviewer correction rates by 37 percent, and achieved 100 percent detection of predefined non-compliance patterns. The framework contributes to laboratory operations management by delivering a scalable compliance decision support architecture that reduces audit cycle effort, standardizes verification processes, mitigates operational risk, and establishes a machine-readable provenance infrastructure for longitudinal quality assurance

References

acac

Downloads

Published

2026-05-29

Citation Check

How to Cite

Rasyid, S. A., Wiranata, A., Usman, K., Wibowo, S. A., Hantoro, G., Syamsul, M., & Prihatini, Y. I. (2026). Evidence-Grounded Hybrid Framework for ISO/IEC 17025 Laboratory Operations Audits: A Neuro-Symbolic Decision-Support Approach. Logistic and Operation Management Research (LOMR) , 5(1), 36–49. https://doi.org/10.31098/lomr.v5i1.4258