Speaker
Description
Quality certification remains one of the most knowledge-intensive and time-consuming activities in the industrial forging sector, regardless of the manufacturing route, including open-die forging, closed-die forging, ring rolling, upset forging and precision forging. In many companies, certification still depends heavily on experienced engineers who manually collect production records, interpret customer specifications, verify laboratory and inspection results, and resolve inconsistencies across multiple technical documents. Although production processes differ significantly, certification invariably requires demonstrating compliance among production data, customer requirements and international standards, while maintaining full traceability of materials, heat treatments, mechanical testing and non-destructive examinations.
This paper presents a hybrid Generative-Neurosymbolic Artificial Intelligence framework for automating both the preparation and the engineering verification of quality certificates. The proposed architecture combines Large Language Models, used to extract and normalize information from heterogeneous engineering documents, with a symbolic reasoning engine based on a forging-domain knowledge graph. The symbolic layer performs deterministic compliance checks against applicable standards, customer-specific requirements and engineering rules, thereby producing explainable and auditable decisions.
Unlike conventional document-automation solutions, the framework introduces a dual verification mechanism. First,it assesses whether customer specifications are technically consistent with the referenced standards before manufacturing begins,enabling early identification of conflicting, incomplete or infeasible contractual requirements. Second, it verifies that production and inspection data satisfy every applicable requirement, including derived properties, process constraints, material characteristics and test results, before authorizing certificate generation.
The proposed approach transforms certification from a manual document-preparation task into a formal engineering verification process. By combining the linguistic capabilities of genAI with deterministic symbolic reasoning,it can reduce engineering effort,improve consistency and traceability,minimize human error,and support explainable, audit-ready certification in safety-critical manufacturing environments. Although developed for forging,the architecture provides a practical pathway for introducing trustworthy AI into other hot-deformation processes where compliance with technical standards, rather than document generation alone,represents the true industrial challenge.
| Speaker Company/University | Giacomo Bottoli |
|---|