AI based decision-making system for tooling design of aircraft product assembly with developed knowledge retrieval...

Area of Science:

• Aerospace Engineering
• Knowledge Management
• Artificial Intelligence

Background:

• Aircraft wing-spar tooling design is complex, requiring efficient access to extensive technical information.
• Traditional information retrieval methods often fall short in providing decision-ready guidance for intricate design tasks.
• Reducing search effort and improving traceability are critical for effective product development.

Purpose of the Study:

• To develop an ontology-based Knowledge Retrieval Practices (KRP) system for aircraft wing-spar tooling design.
• To enhance decision-making by reducing search effort, improving traceability, and providing decision-ready guidance.
• To integrate multiple retrieval modes for comprehensive knowledge access.

Main Methods:

• Formalized a domain ontology and rule-based constraints for tooling design.
• Developed a query-information model mapping natural language to machine-interpretable intents.
• Orchestrated three retrieval modes: ontology-based semantic (OBS), rule-based inference (RBI), and case-based reasoning instances (CBRI).
• Integrated the system within a five-layer platform with Product Development Management (PDM)/Computer Aided Design (CAD) compatibility.

Main Results:

• Achieved a mean task-level accuracy of 93.1%, with a hybrid OBS+RBI+CBRI configuration reaching 96.9% confidence.
• Document-level accuracy ranged from 98-99% on graded tooling corpora.
• Demonstrated consistent gains over traditional retrieval methods in a domain-agnostic stress test.
• Identified specific strengths for each retrieval mode: OBS for conceptual queries, RBI for computable aspects, and CBRI for structural analogies.

Conclusions:

• The developed KRP platform offers a deployable solution for complex assembly tasks, providing explainable and decision-ready guidance.
• Actionable guidance is provided on selecting appropriate retrieval strategies for different query types.
• The study establishes a reproducible evaluation protocol, paving the way for future aerospace knowledge management research and benchmarks.