Application of Explainable Artificial Intelligence (XAI) for Automated Design Verification and Quality Assurance in Complex Engineering Products
Abstract
The rapid evolution of modern engineering industries—spanning aerospace, automotive, precision robotics, and turbomachinery—has introduced unprecedented structural and geometric complexity into engineered assemblies. Traditional design verification and quality assurance (QA) protocols rely heavily on manual engineering review, static rule checking, and computationally intensive Finite Element Analysis (FEA) or Computational Fluid Dynamics (CFD) simulations. Although deep learning surrogate models have recently been deployed to accelerate design iterations, their industrial adoption remains severely constrained by their opaque 'black-box' nature. Engineering teams cannot validate or audit predictions when underlying failure mechanisms and design rule violations are not transparently explained. This paper presents a comprehensive review and technical investigation into the application of Explainable Artificial Intelligence (XAI) paradigms— including SHapley Additive exPlanations (SHAP), Local Interpretable Model agnostic Explanations (LIME), Integrated Gradients, and Layer-wise Relevance Propagation (LRP)—for automated design verification and quality assurance in complex products. We evaluate hybrid physics-informed neural networks (PINNs) and Graph Neural Networks (GNNs) integrated with post hoc XAI mechanisms to verify 3D CAD assemblies and multi-physics design constraints. Numerical benchmark evaluations demonstrate that XAI-guided frameworks reduce automated design verification cycle times by up to 88.4% compared to full FEM simulations, while achieving a defect identification accuracy of 96.8% and providing actionable engineering attribution maps. Furthermore, we outline key deployment challenges, trust metrics, limitations,and future research pathways for interpretability in cyber-physical product lifecycles.
KEYWORDS: Explainable Artificial Intelligence (XAI), Automated Design Verification, Quality Assurance, Physics-Informed Neural Networks, Feature Attribution, Structural Integrity, Computer-Aided Design (CAD).
KEYWORDS: Explainable Artificial Intelligence (XAI), Automated Design Verification, Quality Assurance, Physics-Informed Neural Networks, Feature Attribution, Structural Integrity, Computer-Aided Design (CAD).
Full Text:
PDF 69-81Refbacks
- There are currently no refbacks.