Opening the Black Box: An Explainable, Few-shot AI4E Framework Informed by Physics and Expert Knowledge for Materials Engineering
🤖 AI Summary
Addressing the dual bottlenecks of scarce high-quality experimental data and poor interpretability in AI models for materials engineering, this study proposes a mechanism-driven, few-shot interpretable AI framework. The method embeds physical constraints and domain expertise directly into the model architecture, synergistically integrating three-stage synthetic data augmentation, symbolic regression, and a hybrid global-local nested optimization strategy combining differential evolution with local refinement. This enables the construction of constitutive equations that are both highly accurate and intrinsically interpretable. Using only 32 experimental samples, the framework achieves an 88% accuracy in crack susceptibility prediction—the first quantitative elucidation of thermo-mechanical-metallurgical coupled cracking mechanisms. It significantly enhances process understanding, model generalizability across material systems, and reliability of synthetic data generation. The approach establishes a verifiable, transferable paradigm for AI-for-Engineering (AI4E) in safety-critical applications.
Application Category
📝 Abstract
The industrial adoption of Artificial Intelligence for Engineering (AI4E) faces two fundamental bottlenecks: scarce high-quality data and the lack of interpretability in black-box models-particularly critical in safety-sensitive sectors like aerospace. We present an explainable, few-shot AI4E framework that is systematically informed by physics and expert knowledge throughout its architecture. Starting from only 32 experimental samples in an aerial K439B superalloy castings repair welding case, we first augment physically plausible synthetic data through a three-stage protocol: differentiated noise injection calibrated to process variabilities, enforcement of hard physical constraints, and preservation of inter-parameter relationships. We then employ a nested optimization strategy for constitutive model discovery, where symbolic regression explores equation structures while differential evolution optimizes parameters, followed by intensive parameter refinement using hybrid global-local optimization. The resulting interpretable constitutive equation achieves 88% accuracy in predicting hot-cracking tendency. This equation not only provides quantitative predictions but also delivers explicit physical insight, revealing how thermal, geometric, and metallurgical mechanisms couple to drive cracking-thereby advancing engineers' cognitive understanding of the process. Furthermore, the constitutive equation serves as a multi-functional tool for process optimization and high-fidelity virtual data generation, enabling accuracy improvements in other data-driven models. Our approach provides a general blueprint for developing trustworthy AI systems that embed engineering domain knowledge directly into their architecture, enabling reliable adoption in high-stakes industrial applications where data is limited but physical understanding is available.
Problem
Research questions and friction points this paper is trying to address.
Develops explainable AI framework for materials engineering with limited experimental data
Addresses black-box model interpretability issues in safety-critical industrial applications
Integrates physics and expert knowledge to discover constitutive equations for material behavior
Innovation
Methods, ideas, or system contributions that make the work stand out.
Physics-informed synthetic data augmentation from few samples
Nested optimization for interpretable constitutive equation discovery
Hybrid global-local refinement enabling multi-functional predictive tool
🔎 Similar Papers
2024-01-26arXiv.orgCitations: 5
2024-06-01NeurocomputingCitations: 8
