Applications of scientific machine learning for the analysis of functionally graded porous beams

2024-08-04

Neurocomputing

AI Summary

To address the modeling challenge of mechanical behavior in functionally graded porous beams arising from spatially varying pore sizes, this work proposes a scientific machine learning framework integrating physics-informed constraints with data-driven learning. We pioneer the incorporation of physics-informed neural networks (PINNs) into the constitutive modeling of gradient materials, augmented by a gradient-enhanced loss function and non-uniform B-spline parameterization. This enables end-to-end, boundary-condition-consistent prediction with cross-scale parametric sensitivity analysis. Across diverse pore distribution patterns, the method achieves displacement and stress field prediction errors below 3.2%, while accelerating computation by two orders of magnitude relative to conventional finite element analysis. The model exhibits strong interpretability, supporting real-time parameter inversion and structural optimization. This work establishes a novel paradigm for efficient and high-fidelity modeling of porous functional materials.

Technology Category

• Machine Learning: Structured Learning
• Constraint Satisfaction and Optimization: Constraint Learning and Acquisition
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Web Mining and Content Analysis: Machine learning and data science for the Web

Problem

• Machine Learning
• Porous Beam
• Hole Size Effect

Innovation

• Machine Learning
• Physics-Informed Neural Networks (PINN)
• Data-driven Neural Operators