Physics-Informed Kolmogorov-Arnold Networks for multi-material elasticity problems in electronic packaging
Preprint
Abstract
This paper proposes a Physics-Informed Kolmogorov-Arnold Network (PIKAN) method for analyzing elasticity problems in electronic packaging multi-material structures. The core innovation lies in replacing Multi-Layer Perceptrons (MLPs) with Kolmogorov-Arnold Networks (KANs) within the energy-based Physics-Informed Neural Networks (PINNs) framework. The method constructs admissible displacement fields that automatically satisfy essential boundary conditions and employs various numerical integration schemes to compute loss functions for network optimization. Unlike traditional PINNs that require domain decomposition and penalty terms for multi-material problems, KANs' trainable B-spline activation functions provide inherent piecewise function characteristics that naturally accommodate material property discontinuities. Consequently, this approach requires only a single KAN to achieve accurate approximation across the entire computational domain without subdomain partitioning and interface continuity constraints. Numerical validation demonstrates PIKAN's accuracy and robustness for multi-material elasticity problems. The method maintains high accuracy while significantly reducing computational complexity compared to domain decomposition approaches. Results confirm PIKAN's unique advantages in solving multi-material problems and its significant potential for electronic packaging structure analysis. Source codes are available at https://github.com/yanpeng-gong/PIKAN-MultiMaterial.Keywords
All Related Versions
This publication has 34 references indexed in Scilit:
- An engineered solution to multi-physics of insulated gate bipolar transistor module considering electrical-thermal-mechanical coupling effectAdvances in Engineering Software, 2022
- Machine Learning‐Based Solution for Thermomechanical Analysis of MMIC PackagingAdvanced Materials Technologies, 2022
- Scientific Machine Learning Through Physics–Informed Neural Networks: Where we are and What’s NextJournal of Scientific Computing, 2022
- An isogeometric boundary element method for heat transfer problems of multiscale structures in electronic packaging with arbitrary heat sourcesApplied Mathematical Modelling, 2022
- A physics-guided neural network framework for elastic plates: Comparison of governing equations-based and energy-based approachesComputer Methods in Applied Mechanics and Engineering, 2021
- An energy approach to the solution of partial differential equations in computational mechanics via machine learning: Concepts, implementation and applicationsComputer Methods in Applied Mechanics and Engineering, 2020
- A parallel FFT‐accelerated layered‐medium integral‐equation solver for electronic packagesInternational Journal of Numerical Modelling: Electronic Networks, Devices and Fields, 2019
- Recent advances in nano-materials for packaging of electronic devicesJournal of Materials Science: Materials in Electronics, 2019
- The Deep Ritz Method: A Deep Learning-Based Numerical Algorithm for Solving Variational ProblemsCommunications in Mathematics and Statistics, 2018
- Multilayer feedforward networks are universal approximatorsNeural Networks, 1989