Papers
Topics
Authors
Recent
Search
2000 character limit reached

Deep Autoencoder based Energy Method for the Bending, Vibration, and Buckling Analysis of Kirchhoff Plates

Published 9 Oct 2020 in cs.LG, cs.NA, and math.NA | (2010.05698v1)

Abstract: In this paper, we present a deep autoencoder based energy method (DAEM) for the bending, vibration and buckling analysis of Kirchhoff plates. The DAEM exploits the higher order continuity of the DAEM and integrates a deep autoencoder and the minimum total potential principle in one framework yielding an unsupervised feature learning method. The DAEM is a specific type of feedforward deep neural network (DNN) and can also serve as function approximator. With robust feature extraction capacity, the DAEM can more efficiently identify patterns behind the whole energy system, such as the field variables, natural frequency and critical buckling load factor studied in this paper. The objective function is to minimize the total potential energy. The DAEM performs unsupervised learning based on random generated points inside the physical domain so that the total potential energy is minimized at all points. For vibration and buckling analysis, the loss function is constructed based on Rayleigh's principle and the fundamental frequency and the critical buckling load is extracted. A scaled hyperbolic tangent activation function for the underlying mechanical model is presented which meets the continuity requirement and alleviates the gradient vanishing/explosive problems under bending analysis. The DAEM can be easily implemented and we employed the Pytorch library and the LBFGS optimizer. A comprehensive study of the DAEM configuration is performed for several numerical examples with various geometries, load conditions, and boundary conditions.

Citations (267)

Summary

  • The paper introduces the DAEM that leverages deep autoencoders with a minimum energy principle to analyze bending, vibration, and buckling of Kirchhoff plates.
  • The method employs a mesh-free approach using Monte Carlo integration and modified activation functions to enhance numerical stability and accuracy.
  • Numerical examples in PyTorch demonstrate its efficiency in predicting natural frequencies and critical buckling loads, offering a promising alternative to traditional techniques.

Deep Autoencoder Based Energy Method for Kirchhoff Plate Analysis

The paper introduces a novel computational framework employing a deep autoencoder-based energy method (DAEM) to tackle the mechanical analysis of Kirchhoff plates. Kirchhoff plates, prevalent in engineering due to their computational efficiency and complexity, require sophisticated analytical approaches for accurate bending, vibration, and buckling analyses. The proposed DAEM integrates the features of a deep autoencoder with the principle of minimum total potential energy to provide a versatile solution absent of traditional mesh dependencies.

Technical Overview

The authors leverage the robust feature extraction capabilities of deep autoencoders to identify and approximate complex patterns in the energy system associated with Kirchhoff plates. This AI-powered approach offers an alternative to traditional numerical methods such as finite element, boundary element methods, and others. The core of the DAEM is a feedforward deep neural network architecture that supports unsupervised learning and functions as a nonlinear approximator through unsupervised feature extraction.

The DAEM's objective function centers on minimizing the total potential energy, which encompasses both strain energy and external force potential energy. For vibration and buckling issues, Rayleigh's principle serves as the foundation of loss function construction, and modified activation functions—like the scaled hyperbolic tangent—are introduced to mitigate gradient-related issues often seen in deep neural networks.

Numerical Implementation

Detailed implementation elucidated in the paper includes employing PyTorch for training the DAEM and using an LBFGS optimizer. The configuration of the deep autoencoder is tested across various numerical examples—incorporating different geometries, loading, and boundary conditions—demonstrating the method's applicability to a broad class of Kirchhoff plate problems. Notably, the method promotes a mesh-free environment by utilizing Monte Carlo integration for energy calculations.

Results and Implications

Upon implementation, the DAEM showed proficiency in resolving Kirchhoff plate challenges, accurately predicting bending deformations, natural frequencies, and critical buckling loads. Notable numerical outcomes include improved approximation accuracy and stability, establishing DAEM as not only a viable tool but an efficient and accessible method for complex mechanical analyses. The paper justifies the use of DAEM in dynamic scenarios, elucidating on its computational cost, accuracy, and flexibility compared to classical approaches.

Future Directions

The authors acknowledge the potential of further research to enhance DAEM for broader applications, such as the inclusion of geometric and material non-linearities, and adapting the method for computational fluid dynamics. The development of a more globally optimized algorithm is also considered for ongoing improvement of this AI-driven analysis technique.

In summary, this research provides a comprehensive exploration of a deep autoencoder-based method for Kirchhoff plate analysis, contributing significantly to the domain by introducing machine learning capabilities to traditional mechanical problems and delivering promising results both theoretically and practically. The groundwork laid by this study could pave the path for similar innovations in structural engineering and computational mechanics.

Paper to Video (Beta)

No one has generated a video about this paper yet.

Whiteboard

No one has generated a whiteboard explanation for this paper yet.

Open Problems

We haven't generated a list of open problems mentioned in this paper yet.

Continue Learning

We haven't generated follow-up questions for this paper yet.

Collections

Sign up for free to add this paper to one or more collections.