Abstract:This paper presents a novel framework for predicting computer-aided engineering (CAE) simulation results using <a class="topic-link" href="/topics/computer-science/machine-learning">machine learning</a> (ML). The framework is applied to finite element (FE) simulations of <a class="topic-link" href="/topics/engineering/axial-dynamic">dynamic axial</a> crushing of rectangular crush tubes that are typically used in vehicle <a class="topic-link" href="/topics/engineering/crashworthiness">crashworthiness</a> applications. A virtual <a class="topic-link" href="/topics/engineering/design-of-experiments">design of experiments</a> that varies the size and wall thickness of the FE model is performed to generate the necessary training data. This process generates designs with varying numbers of nodes and elements that are handled by the <a class="topic-link" href="/topics/computer-science/machine-learning-system">ML system</a>. However, the explicit design parameters and meshing techniques that were used to generate the training data remain unknown to the ML system. Instead, 3D convolutional <a class="topic-link" href="/topics/computer-science/neural-networks">neural networks</a> (CNN) <a class="topic-link" href="/topics/engineering/autoencoder">autoencoders</a> are used to process the initial FE model data (i.e., nodes, elements, thickness, etc.) to automatically determine these features in an <a class="topic-link" href="/topics/computer-science/unsupervised-manner">unsupervised manner</a>. A voxelization strategy that operates on the mass of individual nodes is proposed to handle the unstructured nature of the nodes and elements while capturing variations in the wall thickness of the FE models. The flattened latent space generated by the 3D-CNN-autoencoder is then used as input into long-short term memory neural networks (LSTM-NN) to predict the force–displacement response as well as the deformation of the mesh. The training process of both the 3D-CNN-autoencoders and LSTM-NN is systematically studied to highlight the robustness of the framework. The proposed ML system utilizes only 16% of the simulations generated in the virtual design of experiments to achieve good <a class="topic-link" href="/topics/engineering/predictive-capability">predictive capability</a>. Once trained, the proposed framework can predict the deformation of the mesh and resulting force–displacement response of a new design up to <math>∼</math>330 and <math>∼</math>2,960,000 times faster, respectively, than the conventional FE approach with good accuracy. This computational speed up offers design engineers and scientists a potential tool for accelerating the design exploration process with CAE simulation tools, such as <a class="topic-link" href="/topics/engineering/finite-element-analysis">FE analysis</a>.

A machine learning framework for accelerating the design process using CAE simulations: An application to finite element analysis in structural crashworthiness

A machine learning framework for accelerating the design process using CAE simulations: An application to finite element analysis in structural crashworthiness

Machine learning enabled fusion of CAE data and test data for vehicle crashworthiness performance evaluation by analysis

Data-driven Design of Well-Behaved Nonlinear Structures: A Case Study on the Von Mises Truss

A Novel Machine-Learning Based Framework for Calibrating Micromechanical Fracture Model of Ultra-Low Cycle Fatigue in Steel Structures

Finite Element Network Analysis: A Machine Learning based Computational Framework for the Simulation of Physical Systems

Machine-Learning-Accelerated Simulations for the Design of Airbag Constrained by Obstacles at Rest

A Crashworthiness Design Framework Based on Temporal-Spatial Feature Extraction and Multi-Target Sequential Modeling

A machine learning assisted multifidelity modelling methodology to predict 3D stresses in the vicinity of design features in composite structures

Artificial intelligence (AI)-assisted simulation-driven earthquake-resistant design framework: Taking a strong back system as an example

Bridging finite element and deep learning: High-resolution stress distribution prediction in structural components

A Problem-Independent Machine Learning (PIML) enhanced substructure-based approach for large-scale structural analysis and topology optimization of linear elastic structures

Fast fluid–structure interaction simulation method based on deep learning flow field modeling

Machine learning-based crashworthiness optimization for the square cone energy-absorbing structure of the subway vehicle

Machine learning models for predicting axial compressive capacity of circular CFDST columns

Accelerating Analysis for Structure Design via Deep Learning Surrogate Models

Multi‐parameter design optimization of crash box for crashworthiness analysis

Using successive response surface methodology

Application of machine learning for 'what if?' stress analysis

Introducing Finite Element Method Integrated Networks (FEMIN)

A Novel Geometric Nonlinear Reduced Order Modeling Method Using Multi-Fidelity Surrogate for Real-Time Structural Analysis