University of British Columbia
Finite element simulations of welding distortions and predictions using deep Learning methods
Abstract
dc:descriptionFinite element (FE) simulations, while effective, are computationally expensive, particularly for dynamic problems. Neural networks (NNs) offer a more efficient alternative. This study utilizes NNs to predict welding-induced distortions in a T-joint—a critical component in engineering structures due to its enhanced load-bearing capacity. Welding often causes distortions that impact the structural integrity and dimensional accuracy of these components. Conventional methods for controlling such distortions are costly, making machine learning (ML) an effective solution for reducing computational demands. Additionally, optimization tasks frequently require extensive analyses, highlighting the need for a surrogate model. This study develops and implements such a model, offering an efficient framework for optimizing welding processes while maintaining accuracy. In this study, two NNs—a multilayer perceptron (MLP) and a convolutional neural network (CNN)—are employed to predict residual distortions. Two case studies are conducted for each model to investigate the effects of variations in geometry and welding sequences. FE simulations of the gas metal arc welding (GMAW) process are analysed to assess how welding order and direction influence residual distortions. The results demonstrate that selecting an optimal welding sequence and direction can reduce distortions by up to 65%. These FE simulations are then used to train the neural networks. The MLP and CNN models are carefully designed and optimized through architectural and hyperparameter tuning to achieve highly accurate distortion predictions. The findings show that both models are effective; however, the CNN achieves higher accuracy for complex distortion patterns, underscoring its suitability for more intricate scenarios.
Degree
thesis:*- Name thesis:degree_name
- Master of Applied Science - MASc
- Level thesis:degree_level
- master's
- Discipline thesis:degree_discipline
- Mechanical Engineering
- Grantor dc:publisher
- University of British Columbia
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Karimi, Mahdi
- Contributors dc:contributor
-
- Jelovica, Jasmin
Rights
dc:rights- Statement dc:rights
-
- Attribution-NonCommercial-NoDerivatives 4.0 International
- Language dc:language
- eng
Identifiers
dc:identifier.*- Handle dc:identifier
- http://hdl.handle.net/2429/89947