Application of Artificial Intelligence to Solve an Elasto-Plastic Impact Problem 2020-01-0734
Artificial intelligence (AI) dramatically changing multiple industries. AI’s potential to transform Computer Aided Engineering (CAE) can not be overlooked. Conventionally, Finite Element Analysis (FEA) is the simulation of any given physical phenomenon using the numerical technique to obtain an approximate solution to a class of problems governed by partial differential equations. Implementation of AI methods in this area combine human intelligence with numerical solutions to make them more efficient.
The paper attempt to develop a Machine Learning (ML) model to solve an elasto-plastic impact problem. A symmetric short crush tube made of three materials impacted by a moving wall. A learning database to train and validate the model established using finite element simulations. A structured dataset file prepared from CAE simulations to test and validate the ML model. Tube size, gauge and elasto-plastic material properties used as input attributes or features. Effective plastic displacement was the target label to predict. The dataset analysed to understand relations among the features. Multiple ML models experimented and the hyper parameters tuned to improve model fit. The trained model evaluated with an unseen dataset to assess its prediction accuracy in real-world settings. The output of deep learning model compared to the results from a finite element simulation. The established deep learning model shows promising accuracy.
A glossary of basic terms and concepts presented to comprehend the readers with AI. The paper briefs an overall process to set up a deep learning model using TensorFlow and Python libraries. Finally, potential application areas to consider for AI implementation in CAE discussed along with expected practical challenges.