Repository of rainbow deliverable for Pierre, in this work we use Convolutional Neural Network surrogate model in place of finite element solver to predict the deformations of a cantilever beam (4X1X1m dimensions)
Repository of rainbow deliverable for Pierre, in this work we use Convolutional Neural Network surrogate model in place of finite element solver to predict the deformations of a cantilever beam (4X1X1m dimensions).
A cantiliver beam is clamped at one end and is applied force on any node of the upper face. In this way training data is generated by applying 100 random forces on each node of the top face. Generated data is then eventually trained using deep learning libraries. Since the size of the training data file is huge it is not uploaded in the repository. Trained nerual network is then used to predict the deformations of test set, one of the test set result has been displayed in the script folder where we compare deformed mesh of finite element solution and nerual network solution respectively.
