Parameter Determination for Hyper-elastic Material based on a BLSTM Neural Network with Optimized Condition Numbers

Parameters determination is crucial in applications involving hyper-elastic materials. In this paper, a novel method based on a bidirectional long short-term memory (BLSTM) neural network is proposed to learn the relationship between the indentation load-depth data and the hyper-elastic material parameters. Considering the influence of data errors on prediction accuracy, condition number in this inverse problem is analyzed and optimized. In addition, the input data are assigned different weights according to their condition numbers in order to bring more attention on the reliable data. Our network has been tested by using a finite element simulation dataset and has shown excellent effects results for identifying hyper-elastic material parameters.