Wednesday  |  Monongahela  |  11:10 AM–11:30 AM

#13556, Deep Neural Network to Identify Nonlinear Viscoelastic Constitutive Model

Modeling nonlinear viscoelasticity for soft solids is a challenging problem for a number of reasons. In many cases a large number of material parameters are needed to capture material response and validation of models can be hindered by fitting problems. Previously, we have developed a Gaussian Process and Singular Value Decomposition (GP&SVD) approach to determine the material parameters of a constitutive model describing the mechanical behavior of a soft, viscoelastic PVA hydrogel with four parameters, which works very well. However, GP is limited when the number of model parameters is larger than 10, where the size of the training set usually needs to be larger than 5000. To increase the scalability of our method, we replace GP with deep neural network (DNN) for constitutive models with more parameters. A PA constitutive model with 13 independent parameters is fit using our method. Tens of thousands of stress histories generated by the constitutive model constitute the training sets. The low-rank representations of stress histories by Singular Value Decomposition (SVD) are taken to be random variables which can be modeled via DNN with respect to the material parameters of the constitutive model. We obtain optimal material parameters by minimizing an objective function over the input set. Results so far show that DNN&SVD has great potential in fitting constitutive models with 10 more parameters.

Jikun Wang   Cornell University
Chung-Yuen Hui   Cornell University
Alan Zehnder   Cornell University