Mechanical characterization of heterogeneous hyperelastic membrane using inverse methods

Abstract

Many soft biological tissues are heterogeneous, having different properties at different locations. Characterizing these tissues is very important for virtually testing potential medical technologies or protocols. Some synthetic thin structures used in various manufacturing processes are also complex in material composition. Determination of the mechanical properties of these structures is critical for industrial production. These materials can undergo very large deformation in actual applications and their behavior is nonlinear. This makes their characterization difficult and there is a need for an efficient method. In this study, hyperelastic material properties of a heterogeneous synthetic flat membrane with two constituent materials are determined using two inverse methods. One method is the traditional Finite Element Model Updating Method and the other is based on machine learning using a deep neural network. Inverse modeling was done in moderate strain range (engineering strain up to 37 %) with the Neo-Hookean material model and in the large strain (engineering strain up to 93%) range using the Yeoh model. Both the inverse methods were found to have very good accuracy. Accuracy in the moderate strain range was slightly higher than that of the high strain range. For both strain range, the coefficient of determination values were very close to 1 for both the stress-strain curves and the work-energy curves, which indicates very good accuracy. The machine learning method was six orders of magnitude faster than the Finite Element Model Updating Method.