4 - 4D printing modeling via machine learning

Abstract

This chapter presents a case study guiding to the development of machine learning (ML) models for predicting the behavior of nonlinear four-dimensional (4D)-printing problems. The aim of this study is to develop a ML model to predict the bending angle of a 4D-printed soft pneumatic actuator (SPA) and to investigate the influence of the input variables on its bending. This includes the development of a finite element model (FEM) to accurately simulate experimental actuation and conduct a series of simulations in order to obtain training data for the ML modeling. First, the effectiveness of using a linear model to estimate the bending angle of a hyperelastic material, through the use of an analytical model, is analyzed. Then, a FEM method using an Ogden 3-parameter hyperelastic material model in ANSYS simulation was developed to accurately model the deflection of the actuator in response to the input pressure for variable geometries, including the bellow shape, height, width, and bottom-layer thickness. More than a thousand data training samples from the FEM simulations generated to use as training data for the ML model. The ML model is developed to predict the bending angle of the 4D-printed SPA. The ML model accurately predicted FEM and experimental data and proved to be a viable solution for 4D printing modeling of soft robots and dynamic structures. This work helps to understand how to best develop models for nonlinear problems using ML methods.