4D printing soft robots guided by machine learning and finite element models

• Published in: Sensors and actuators. A. Physical. Vol. 328; p. 112774
• Main Authors: Zolfagharian, Ali, Durran, Lorena, Gharaie, Saleh, Rolfe, Bernard, Kaynak, Akif, Bodaghi, Mahdi
• Format: Journal Article
• Language: English
• Published: Lausanne Elsevier B.V 01.09.2021; Elsevier BV
• Subjects: 4D printing; Actuation; Actuators; Finite element method; Finite element modeling; Machine learning; Mathematical models; Pneumatics; Printing; Robots; Soft pneumatic actuators; Soft robotics; Training; Soft robotics; Soft pneumatic actuators; 4D printing; Finite element modeling; Machine learning
• ISSN: 0924-4247; 1873-3069
• DOI: 10.1016/j.sna.2021.112774

[Display omitted] •A machine learning approach is introduced to estimate 4D-printed soft robot bending behavior.•A finite element model is developed to reflect the nonlinear behavior of the pneumatic soft actuators.•The 4D-printed soft actuator geometrical requirements are obtained via machine learning for specific tasks.•The experimental tests validate the efficacy of the presented machine learning model for 4D printing soft robots. This paper presents a method for four-dimensional (4D) printing of soft pneumatic actuator robot (SPA)s, using nonlinear machine learning (ML) and finite element model (FEM). A FEM is developed to accurately simulate experimental actuation to obtain training data for the ML modeling. More than a thousand data training samples from the hyperelastic material FEM model generated to use as training data for the ML model, which was developed to predict the geometrical requirements of the 4D-printed SPA to realize the bending required for specific tasks. The ML model accurately predicted FEM and experimental data and proved to be a viable solution for 4D printing of soft robots and dynamic structures. This work helps to understand how to develop geometrical soft robots’ designs for nonlinear 4D printing problems using ML and FEM.