Generative Design Procedure for Embedding Specified Planar Behavior in Modular Soft Pneumatic Actuators.

Abstract

A modular actuator construction consisting of smaller articulating units in series was designed to construct soft pneumatic actuators. These units are constructed to have a preferential bending direction using Mold Star 15 and Smooth-Sil 950 silicone. By changing the orientation of each unit, a different deformed actuator shape can be achieved. A design tool was developed where a genetic algorithm was coupled with a nonlinear finite element solver. This design tool optimizes the design using the genetic information available in the initial population over multiple generations and presents a candidate design that best resembles a desired profile specified as the objective function. A two-dimensional reduced order model was developed that reduces the time for each function evaluation from ≈20 min for a three-dimensional (3D) numerical analysis to ≈45 s. The design tool was tasked to solve design targets ranging from sine and cosine functions of various amplitudes to final actuator tip positions. In each case, the inflated actuator resembled the desired profile. Selected physical actuators were cast and tested. 3D scanning was used to capture the inflated shape and compared it to the numerical solution. A quantitative comparison showed a maximum average deviation of <2.5% of the uninflated actuator length between the physical and numerical models. The proposed design tool proved successful in designing shape matching actuators with close agreement to physical models.