Machine Learning Accelerated Topology Optimization and its Applications in Pneumatic Soft Actuator Design

Abstract

Structural optimization is a popular tool for designing smart structures for a given set of conditions, constraints, and objectives. This can involve selecting the optimal material, shape, size, and distribution of the structural elements to achieve the desired performance while minimizing weight, cost, or other factors. The structural optimization is often an iterative process, which can be costly in computational time, in particular when designing pressure-driven soft actuators, involving complex and large size physical problems. This thesis aims to propose, prove, and validate a novel framework to implement Machine Learning (ML) techniques to accelerate structural optimization when solving a wide range of design problems, including topology optimization problems, topology optimization problems with design-dependent loading, reliability-based topology optimization problems, and the design and development of soft pressure-driven actuators.