A Structural Optimization Problem Formulation for Design of Compliant Gripper Using a Genetic Algorithm


This paper demonstrates the automatic design of a compliant gripper by structural optimization using genetic algorithms. Compliant mechanisms are single-piece jointless structures that use compliance (elastic deformation) as a means to achieve motion. As such, they have many advantages compared to conventional rigid-link mechanisms and so can be created as a replacement for their rigid-link counterparts, especially when the applications are in the micro-dimensional scale. Recently, relatively simple compliant mechanisms have been successfully synthesized by applying structural optimization methods because these methods automatically determine the topology and shape of structures based on any given desired structural criteria. The mechanism designed in this paper is meant to be able to grip an object and convey it from one point to another. Such a mechanism has useful applications in MEMS and various automation devices, but it is relatively complex. They are difficult to design mainly because their motion has to be analyzed by finite element methods and the relationship between their geometry and their elastic behavior is highly complex and non-linear.
The synthesis of such a mechanism is to be achieved by a structural optimization approach using a genetic algorithm as the optimizer and a special morphological representation for defining the design geometry. The problem is formulated as a discrete multiobjective constrained optimization problem. The genetic algorithm is a multiobjective algorithm with constraint handling, based on maintaining separate non-domination (Pareto) rankings for objectives and constraints satisfaction, thus enabling an intelligent selection of solutions for cooperative mating which eliminates the need to prescribe penalty function parameters commonly used for constraint handling. In addition, a recently developed morphological geometric representation scheme is used to define the topology and shape of the structure via an arrangement of skeleton and surrounding material. This technique facilitates the transmission of topological/shape characteristics across generations in the evolutionary process, and will not render any undesirable geometric features such as disconnected segments, checkerboard patterns or single-node hinge connections. A non-linear finite element program has also been used for the large-displacement analysis of the structure, and the program is integrated with the genetic algorithm to form an overall working framework for structural optimization.
Some tentative formulations of the optimization problem needed to achieve the required elastic/structural behavior of the design have been developed. The resulting geometries obtained here are clearly defined due to the discrete nature of the geometry representation, unlike in the homogenization or material density methods of topology optimization which require the prescription of some threshold point to interpret whether the resulting material density values in the elements indicate solid material or void. Optimal resulting designs have been obtained that are valid and practical for realization/fabrication.