Automated type and dimensional synthesis of planar mechanisms using numeric optimization with genetic algorithms

Mechanism type synthesis, or topology optimization, is hampered by the lack of established, well-known design rules, and designers cannot grasp the space of possible designs and the impact of all design variables on a mechanism's performance. Realistically, a human can only design and evaluate several candidate mechanisms, though there may be hundreds of competitive designs that should be investigated. In contrast, an automated approach to mechanism type synthesis can create thousands of designs and measure the performance of each one. A new methodology for automated type and dimensional synthesis of planar mechanisms with revolute joints is presented. The methodology has the following features: (i) the mechanism topology is explicitly included as one design variable within a numeric optimization framework; (ii) the complex kinematic models are formulated by using an effective multibody system modeling technique; (iii) with the genetic algorithms as a synthesizer and a local search method as a kinematic solver, both discrete topological variables and continuous parameters are optimized simultaneously; (iv) the synthesizer accounts for the significance of different metrics through a requirement prioritization scheme; (v) high-performance computing techniques are easily embedded into this methodology. The proposed methodology has been applied to four mechanism design problems: one topology-based optimization and three kinematic synthesis problems. Numerical experiments have shown its applicability to general mechanism kinematic synthesis problems. With moderate effort, this methodology can be extended to tackle more complex dynamic synthesis problems due to the flexible and extensible software architecture. The author's methodology is different from other work in the literature in a fundamental way: it uses numeric optimization rather than domain-specific rules. The optimization-based approach allows a designer to explore an area for which design heuristics are difficult or impossible to establish. It is the first fully automated approach to solving a genuine mechanism type synthesis problem using a numeric optimization framework. Thus, this methodology constitutes a significant and original contribution to the field of mechanism and machine theory.