Simultaneous Design of Two-stage Gearboxes: An Application of Gravitational Particle Swarm Algorithm

In this paper, we propose a novel gearbox design method called multiple gearbox optimization (MGO), that can simultaneously design multiple patterns of feasible gearboxes. This MGO consists of a penalty handling method and a multimodal optimization method. The handling method converts an existing constrained gearbox design problem into an unconstrained one. The multimodal optimization method is developed by making our previously proposed gravitational particle swarm algorithm (GPSA) to solve mixed-integer optimization problems. As a result, our MGO successfully obtained an average of 14.18 design patterns in a single run that satisfied all design constraints. Statistical tests indicated that the performance of our MGO was significantly superior to some conventional methods in terms of the number of design patterns and the volume/weight of gearboxes.