Optimum Design of a Spur Gear Using a Two Level Optimization Approach

In this paper, we present a two level optimization approach in order to enhance the design process of a one-stage speed reducer. The proposed design methodology is performed using genetic algorithms which are judiciously combined with the use of: i) analytical models (1st level) and ii) Finite Element Method (FEM) based models (2nd level), to evaluate design candidates. Indeed, the use of CAD-CAE tools to develop higher fidelity FEM models allows to re-evaluate the attained first level designs, while accounting for new design parameters and advanced aspects which have been ignored in the first level. In order to minimize the computational burden, a metamodel based optimization technique is adopted at this second level. To illustrate the efficiency of the proposed approach, a case study of a spur gear based reducer is presented where the design of experiments is built using Hypercube Latin Sampling and surrogate models are constructed using Radial Basic Functions.