A geometric programming approach to the optimization of mechatronic systems in early design stages

This paper evaluates geometric programming as a solver to optimize mechatronic system design in a holistic manner to aid early design decisions. Mechatronic systems design optimization requires complex and often non-convex functions as design objectives and constraints. Currently the solutions are primarily based on randomized search methods, e.g., genetic algorithms, and they are time-consuming. This paper converts complex constraints and objectives into approximate posynomial forms, which can then be used with disciplined convex optimization to significantly reduce the computation time for optimization. The approach is compared to the previous research using a mechatronic servo system design case study consisting of a motor, a shaft, two planetary gears and a rotational load. The result confirms that the geometric programming approach improves both computation speed and accuracy.