Comparison of rotor arrangements of Transverse Flux Machines for a robotic direct drive optimized by genetic algorithm and Regression Tree Method

Abstract

Articulated robotics applications typically have a demand for high torque at low speed. However, conventional electrical machines cannot generate a reasonable amount of torque directly by electro-magnetics. Therefore, gearboxes are used to convert speed and torque, accepting loss of mechanical power due to additional friction. Although geared solutions for robotic drive trains already offer exceedingly high torque densities, they are limited by the drawbacks of high reduction gears, such as non-linearities in friction, complex flexibility effects, and limited service life of mechanics in contrary to direct drive solutions. The Transverse Flux Machine with the high gravimetric torque density may be a solution for reducing or eliminating the need for a gearbox. Using a genetic algorithm, the proposed Transverse Flux Machines are optimized. To enhance the optimization’s speed, the machines’ calculations done by Finite-Element-Analysis of selected generations are replaced by a Regression Tree Model whose results are verified after a defined expired model service life with a subsequent adjustment of the model. The eligibility of different arrangements the Transverse Flux Machines’ rotor are compared regarding the application as low-speed direct drive in robotics, also compared to similar Radial Flux Machines. The optimized Transverse Flux Machines have a higher efficiency due to lower copper loss and a higher active gravimetric torque density. However, the Radial Flux Machines have higher total torques and power factors.