Application of Machine Learning to Maximize Power Density of Propulsion Motors for VTOL Applications

Various measures are taken to maximize the power density of electrical machines for aircraft propulsion applications as minimizing the weight penalty is a key to commercializing such technologies. Material selections and setting appropriate electrical and magnetic loading are critical to achieve an optimum balance between power density and other performance indicators. Electromagnetic analysis combined with detailed drive system simulations show that severe nonlinearities in electrical machines can lead to control instabilities. This issue needs to be overcome for further improvements in power density of electric propulsion systems. Simple machine learning solutions can be applied to identify the nonlinear behavior of electrical machines with significant magnetic saturation. The approach involves a combination of off-line learning based on data produced by finite element analysis tools and on-line learning using practical data captured during equipment calibration procedures in the laboratory. The optimal type of learning algorithm can be selected based on the specific characteristics of the motor design.