Affiliations: [a] College of Mechanical and Electrical Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing, China | [b] China Aerospace Science and Industry Corporation Limited, Institute of Magnetic Levitation and Electromagnetic Propulsion, Beijing, China
Correspondence:
[*]
Corresponding author: Jin Zhou, College of Mechanical and Electrical Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing, China. Tel./Fax: +86 025 84892905; E-mail: meejzhou@nuaa.edu.cn
Abstract: The high speed magnetic levitation turbine generator used in Organic Rankine cycle (ORC) system will produce wind friction loss due to the high density of work environment and high eddy current loss in the part of permanent magnet (PM) rotor, which obviously effect the performance and stability of PM. In order to minimize the loss of PM rotor, this paper presented a comprehensive optimization design for the stator and rotor structure of turbine-generator. Firstly, the physical performances coupled with design variables, such as loss, generator power, rotor dynamic, strength and axial force of AMB were analyzed, mainly analyzed the structural variables coupled with design variables due to the effects of single constraint and the decoupling methods, which eliminated irrelevant variables in the final optimization. Neural network was used to establish agent models for each physical performance and brought into comprehensive optimization process of total PM rotor, which greatly improved the efficiency of optimization. The result of optimization is that the total rotor loss decreased 15.3% and all of other physical performance meet the requirements.
