Design parameter sensitivity analysis and performance prediction of a novel direct drive double dumbbell flux switching linear generator

Pooja Khatri,Xu Wang

DOI: https://doi.org/10.1007/s40722-021-00217-8

2021-11-15

Journal of Ocean Engineering and Marine Energy

Abstract:In this paper, a novel single-phase double dumbbell flux-switching linear generator (DDFSLG) has been proposed for ocean wave energy conversion. The novelty lies in the structure with a long permanent magnet inserted inside the hollow steel dumbbell and the copper coil being wound around the outside of the hollow dumbbell, which hinders the direct interaction between inner stator slots and permanent magnet edges and reduces the cogging force. The originality is the design and optimization method through a hybrid of electromagnetic finite element modelling and machine learning methods. The design has been simulated and analyzed using ANSYS electromagnetics finite element analysis (FEA) software. The output performance of DDFSLG has been compared with that of the existing surface-mounted linear vernier hybrid machine (SMLVHM). To validate the ANSYS FEA software for simulation of the novel design, a single-phase conventional linear permanent magnet generator (CLPMG) is also designed and analyzed by the ANSYS FEA and MATLAB Simulink software. The simulation model of ANSYS FEA software has been validated by the analytical model of the MATLAB Simulink software. The results of the MATLAB Simulink model are in concordance with those of the ANSYS FEA model. A regression relationship between input design parameters and target peak power output of DDFSLG has been established to predict the peak power output from the design parameters. The sensitivity of the design parameters and their interactions have been analyzed through the response surface method. The analysis of the variance method is used to validate the prediction model. Genetic algorithm has been applied to further optimize the design parameters of the stator and magnet configuration for the maximum peak output power. The optimal results of the prediction model have been verified. The response surface method plus the genetic algorithm has been proved to be a useful tool for optimizing the design of the linear tubular generator. Results show that the optimized design has increased the induced voltage, air gap flux density, power output up to a significant level which effectively improves the generator performance.