Dataset for Electric Machine Performance Indices Prediction with Handling of Geometric Constraints

The inaccurate geometric modeling of an interior permanent magnet synchronous motor (IPMSM) would lead to incorrect performance analysis results or unplanned termination of massive finite element analysis (FEA) simulation runs. Therefore, it is crucial to ensure the feasibility of the IPMSM geometric model before analyzing its performance with FEA tools. This paper presents a bottom-up geometric modeling approach and constraint violation repair method for different parts of an IPMSM to ensure that an accurate model can be obtained. The mathematical models of geometric constraint are derived from the stator and rotor with specific structures to ensure its design variables are adjusted appropriately. The complexity of geometric constraint and interdependence on other design variables result in geometric constraint violations that should be repaired with simple, complex, or coupled operators. This paper also provides the methodology for generating comprehensive IPMSM datasets using the Latin Hypercube Sampling method with geometric violation repair operators, demonstrating the efficiency and effectiveness of the proposed method for handling the geometric constraint violations. The dataset can be employed to assess the performance of surrogate models or explore new models for characterizing the performance of IPMSMs. The prediction of the baseline with the Radial Basis Function (RBF) model exhibits good consistency with the FEA results. The proposed geometric modeling approach and constraint handling method are universal for any pre-processor and massive simulation modules of electric machines within FEA analysis. Moreover, the proposed method can be extended to construct geometric models for synchronous machines, induction machines, SynRM, and other innovative structure electric machines.