Performance Optimization of an Axial Compressor Using a Novel Multifidelity Surrogate Model Based on Flow Field Extraction

During the utilization of efficient optimization algorithms for axial compressors, the construction of a precise performance prediction surrogate model stands as a pivotal step. To reduce the cost of constructing the surrogate model while ensuring prediction accuracy, a novel multifidelity surrogate model based on flow field extraction (FFMFS) is proposed in this paper. In constructing FFMFS, two sets of samples with different fidelity are employed for model training, and six important flow field variables in axial compressors are extracted to modify the performance deviation between low-fidelity (LF) and high-fidelity (HF) results. Based on the proposed FFMFS, the aerodynamic performance of a 1.5-stage subsonic axial compressor is optimized, and the numerical method used in the optimization is validated on a 3.5-stage axial compressor test bench. During optimization, adjustments are made to the rotor blade profile, taking into account a total of 28 design variables and six objective functions. The FFMFS constructed for this compressor demonstrates a high prediction accuracy with a $R2$ value of 0.96, while also significantly reducing the sample generation cost. The optimization results show that the compressor efficiency and pressure ratio are significantly improved across the entire operating range. As a result of adjusting the rotor blade profile, the flow loss inside the compressor is evidently reduced. This work provides a new framework for constructing MFS with flow field information of axial compressors.