Discrete cavities with guide vanes were developed and optimized to improve the operating stability of a centrifugal compressor. Various combinations of search algorithms and surrogate models were tested to find the best optimization methods. Aerodynamic analysis was performed using three-dimensional Reynolds-averaged Navier–Stokes equations. The numerical results obtained for the total pressure ratio and adiabatic efficiency were validated with experimental data for the centrifugal compressor with a smooth casing. The yaw and pitch angles of the guide vanes and axial distance between cavities were selected as design variables. The stall margin was used as an objective function for the design optimization. Latin hypercube sampling was used to select 27 sample points in the design space. The best combination was found by testing four surrogate models (response surface approximation, Kriging, radial basis neural network, and deep neural network models) and three searching algorithms (a genetic algorithm, particle swarm optimization, and hybrid PSO-GA). Hybrid PSO-GA with the DNN model showed the best overall results. The optimum design showed increments of 13.36% and 3.78% in the stall margin compared to compressors with a smooth casing and the reference cavity design, respectively.

中文翻译：

---

基于优化技术对比分析的离心式压缩机带导叶离散腔优化

开发并优化了带导叶的离散腔，以提高离心压缩机的运行稳定性。测试了搜索算法和代理模型的各种组合以找到最佳优化方法。使用三维雷诺平均 Navier-Stokes 方程进行空气动力学分析。获得的总压比和绝热效率的数值结果通过具有光滑外壳的离心压缩机的实验数据进行了验证。导向叶片的偏航角和俯仰角以及空腔之间的轴向距离被选为设计变量。失速裕度用作设计优化的目标函数。拉丁超立方抽样用于在设计空间中选取 27 个样本点。通过测试四种替代模型（响应面近似、克里金法、径向基神经网络和深度神经网络模型）和三种搜索算法（遗传算法、粒子群优化和混合 PSO-GA）找到了最佳组合。带有 DNN 模型的混合 PSO-GA 显示出最好的整体结果。与具有光滑外壳和参考空腔设计的压缩机相比，优化设计的失速裕度分别增加了 13.36% 和 3.78%。