基于改进型动量BP神经网络算法的风扇叶片性能优化研究

为探索叶片优化设计的新途径,获得具有更高气动性能的各类叶片,使用了一种改进型变学习率动量BP神经网络算法和非均匀有理B样条函数法相结合的方法,对某风扇压气机的第2级转子叶片特别是针对其根部基元级进行了气动优化改型设计;通过对转子叶片和整个风扇压气机优化设计前后气动性能的对比分析,表明该优化设计方法确实可行,能够改善转子叶片以及整个风扇压气机的气动性能,使整机的效率和压比分别增长了0.379 4%以及0.225 4%,达到了优化的目的。

An Improved Momentum Back-Propagation Neural Network for Optimizing the Aero-Dynamical Performance of a Fan Blade

We present a new method for blade shape optimization with better aero-dynamical performance;such a method combines a variable learning rate momentum back-propagation neural network(VLMBPNN) algorithm with non-uniform rational Bspline(NURBS) method.This method is used to reshape the second rotor blade of a two stage fan compressor,especially the hub section,to improve its aero-dynamical performance.The computational fluid dynamics(CFD) results of the optimized rotor blade and the optimized whole fan compressor are obtained and compared with those of the original rotor and the original fan compressor.The results show that this method is feasible and can improve the flow field structure of the rotor blade as well as the whole fan compressor.The results also indicate that the isentropic efficiency and total pressure ratio of the whole fan compressor are increased by 0.379 4 percent and 0.225 4 percent respectively.