基于神经网络的涡轮叶片结构模态仿真

针对工作在复杂环境下的航空发动机涡轮叶片模态特性在工作环境变化下会发生相应的变化,应用有限元分析软件ANSYS对航空发动机低压一级涡轮叶片进行了模态仿真,并在ANSYS概率设计模块中应用响应表面法和蒙特卡洛法模拟了旋转软化、应力强化、温度梯度和材料参数的分布,以这四类变量的分布为神经网络的训练样本,应用遗传算法优化后的三层BP神经网络对该样本进行训练,得出固有频率的分布,从而得到了以上四类变量的变化对固有频率的影响.仿真结果与实际吻合较好.

Modal Simulation of Turbine Blade Based on Artificial Neural Network

The blade of an airo-engine turbine works in complex environment,its modal parameters are changed with the change of work environment.So the modal simulation of an air-engine turbine blade is made by using FEM software ANSYS.The parameter scatter is the training sampling of the three layer artificial neural network,which is optimized by genetic algorithms.The frequency is calculated by this method.So the influence of the four kinds of parameterss on the frequency is gotten.The result is in accord with the fact.