【摘要】 离心叶轮叶片在高速旋转时容易因裂纹扩展出现断裂破坏。由于叶片几何形状和载荷均很复杂，采用公式法计算I型裂纹的应力强度因子不可避免地存在着误差。扩展有限元法分析应力强度因子虽精度较好，但一般要花费大量的计算时间且有时收敛困难。首先基于ABAQUS软件扩展有限元法模块，仿真分析了不同的裂纹起始位置和裂纹长度下的叶片裂纹尖端应力强度因子，得到其与裂纹长度和起始位置的关系。接下来，基于断裂力学理论知识，检验了公式法估算应力强度因子的准确度。最后，以扩展有限元法的仿真结果为训练数据，以叶片裂纹位置和裂纹长度为输入参数，建立了裂纹尖端应力强度因子的多层反向传播人工神经网络(back propagation artificial neural network, BP-ANN)。算例表明，BP-ANN的预测精度优于公式法，并可有效减少扩展有限元法的仿真次数，推进断裂力学在离心叶轮可靠性设计中的应用。更多还原

【Abstract】 Centrifugal impeller blades are prone to fracture failure due to crack propagation when rotating at high speed. Due to the complex blade geometry and load, there is an error in calculating the stress intensity factor of mode-I crack by formula. Due to the complex blade geometry and load, errors are generated inevitably in calculating the stress intensity factor of mode I crack using theoretical formulae. Although extended finite element method can be used to obtain stress intensity factors with good accuracy, the computation is generally time consuming, and sometimes hard to converge. Firstly, the stress intensity factors at the crack tips of the blade were simulated considering different crack start positions and crack lengths using ABAQUS’s XFEM module, and their relationship with crack length and starting position was obtained. Next, the accuracy of the formulaic estimate on the stress intensity factor was tested based on the theory of fracture mechanics. Finally, a multi-layer back propagation artificial neural network(BP-ANN) with the stress intensity factor of the crack tip was established using the simulation results of the extended finite element method as the training data and the crack position and crack length of the blade as the input parameters. The example shows that the predicting accuracy of BP-ANN is better than that of the formulaic method. With the BP-ANN, the simulation times of the extended finite element method can be effectively reduced which is beneficial to promoting the application of fracture mechanics in the reliability design of centrifugal impellers.更多还原