利用修正的BP神经网络进行框架结构损伤诊断

提出了一种基于修正的BP神经网络方法的结构损伤诊断方法，该方法通过基于误差曲面调整权值以及引入动量因子算法来避免出现局部极小，并通过自适应调整学习速率来提高学习效率，从而解决了传统BP神经网络在实际应用中存在的收敛速度慢和存在局部极小的问题。算例表明，该方法能够不仅较为准确的检测出结构的不同程度损伤，而且提高了计算效率，同时具有较强的鲁棒性。

Structural damage detection using amendatory BP neural networks

A new amendatory back - propagation neural network has been proposed in this paper. This method can avoid local minima by modulating modulated values through error surface, and by adopting a dynamic momentum algorithm. And it can accelerate steepest descent by self - adapting to modulate the study efficiency. This method solves two practical problems encountered by the traditional back - propagation method ,i. e. ,slow learning progress and convergence to a false local minimum. Numerical results show that the method may enhance computational eff-ciency as well as detection accuracy, and it also has a good robustness