基于模态应变能与神经网络的钢网架损伤检测方法

Damage diagnosis of spatial trusses based on modal strain energy and neural network

神经网络通过对样本的学习,获得结构模态参数与损伤之间的映射关系。目前基于神经网络的损伤检测已经越来越广泛地使用在非破坏性损伤诊断当中。但对于大型结构而言,它的训练样本数量过大,将消耗大量的计算。所以如何降低神经网络的计算量使其可用于大型结构的损伤诊断是一个亟待解决的问题。为了解决这个问题,提出了空间钢网架损伤的两步诊断法:第一步,利用模态应变能对结构损伤的敏感性,判断出结构损伤的可能位置;第二步,利用神经网络从可能发生损伤的杆件中定位出实际损伤的位置,并进行损伤程度的判断。利用一个空间网架作为数值算例,进行可行性验证。结果表明此方法可以准确判断出结构的损伤位置以及损伤大小,是一种行之有效的方法。

According to network training, Artificial Neural Network （ANN） obtains the relationship between modal parameters and structural damage. Recently, ANN has been widely used for nondestructive damage diagnosis. For large structures, there are a great many of training samples that lead to a large amount of computations, thus ANN is not an efficient method. In order to solve proposed. First the modal strain e damage approximately; then, the e such problem, a two-step approach to damage identification mergy, which is sensitive to the structural damage, is used to neural network technique is applied to find the of spatial trusses is locate the structural damage from those possible locations, and to assess the degree of damage. A numerical example for a spatial truss is provided to verify the practicability of the method. The results show that this two-stem method is effective to locate damage and assess damage degree.