基于数据融合的张弦桁架损伤识别方法与试验研究

针对张弦桁架采用单损伤指标识别损伤位置易受到干扰甚至产生误判的问题，提出基于曲率模态差和模态柔度差曲率融合的损伤识别方法。基于D-S证据矩阵和加权平均两种数据融合准则，建立张弦桁架损伤识别的单次融合和两阶段融合识别流程。采用有限元软件ANSYS建立张弦桁架分析模型，利用减小构件截面积模拟弦杆损伤，分别采用非融合单损伤指标、单次数据融合和两阶段融合方法进行损伤识别分析。结果表明：单损伤识别指标的抗干扰能力较低，尤其对支座附近下弦杆的识别效果差；基于数据融合的损伤识别方法能够综合多种指标损伤识别结果，有效提高损伤位置识别的准确性；与加权平均数据融合准则相比，基于D-S证据矩阵准则的数据融合方法更能有效降低非损伤位置干扰，识别效果更好；两阶段融合方法效果优于单次融合方法，损伤位置的损伤概率愈发接近于1，非损伤位置的损伤概率愈发趋近于0。最后开展了一榀张弦桁架模型的损伤识别试验，通过试验验证了当所获取模态数据受到环境或测试噪声等因素影响下，文章提出的基于数据融合的损伤识别方法仍具有较好的鲁棒性，能够有效识别实际张弦桁架结构的损伤位置。但由于曲率模态差以及模态柔度差曲率难以识别损伤程度，因此文章方法仅能够提高损伤位置的识别精度。

Analytical and experimental research on damage identification of #br#cable-stayed arch-truss based on data fusion

 Cable-stayed arch-truss is susceptible to interference or even suffering from misjudgment when using the diagnosis method of single damage index for damage identification. In order to solve this problem,a damage identification method,using recognition results of data fusion of the curvature modal difference and the modal flexibility difference curvature,is presented in this study. Based on two fusion criterions of D-S evidence matrix and the weighted average of the two criteria,the single-stage fusion and two-stage fusion methods are constructed. An analytical model is established using ANSYS and the rod damage is simulated by reducing sectional area of rods. Non-fusion,single-stage and two-stage fusion method are used to identify the rod damage separately. Results show that non-fusion single index has worst anti-interference ability,especially for the bottom chord near supports. The damage identification accuracy is effectively improved by using data fusion. Compared with weighted average criterion,D-S evidence matrix criterion is more effective in reducing the interference of non-damage locations and has better identification effect. Two-stage fusion method is more effective than single-stage fusion method,with the damage probability closer to 1 at damage locations and 0 at non-damage locations. Finally,a damage identification test of cable-stayed arch-truss model is carried out. Test results show that the present method is robust enough in identifying the damage of the actual structure,even under the influence of environmental or measurement noises. However,for the curvature modal difference and the modal flexibility difference curvature,it is difficult to identify the damage degree. Therefore,the method proposed here can only improve the recognition accuracy of the damage location. 