In this paper, the robustness of subspace-based identification methods for modal analysis is assessed for damaging systems including local Coulomb-friction mechanisms (or equivalently elasto-plasticity). The underlying objective is to capture damage-induced modal feature changes of structures, mainly frequency drop-off and damping increase, knowing that the presence of components with participating friction forces also affects modal signature. The performance of several modal detection procedures is assessed using the data acquired during the JVP (french for ‘Jonction Voile-Plancher’) test campaign during which a reinforced-concrete specimen, anchored to a shaking table by the means of swivels, is submitted to several kind of ground motion inputs of varied level. A comparison between Input/Output versus Output-Only identification is made using both seismic and broad-band random inputs and the potentialities of the modal selection procedures are highlighted for different cases where the two kind of non-linearities (damage occurrence and friction at swivels) are activated in different proportions. Regarding the presence of friction forces when conducting modal analysis, experimental results are backed up by a numerical study that enables a theoretical quantification of the apparent frequency and damping ratio values that should be expected after modal analysis for varied input-to-nonlinear forces ratios.

在本文中，针对包括局部库仑摩擦机制（或等效的弹塑性）在内的损坏系统，评估了用于模态分析的基于子空间的识别方法的稳健性。基本目标是捕获结构的损伤引起的模态特征变化，主要是频率下降和阻尼增加，知道具有参与摩擦力的组件的存在也会影响模态特征。使用 JVP（法语为“Jonction Voile-Plancher”）测试活动期间获得的数据评估了几种模态检测程序的性能，在此期间，通过旋转装置固定在振动台上的钢筋混凝土样本被提交给几种不同级别的地面运动输入。输入/输出与仅输出识别之间的比较是使用地震和宽带随机输入进行的，并且针对两种非线性（损坏发生和旋转摩擦）的不同情况强调了模态选择程序的潜力) 以不同的比例被激活。关于进行模态分析时摩擦力的存在，实验结果得到了数值研究的支持，该数值研究能够对表观频率和阻尼比值进行理论量化，这些值在对不同的输入与非线性力比进行模态分析后应该是预期的。