Department of Mechanical Engineering, The Hong Kong Polytechnic University, Kowloon, Hong Kong Special Administrative Region.
School of Aerospace Engineering, Xiamen University, Xiamen 361005, PR China.
Ultrasonics. 2022 Feb;119:106607. doi: 10.1016/j.ultras.2021.106607. Epub 2021 Oct 4.
Earlier, an ameliorated MUSIC (Am-MUSIC) algorithm is developed by the authors [1], aimed at expanding conventional MUSIC algorithm from linear array-facilitated nondestructive evaluation to in situ health monitoring with a sparse sensor network. Yet, Am-MUSIC leaves a twofold issue to be improved: i) the signal representation equation is constructed at each pixel across the inspection region, incurring high computational cost; and ii) the algorithm is applicable to monochromatic excitation only, ignoring signal features scattered out of the excitation frequency band which also carry information on structural integrity. With this motivation, a multiple-damage-scattered wavefield model is developed, with which the signal representation equation is constructed in the frequency domain, avoiding computationally expensive pixel-based calculation - referred to as frequency-domain MUSIC (F-MUSIC). F-MUSIC quantifies the orthogonal attributes between the signal subspace and noise subspace inherent in signal representation equation, and generates a full spatial spectrum of the inspected sample to visualize damage. Modeling in the frequency domain endows F-MUSIC with the capacity to fuse rich information scattered in a broad band and therefore enhance imaging precision. Both simulation and experiment are performed to validate F-MUSIC when used for imaging single and multiple sites of damage in an isotropic plate waveguide with a sparse sensor network. Results accentuate that effectiveness of F-MUSIC is not limited by the quantity of damage, and imaging precision is not downgraded due to the use of a highly sparse sensor network - a challenging task for conventional MUSIC algorithm to fulfil.
早些时候，作者[1]开发了一种改进的 MUSIC（Am-MUSIC）算法，旨在将传统的 MUSIC 算法从线性阵列辅助无损评估扩展到具有稀疏传感器网络的原位健康监测。然而，Am-MUSIC 仍存在两个需要改进的问题：i）在检测区域的每个像素处构建信号表示方程，这会导致较高的计算成本；ii）该算法仅适用于单色激励，忽略了激励频带之外散射的信号特征，这些特征也携带结构完整性的信息。受此启发，开发了一种多损伤散射波场模型，利用该模型在频域构建信号表示方程，避免了基于像素的计算，这种计算方法被称为频域 MUSIC（F-MUSIC）。F-MUSIC 量化了信号表示方程中固有于信号子空间和噪声子空间之间的正交属性，并生成了被检测样本的全空间频谱，以可视化损伤。在频域进行建模使 F-MUSIC 具有融合宽带中丰富信息的能力，从而提高成像精度。在具有稀疏传感器网络的各向同性板波导中对单个和多个损伤位置进行成像时，分别进行了仿真和实验来验证 F-MUSIC。结果强调，F-MUSIC 的有效性不受损伤数量的限制，并且由于使用了高度稀疏的传感器网络，成像精度不会降低，这是传统 MUSIC 算法难以完成的任务。
