基于超声导波多角度稀疏表示的复合材料损伤定量评估方法研究

With wide application of composite structures in many fields (e.g., aviation, energy, transportation and ships), condition monitoring and damage evaluation of in-service composite structures become more and more demanding. Ultrasonic guided wave is one of the most promising structural health monitoring techniques, however, interpretation of complicated guided wave signals is difficult. Through sparse decomposition, complicated guided wave signals could be transformed into concise expression with specific physical meaning, which contributes to signal interpretation and the further damage evaluation. The motivation and objective of this project is to establish a systematic strategy for guided wave multi-dimension sparse representation. Firstly, an underdetermined equation in frequency-wavenumber domain based on far-field guided wave is built, and the dispersion curve is obtained by solving the equation to characterize guided wave propagation. Subsequently, the matching mechanism between the guided wave signal and the over-complete dictionary is studied to improve the accuracy of time domain sparse decomposition, thus the waveform component caused by damage could be clearly separated, recognized and extracted. Finally, the methodology for feature extraction is established and the evolution of feature associated with damage parameter is investigated. On this basis, an information fusion technique based on spatial domain sparse representation is proposed by utilizing the evolution information and damage distribution characteristic to realize quantitative evaluation of composite structures.

随着复合材料在航空、能源、交通、船舶等领域的广泛应用，对其服役过程中状态监测和损伤评估技术的要求也越来越高。超声导波是公认的最具潜力的结构健康监测技术之一，然而，复杂导波信号的信息解读能力不足是其开展有效应用的突出瓶颈。运用稀疏分解理论可将复杂信号转化为简洁且物理意义明确的表达形式，可望对准确揭示导波信号物理信息的构成，进而实现复合材料损伤评估产生重要推动。本项目围绕超声导波多角度稀疏表示方法开展研究，通过建立基于声波远场传播模型的频率-波数域多参稀疏欠定方程，获取表征导波传播机理的频散曲线；在此基础上，揭示导波信号特征和过完备字典的最优匹配机制，提高时间域稀疏分解精度，实现损伤衍生成分的分离、辨识和提取；进一步，研究损伤特征指标提取方法和指标随损伤参数的演变规律，结合损伤分布规律，建立基于空间域稀疏表示的全信息融合技术，实现复合材料损伤定量评估。

随着复合材料在航空、能源、交通、船舶等领域的广泛应用，对其服役过程中状态监测和损伤评估技术的要求也越来越高。超声导波是最具潜力的无损检测技术之一，然而，复杂导波信号的信息解读能力不足是其开展有效应用的突出瓶颈。运用稀疏分解理论可将复杂信号转化为简洁且物理意义明确的表达形式，可望对准确揭示导波信号物理信息的构成，进而实现复合材料损伤评估产生重要推动，是超声导波损伤检测的一个前沿科学问题。本项目围绕超声导波多角度稀疏表示方法开展研究，首先，构建了适用于层合板结构的超声导波声场模型，拓宽了传统各向同性介质超声建模方法的应用范围，提出了过完备字典与导波声场模型的最优匹配准则，提高了时间域稀疏分解精度；进一步，提出了匹配时频域超声传播特征的信号处理算法，利用先验声场模型显著提升了小波变换与稀疏分解算法性能，提高了信息解析能力，进而提出了基于时频域导波信号的损伤特征指标提取方法，克服了传统窄带特征指标的局限性；在此基础上，建立了适用于多输入多输出系统的广义空间域稀疏成像算法，克服了检测效率与成像性能的相互制约限制，实现了效率与性能的同时优化；最后，构建了可解释超声导波多层卷积稀疏编码网络模型，充分利用了宽带信号携带的丰富结构信息，实现了碳纤维增强复合材料分层损伤的轮廓定量评估。综上所述，本项目的研究提出了超声导波传播机制建模、时-频-空多作用域稀疏分解、特征参数全信息融合、损伤评估智能模型构建等一系列方法和策略，显著提高了检测结果的准确率，为实现超声导波损伤定量评估提供了新思路。基于研究成果，负责人在本领域权威期刊发表SCI论文8篇，均为JCR Q1区，其中第一或通讯作者论文7篇，申请国家发明专利2项，为在役复合材料结构的状态评估、健康维护和有效管理奠定了理论和方法基础。