针对环形阵列超声导波相控阵全聚焦成像法伪像多、成像质量较低的问题，提出一种适用于薄壁结构损伤识别定位的全聚焦加权成像方法。首先根据Lamb波散射模型和频散先验曲线构造单模态频散字典，构建多模态响应信号与频散字典间的稀疏模型并求解，使得多模态响应信号在单模态字典下稀疏表示，实现超声导波模态分离；其次引入虚拟时间反转技术消除全矩阵数据中单模态响应信号的频散，矫正相位偏差，补偿幅值衰减；最后将损伤成像区域划分成若干网格，并假设每个网格点均有潜在损伤的可能，以此构造潜在损伤原子库，构建单模态响应信号与潜在损伤原子间的稀疏模型并求解得到权重因子，实现损伤加权全聚焦成像。试验结果表明，所提的全聚焦加权成像方法能够实现铝板结构单损伤、多损伤高精度定位和高分辨率成像，最大误差为63 mm。

Here, aiming at problems of many artifacts and lower imaging quality of annular array ultrasonic guided wave phased array full focus imaging method, a full focus weighted imaging method suitable for identifying and locating damage in thin-walled structures was proposed. Firstly, a single-modal frequency dispersion dictionary was constructed according to Lamb wave scattering model and frequency dispersion prior curve, and a sparse model between multi-modal response signal and the frequency dispersion dictionary was constructed and solved to make a multimodal response signal be sparsely represented under the single-modal dictionary, and realize ultrasonic guided wave mode separation. Secondly, the virtual time reversal technology was introduced to eliminate frequency dispersion of single-modal response signal in full matrix data, correct phase deviation, and compensate for amplitude attenuation. Finally, damage imaging area was divided into several grids, and it was assumed that each grid point has the possibility of potential damage. Based on this assumption, a potential damage atom library was constructed, and a sparse model between single-modal response signal and potential damage atoms was constructed and solved to obtain weight factors, and realize damage weighted full focus imaging. The experimental results showed that the proposed full focus weighted imaging method can realize high-precision positioning and high-resolution imaging of single and multi-damage of aluminum plate structures, the maximum error is 63 mm.