基于改进式SL0算法的超声混叠信号分离

Ultrasonicaliasing signal separation based on improved SL0 algorithm

由多层等厚铝合金板真空扩散焊接制成的航空构件在超声无损检测过程中常常因为不同层的缺陷在水平于声波传播方向上发生重叠,导致上、下层缺陷回波发生混叠,影响下层缺陷的判别。因此混叠信号的分离对重叠缺陷的检出具有重要意义。在传统的平滑化L0算法的基础上,改用双曲反正切函数和修正牛顿迭代法来逼近L0范数的最小值,同时将投影条件中的等式约束放宽为不等式约束,再利用凸优化方法进行求解,最终实现了含噪超声重叠信号的分解与重构。实验表明,对于含噪信号,相比于其他稀疏表示方法,改进后的SL0算法在稀疏能力和重建效果上都有着更好的表现。同时,稀疏分解得到的信号参数和实际测得的真实数据也具有很好的一致性,证明了该算法的准确性。

In the process of ultrasonic non-destructive testing, the defects of different layers inside the aerospace components made of multilayer aluminum alloy plates of equal thickness by vacuum diffusion welding often overlap horizontally in the direction of sound wave propagation. This results in aliasing of echoes of the upper and lower defects, affecting identification of the defects of the lower layer. Therefore, separation of aliasing signals has great significance to detection of overlapping defects. In this paper, based on the traditional SL0 algorithm, the hyperbolic arctangent function and the modified Newton iteration method are used to approximate the minimum value of the L0 norm. Meanwhile, the optimal solution is obtained by convex quadratic programming with inequality constraints. Finally, the decomposition and reconstruction of noisy ultrasound overlapping signals are realized. The results reveal that compared with other sparse representation methods, the improved SL0 algorithm has better performance in both sparse ability and reconstruction effect for noisy signals. Besides, the parameters of the signal obtained by sparse decomposition are consistent with the real data obtained by actual measurement, which proves the accuracy of the algorithm.