Ultrasonic Guided Wave Monitoring System Using Histogram Transformer for Enhanced Damage Detection
Xiao Ying, Zhao Wang, Linfeng Wang, He Sun, Haibo Li, Yantao Liu, Fuzai Lv, Jianping Li, Yang Liu · Advanced Devices & Instrumentation · 2025
In response to the challenges of high integration and high-precision imaging in ultrasonic guided wave (UGW) monitoring systems, we have developed an advanced UGW monitoring system employing a convolutional neural network (CNN) with histogram transformer block (HTB) algorithm for enhanced damage detection. This system utilizes field-programmable gate array-based direct digital synthesis to generate excitation signals, and high-integration, multi-channel parallel front-end devices to capture the UGW signals. The system’s core processing unit is built on a Zynq device, enabling parallel processing of multi-channel UGW data. In the system, we have implemented a CNN-HTB-based method to achieve high-precision damage imaging. Compared to the reconstruction algorithm for probabilistic inspection of damage, this approach provides precise information on the location, shape, and thickness of the damage. Compared to the algorithm based on CNN, it improves the training speed and imaging accuracy. Validation experiments on aluminum plates demonstrated that the imaging results achieved a correlation coefficient of 0.9807, surpassing the CNN method, which yielded a correlation coefficient of 0.9405. Additionally, the system’s design features a highly integrated industrial notebook, offering a stable and reliable hardware platform for industrial applications.