Laser ultrasonic guided waves deep subdomain transfer detection method for air-conditioning condenser pipelines

Abstract The air-conditioning condenser (ACC) is a critical component of air conditioning systems, where the curved pipe connections pose challenges for traditional contact-based damage detection methods due to the variety of damage types and complex geometric shapes. To address the limitation in structural damage identification accuracy caused by discrepancies between the material property parameters of the simulated model and the actual structure, a laser ultrasonic guided wave transfer detection method based on deep subdomain adaptation is proposed. First, a finite element model of the condenser pipeline is established to obtain detection signals under laser-simulated excitation. Next, wavelet decomposition is employed to extract key frequency band information from the signals, and a deep subdomain adaptation network is utilized to align the feature spaces of samples with different labels, thereby improving the recognition accuracy of damaged and normal samples. Finally, a dedicated software/hardware system for condenser detection is developed to conduct experiments on laser ultrasonic guided wave feature transfer detection for condenser damage. This approach provides a novel method for ACC health monitoring using laser ultrasonic guided waves. Experimental results indicate that the proposed method achieves a damage identification accuracy of 84.78% for ACC pipelines. In the future, a multi-point excitation mode will be considered to achieve full coverage detection of guided wave signals in complex condenser pipeline structures, thereby improving damage identification accuracy.