Multimodal thermal wave imaging and pseudo-3D morphological reconstruction for quantitative detection of micron-scale cracks in solder joints

The reliable detection of micro-scale cracks in solder joints and substrates is critical for ensuring the long-term performance of mechanical electronic devices. Active infrared thermography offers a non-contact and efficient inspection modality, yet the influence of excitation signal waveforms and feature extraction algorithms on detection performance remains insufficiently characterized. This study systematically investigates micron crack detection and depth-resolved morphological mapping using multi-source thermal modulation excitation combined with frequency-domain, time-domain, and statistical feature extraction algorithms. Defect detectability, regional consistency, and internal feature saliency are evaluated using six metrics: CNR, SNR, DIS, BD, RSNR, and PSD. The results demonstrate that the feature points down-sampling algorithm achieves reliable defect identification under square and sine wave excitation, while Fourier transform and dual-channel digital lock-in algorithm provide complementary phase-based enhancement. Under pulse excitation, polynomial fitting correlation algorithm effectively resolves edge interference. Among all evaluated methods, the feature points down-sampling algorithm achieves the best overall performance, with BD values exceeding the average of all algorithms and RSNR reaching up to 0.95 under square-wave excitation. Three-dimensional morphology reconstruction is achieved with high fidelity, the relative reconstruction error follows a power-law relationship with crack width, expressed as y = 5.43 x ^−0.48, R ² = 0.9219, with errors below 10% for crack width larger than 0.2 mm. These findings establish an optimal strategy for high-precision solder joint crack detection and provide a quantitative basis for reliability assessment in mechanical electronic manufacturing.