Data quality improvement and geometric information recovery for resistance spot welding with Spatial-Temporal Fast Fourier Transform

• Published in: Quality engineering Vol. 37; no. 4; pp. 693 - 705
• Main Authors: Xu, Boyang, Guo, Shenghan
• Format: Journal Article
• Language: English
• Published: Milwaukee Taylor & Francis 02.10.2025; Taylor & Francis Ltd
• Subjects: Destructive testing; Fast Fourier Transform; Fast Fourier transformations; Fourier transforms; Geometry; geometry recovery; Image quality; Information retrieval; Manufacturing; manufacturing thermography; Nondestructive testing; Quality assessment; Quality improvement; Recovery; Resistance spot welding; spatial-temporal; Thermal imaging; Thermal resistance; Thermography; Video
• ISSN: 0898-2112; 1532-4222
• DOI: 10.1080/08982112.2025.2509008

Geometric information is crucial to evaluating manufacturing products quality. Traditional quality evaluation relies on destructive testing, which destroys certain finished parts to measure the geometry. With the advancement of sensing technology, infrared (IR) thermography can be collected during manufacturing processes to provide evidence for nondestructive quality evaluation of manufacturing products. However, thermal images may be subject to low data quality that impedes effective information extraction from the data. Recovering a functional description of the parts geometry remains an open question. In this study, Spatial-Temporal Fast Fourier Transform ('ST-FFT') was introduced as imaging quality improvement (IQI) for thermal videos. Least Squares Estimate was used to recover functional descriptions of parts geometry from improved thermal images. The proposed method was validated using seven offline thermal videos of lab-based resistance spot welding processes. The results demonstrated that the variations of geometry recovery ratio of major to minor axes of the ellipse weld nugget calculated without image quality assessment (IQA) are much larger than the ratios calculated with IQA. The ratios estimated from the "ST-FFT" images are closest to ground truth values for all cases. Therefore, the proposed framework enables systematic processing and analysis of manufacturing thermal videos, including IQA, IQI, and geometry recovery.