Detecting subtle cosmetic defects in automotive skin panels

Cosmetic defects, such as 'hollows', are deviations in topology of automotive skin panels that form as a result of springback at the end of the forming process. These deviations are usually too small and local to be detected by discrete measurements of the panel but become visually apparent after the application of paint. As a result, the perceived quality of a panel may become unacceptable and considerable time may be dedicated to minimizing their occurrence through tool modifications. This paper proposes that there are three aspects to the problem. The first is the springback of the panel, the second is the optics of the painted panel, and the third is the ability of human sight to perceive a defect. In particular, it is argued that hollows cause optical distortions that inform the human eye of the presence of a defect. The paper then suggests that signal processing techniques, in particular the wavelet transform, provide a way to relate the geometry of a hollow to the resulting optical distortion. The transform was applied to two physical parts and the paper will discuss the effectiveness of the transform in locating and quantifying the relative severities of the defects that were present