Wavelet Transforms for Quality Engineering Applications

We propose a process control and analysis technique for two-dimensional, irregularshaped parts whose digitized images are transformed of a discrete wavelet transform (DWT) algorithm. Image processing methods are applied to obtain two-dimensional part profiles and changed to one-dimensional boundary discrete signal in order to simplify computation complexity. Wavelet transform is employed to extract “signature” of the signal. Wavelet coefficients, which are arranged in layers, are chosen for the signal compact representation. Simulation tests are made to study the relationship between sample size and wavelet coefficient statistics that are useful to distinguish dimensional variations among parts due to production inperfection. Initial results show that this technique is very promising for automatic inspection and quality control in modern production environments.