The mad, the bad, and the different: Essays in honor of Simon Dinitz : Edited by Israel L. Barak-Glantz and C. Ronald Huff. Lexington Books (D.C. Heath and Company, Lexington, Massachusetts 02173), 1981, 280 pp., hardcover—$23.95

Abstract

Automakers today are challenged with improving the quality of spot welded structures, while reducing the costs associated with quality control. While ultrasonic C-scan testing has become a mainstay of quality control programs, it is not suitable for inspection of every weld in a high volume production due to the considerable operator skill required, and the requirement to take parts off-line for inspection. The dynamic resistance curve of a weld is known to contain information about the development of a weld, however there is not always a clear way to relate that directly back to the quality of a weld and exactly how a system may have arrived at that decision. Because the dynamic resistance data is available inline for every weld, developing a method of determining weld quality from the dynamic resistance would allow for process faults to be diagnosed sooner than would be possible with periodic off-line inspection.In this paper we present a method of estimating nugget diameter directly from the dynamic resistance obtained during welding, by means of Principal Component Analysis (PCA), autocorrelation and multilinear regression. The accuracy of estimated nugget diameters is compared to ultrasound inspection in a production environment. The nugget diameter estimated by the dynamic resistance was found to be more accurate than ultrasound, with Mean Squared Error values of 2.26 for Ultrasound and 0.33 for the Dynamic Resistance Method. For welds with misaligned electrodes, the effectiveness of Ultrasound dropped significantly when the probe was unable to sit flat on the weld surface. The method presented in this paper is suitable for inspection of every weld in a high-volume production, and has been shown to outperform ultrasonic inspection in estimation accuracy.