Share Email Print
cover

Proceedings Paper

Spectroscopic analysis technique for arc-welding process control
Author(s): Jesús Mirapeix; Adolfo Cobo; Olga Conde; María Ángeles Quintela; José-Miguel López-Higuera
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

The spectroscopic analysis of the light emitted by thermal plasmas has found many applications, from chemical analysis to monitoring and control of industrial processes. Particularly, it has been demonstrated that the analysis of the thermal plasma generated during arc or laser welding can supply information about the process and, thus, about the quality of the weld. In some critical applications (e.g. the aerospace sector), an early, real-time detection of defects in the weld seam (oxidation, porosity, lack of penetration, ...) is highly desirable as it can reduce expensive non-destructive testing (NDT). Among others techniques, full spectroscopic analysis of the plasma emission is known to offer rich information about the process itself, but it is also very demanding in terms of real-time implementations. In this paper, we proposed a technique for the analysis of the plasma emission spectrum that is able to detect, in real-time, changes in the process parameters that could lead to the formation of defects in the weld seam. It is based on the estimation of the electronic temperature of the plasma through the analysis of the emission peaks from multiple atomic species. Unlike traditional techniques, which usually involve peak fitting to Voigt functions using the Levenberg-Marquardt recursive method, we employ the LPO (Linear Phase Operator) sub-pixel algorithm to accurately estimate the central wavelength of the peaks (allowing an automatic identification of each atomic species) and cubic-spline interpolation of the noisy data to obtain the intensity and width of the peaks. Experimental tests on TIG-welding using fiber-optic capture of light and a low-cost CCD-based spectrometer, show that some typical defects can be easily detected and identified with this technique, whose typical processing time for multiple peak analysis is less than 20msec. running in a conventional PC.

Paper Details

Date Published: 11 October 2005
PDF: 6 pages
Proc. SPIE 5948, Photonics Applications in Industry and Research IV, 594824 (11 October 2005); doi: 10.1117/12.621939
Show Author Affiliations
Jesús Mirapeix, Univ. de Cantabria (Spain)
Adolfo Cobo, Univ. de Cantabria (Spain)
Olga Conde, Univ. de Cantabria (Spain)
María Ángeles Quintela, Univ. de Cantabria (Spain)
José-Miguel López-Higuera, Univ. de Cantabria (Spain)


Published in SPIE Proceedings Vol. 5948:
Photonics Applications in Industry and Research IV
Ryszard S. Romaniuk; Stefan Simrock; Vladimir M. Lutkovski, Editor(s)

© SPIE. Terms of Use
Back to Top