A voltage signal analysis method for the welding process using the rice frequency

This paper investigates the Rice frequency f R , defined as the square root of the ratio of the second to the zeroth spectral moment, as a descriptor of arc voltage signals in gas metal arc (GMA) welding. Short-Time Rice Frequency Analysis (STRFA) was applied to signals recorded in 89 controlled welds, 80 of which were analysed after documented exclusions, covering a reference condition and eleven deliberately introduced process disturbances. The disturbed zones of every weld were annotated from the recorded process channels and from photographs of the specimens. This allowed the disturbed segments to be compared with undisturbed segments of the same weld, which removes the influence of differences between welding sessions. When all windows of whole welds were pooled and compared with the reference welds, the effects of local disturbances were diluted. Within welds, shielding gas interruption decreased the mean f R by 24.5% (5/5 welds), and a 16% increase in welding current increased it by 155% (7/7 welds). The Rice frequency decreased monotonically with root gap (Spearman ρ = −0.90), groove angle (ρ = −0.76) and bevel hole diameter (ρ = −0.59), whereas travel speed and localised oil contamination had no consistent effect. The direction of the response reversed between window widths of about 10 and 50 ms. Long windows reflect the metal transfer frequency, while short windows reflect fast arc voltage fluctuations. The Rice frequency is therefore a computationally inexpensive, two-scale descriptor suitable for online welding monitoring, best used together with other signal features.