By: Wei Huang, Radovan Kovacevic

Research Center for Advanced Manufacturing, Southern Methodist University

High-strength steels are being found of great applications in different industries because of their high strength-to-weight ratio. In order to achieve a high quality weld with full weld penetration during the laser welding process of the high strength steel, acoustic monitoring based on a single microphone is applied to sense the weld penetration in this study. As the biggest obstacle and challenge for acoustic monitoring during the welding process, the intensive background noise in the industrial environment great limits the broad applications of microphone as the sensor to monitor the weld quality such as the penetration depth. In order to overcome this obstacle posed by the background noise, a noise reduction method called spectral subtraction is applied to effectively reduce the background noises from the acoustic signals of interest acquired during the laser welding process. By analyzing the denoised acoustic signals both in the time domain and frequency domain, the different acoustic signatures of the acoustic signals from full weld penetration and partial weld penetration are obtained. The results indicate that the full weld penetration produces a higher sound pressure than the partial weld penetration and the corresponding acoustic signals also have different frequency distributions from 500 Hz to 1500 Hz. Based on these differences both in the time domain and frequency domain, two acoustic signatures (SPD and BP) are extracted from the denoised signals and used to distinguish the weld penetration states. Meanwhile, the relationship between these extracted acoustic signatures and the depth of weld penetration is also characterized by a neural network and a multiple regression analysis. As shown in the figure, the original acoustic signal acquired during a laser welding process under different welding parameters is greatly contaminated by the intensive background noises. After applying the noise reduction method and proper digital signal processing methods, the denoised acoustic signal, along with the acoustic signatures extracted based on this denoised signal are obtained and clearly distinguish the penetration states by setting a proper threshold. This method can be effectively and efficiently used to monitor the weld penetration during the laser welding process, which offers an alternative to other sensing techniques for laser welding process.

The above brief overview was extracted from its original abstract and paper presented at The International Congress on Applications of Lasers & Electro-Optics (ICALEO) in Orlando, FL. To order a copy of the complete proceedings from this conference click here

paper #1502