| As one of the most economical and efficient ways to join metals, welding is always vital to the economy of a country because up to 50% of the gross national product is related to welding in one way or another. Welding is a very complex process and its final weld quality will be affected by a number of process parameters. Therefore, in order to improve the weld quality and increase the productivity, different sensing, monitoring and control methods and systems are studied and developed for various welding applications. In this study, two sensing techniques, one based on acoustic monitoring of the air-borne sound during welding process, and the other one to visually monitor the geometrical features of weld joint before, during and after welding, are studied and implemented. A single microphone is employed to on-line monitor the laser welding process. By applying different digital signal processing methods such as spectrum analysis and a noise reduction method called spectral subtraction, the acoustic signals acquired during the welding process are processed and analyzed. Based on the processing and analysis results, the feasibility of using these acquired acoustic signals to monitor the depth of weld penetration in laser welding is studied. Meanwhile, in order to further understand the acoustic phenomenon associated with the laser welding process, the mechanisms of the differences in the acoustic signatures from different weld penetrations are investigated. In addition, the relationship between the acquired acoustic signatures and the depth of weld penetration is also characterized as nonlinear by using a neural network and a multiple regression analysis. In order to remove the complex and intensive background noise contained in the acoustic signals captured during the laser welding process, both the spectral subtraction noise reduction method based on a single microphone and the beamforming noise reduction method based on a microphone array are applied to process the acquired acoustic signals. The welding of galvanized steels in lap joint configuration is also acoustically monitored on-line by a single microphone. The presence of weld defects such as spatters and blowholes are easily identified from the acoustic signatures by applying spectrum subtraction and wavelet analysis. In addition to the investigation of using air-borne sound in monitoring the weld quality, a laser-based machine vision system is also developed to perform different tasks such as the weld joint identification, automatic seam tracking, and post-weld quality inspection. The developed vision system is employed to achieve accurate measurement of the position and geometry information of different weld joints. Based on the obtained position information of the weld joint, automatic seam tracking is also achieved by the developed vision system for welding applications. As for the purpose of post-weld quality inspection, the developed vision system is applied to achieve a real-time 3D profiling, which could be used to scan the weld surface and obtain the image of the bead appearance after welding process and therefore, it is capable of detecting the presence of weld defects and identify their positions and sizes. |
