High precision, repeatability, and quality are the three vital requirements in laser welding production. For accurate real-time tracking and inspecting the laser welding process, the high-performance sensors are extremely demanded. Monitored signal reliability can be significantly increased by using high resolution, digital CMOS sensors and high-speed, real-time image processing technologies. This feature presents the latest developments in high-performance optical joint tracking systems and optical inspection systems based on these technologies. Using a coaxially aligned CMOS imaging detector, the optical signals emission of the plasma during CO2 laser welding was studied. The camera images taken from the process were analyzed with image-processing algorithms. Compared with the lateral systems, coaxial arrangement of the camera allows observing the significant process characteristics. Experimental evidence shows that the system can monitor the instability of the keyhole, the gap caused by the welding distortion, and the deviations from the desired welding path. By the image analysis, the spatially distribution intensity of the plasma emission was analyzed, and it can be correlated to the penetration state and the penetration depth. Thus the laser welding process and the weld quality can be evaluated.
The spectra of the optical signals emitted by plasma during laser welding were studied using wavelet analysis. In the presence of the wavelet analysis, the detection of the pool penetration defects in welding process was realized. Comparing optical signals in welding process with and without joint penetration problem, the significant differences can be observed. By continuous wavelet transform, the coefficient lines of some scales have rapid fluctuation at the positions where pool penetration defect occurred, and the maximal coefficient values at the signal breakdown correspond to the length of the defect. In discrete wavelet analysis, the amplitude of low-frequency component of optical signals is fallen down remarkably when the welding defects occur. Correspondingly, the high-frequency component of the signals is almost decline to zero simultaneously. Considering that wavelet analysis can decompose the optical signals, extract the characteristic information of the signals and define the defects location accurately, it can be used to implement process-control of laser welding.