In the laser welding, the welding fusion depth can characterize the weld bonding strength and affect the welding quality. An accurate measurement of the keyhole depth can prevent potential welding defects and improve the laser welding quality. However, conventional welding monitoring methods cannot achieve accurate keyhole depth measurements. The measuring beam of the optical coherence tomography (OCT) can be coaxial with the machining laser and has horizontal and longitudinal resolutions at the micron level. It is a proven detection method that can measure the keyhole depth in a nondestructive manner during welding, thus achieving a high-precision measurement of the keyhole depth. However, the multiple reflections of the OCT detecting beams in the metal keyhole, as well as the welding splash during the measurement process and other factors, may lead to the measurement signals of OCT not only containing the signals reflected from the bottom or side wall of the keyhole but also containing the metal vapor and welding splash reflections as well as the signals reflected from the inner wall of the keyhole. Consequently, OCT cannot accurately measure the depth of the bottom of the keyhole. To solve this problem, a method of laser welding keyhole depth detection using polarimetric OCT is proposed in this study, which eliminates the influence of multiple reflected lights on the keyhole depth measurement and improves the measurement accuracy.
Based on the principle that the phase retardation of light reflected on a metal surface varies with the incidence angle, the polarization sensitive OCT (PS-OCT)system is used to image a laser-welded keyhole. Circularly polarized light is incident at the bottom of the vertical keyhole, and a phase retardation of π/2 is considered as the reference value. The phase retardation obtained from the keyhole measurements using PS-OCT is compared with the reference value. The PS-OCT signal corresponding to the phase retardation, which is the same as the reference value, is marked as the effective signal, and the remaining PS-OCT signal generated by multiple reflections is marked as an invalid signal. An effective PS-OCT signal is used to accurately calculate the keyhole depth. Two keyhole models of copper and aluminum are established, and a ray-tracing method is used to simulate the proposed method. Based on the simulation, a keyhole imitation sample is designed and the OCT is used to illustrate the multiple reflection artifacts of the sample. Subsequently, a PS-OCT experimental system is built, and M-scan- and B-scan measurements are performed at different positions of the samples using polarimetric OCT. The depth screened using the proposed method is compared with that using the commonly used 80-percentile filtering method.
Simulation parameters are determined based on the X-ray images (Table 1), and a ray-tracing simulation analysis is performed using a software [Fig. 4(a)]. The results demonstrate that the proposed method can significantly improve the accuracy of the keyhole depth calculation by screening the PS-OCT signals [Fig. 4 (c)] and is better than the current commonly used 80-percentile filtering method [Fig. 4 (b)]. The keyhole sample is designed [Fig. 6 (a)], and multiple artifacts in the keyhole are first described based on the scanning sample of an ordinary OCT system [Fig. 6 (b)]. Based on the established PS-OCT experimental system, M-scan imaging (Fig. 7) and B-scan imaging (Fig. 8) are performed at different locations of the samples, and the original PS-OCT images, corresponding phase retardation images, and PS-OCT images screened in this study are obtained. It can distinguish between single reflected signals, multiple reflected signals, and stray light signals in the keyhole, which effectively improves the accuracy of the keyhole depth measurement.
The existence of multiple-reflection artifacts is proven in the simulation, and the measurement errors of the OCT keyhole depth are 16.8% and 45.6% for copper and aluminum, respectively. The measurement errors of the keyhole depth are 3.27% and 6.42% for copper and aluminum, respectively, which are determined by using the 80-percentile filtering method. The errors in the keyhole depth measured by using this method are 1.05% for copper and 0.12% for aluminum. The designed keyhole samples are imaged using the established polarimetric OCT system. It is proven that multi-reflection artifacts are eliminated after screening by the proposed method, and a real image is obtained, which can reflect the real structure of the sample. The keyhole depth measurement error obtained using the original PS-OCT image combined with the 80-percentile filtering method is 23%, whereas that obtained using the proposed method is only 1.6%.
