Deviation Status Detection Based On Visual Characteristics In Laser Welding Of Sandwich Constructions

Laser welded steel sandwich constructions have gained great popularity for their good properties of specific strength and vibration absorption. Meanwhile, their welding quality is a critical point for their comprehensive performance. The special welding mode for the T-joint of I-core sandwich plates proposes new challenges such as that the incident laser should be aligned with the web plate.In order to develop a novel method of deviation detection based on the visual sensing for laser welding of I-core sandwich plates, an experimental system consisting of TLFCO2 laser welding equipment, a shielding gas(Helium) device and a CMOS high speed camera was built up.The influence of the increasing of deviation on weld pool is studied on the basis of weld appearance characteristics as well as the cross section macro-structure of weld joint. It is found that the bottom of weld pool would collapse if the deviation increases to a certain extent. The changes of size and fluid flow characteristics of molten pool due to the collapse can be detected with the help of weld pool images.The edge detecting of weld pool is very difficult as a result of the interference of plume plasma, therefore, a modified Snake algorithm is adopted for image processing. Afterwards, three characteristic parameters, including length and area of the stern of the weld pool and difference of area are calculated. Furthermore, the way in which the weld pool image is affected by deviation is found and finally the monitoring method based on the characteristics of weld pool stern is designed.The downward plume, which is the direct phenomena of the keyhole rupture when the deviation becomes intolerant, is almost unable to monitor. Therefore, the upward plume carrying the relevant information is captured by high speed camera. The characteristics of instantaneous plasma plume are calculated and their process parameters are extracted statistically both in time domain and frequency field. Furthermore, principal component analysis(PCA) is employed to synthesize the information of all feature parameters and a complex indicator is employed to evaluate the deviation status. Experimental results demonstrate that the monitoring method based on the characteristics of vapor plume is convenient, effective and promising in production.