Identification of high-power fiber laser welding penetration based on fuzzy clustering algorithm

Weld penetration is an important standard which evaluates the laser welding quality. However, it is difficult to directly measure the weld penetration during laser welding. Controlling the weld penetration can only be carried out through an indirect estimation of weld penetration by other sensing signals. A novel approach of weld penetration status identification based on molten pool infrared image processing is proposed during butt joint high-power fiber laser welding of type 304 austenitic stainless steel plates with laser power of 6 kW. An infrared high speed camera was used to capture the dynamic infrared images of molten pool. Characteristic parameters of a molten pool image were extracted by image processing, and fuzzy clustering algorithm was used to explore the relation between the molten pool parameters and the weld penetration status. An identification model of weld penetration status was established by using the fuzzy C-mean (FCM) and Gustafson-Kessel (GK) algorithm, respectively. Welding experiments confirmed that the molten pool surface characteristic had a close inherent relationship with the weld penetration, and the weld penetration status could be estimated by the proposed GK fuzzy clustering identification model, which could provide an experimental basis for detection and control of weld penetration.