Study on penetration recognition method of laser welding for Q235 steel based on signal mesoscopic extraction and statistical analysis

Laser welding penetration detection is an important link to realize intelligent welding manufacturing, but how to penetrate shelter objects above the keyhole, break through signal effective extraction of accurate location under the mesoscopic scale, which in the penetration feature area of the keyhole, and effectively analyze the complex optical detection signal, these are the main technical difficulties to reliably identify the penetration behavior of laser welding by direct test method. In this paper, a new method of laser welding penetration recognition of Q235 steel is studied. It includes, using the fluorescence radiation source of the keyhole inner wall as the direct detection signal, penetrating shelter objects and making the keyhole internal morphological feature perspective by using the characteristics of different material spectrum segments, obtaining fluorescence doubling image of the keyhole inner wall by using optical imaging, and making the effective penetration recognition signal maximized and enhanced by locating the sensing chip directly to the penetration feature area by pinhole uptake and mesoscopic location methods, under certain welding conditions, finally, identifiing the detection signal by statistical probability feature recognition, which realized the abstract separation of welding penetration feature information. And the method has obtained the reliable penetration detection results.