DBSCAN clustering model for parameter inversion using laser cutting edge morphology characteristic in Zr-4 alloy

Laser cutting, as an efficient, high-quality, non-contact metal cutting technology, has the potential to replace traditional manufacturing processes for Zircaloy-4 (Zr-4 alloy) cladding materials of nuclear reactors. However, the stability of the laser cutting process has a significant impact on the quality and service safety of Zr-4 alloy and its key components in nuclear engineering. Therefore, this work first proposes a novel approach for recognizing abnormal fluctuations in the laser cutting process using cutting edge morphology characteristics, thereby ensuring the stability of the process. Firstly, the cutting edge images are captured using an ultra-depth of field microscopy, and the edge morphology feature parameters (the length (L) of vertical striations, the inclination angle (θ) of inclined striations, and surface roughness (Ra)) are measured. Secondly, a density-based spatial clustering of applications with noise (DBSCAN) model for process parameter inversion is established with only 2 cutting edge feature parameters (L and θ, Comprising 386 pairs of data) as model input. Then, a three-standard fusion method is proposed to optimize the model and the model can identify process parameters (laser power, defocus amount, cutting speed, and auxiliary gas pressure, etc.) abnormal fluctuations at 80% accuracy. Finally, by incorporating Ra as an additional input feature parameter along with L and θ, the model can identify process parameters abnormal fluctuations at 100% accuracy. This work effectively recognizes abnormal fluctuations of process parameters during laser cutting of Zr-4 cladding materials, thus benefiting the control of these fluctuations and quality management in metal sheet laser cutting.