In view of the automatic detection problem of silicon nitride removal on the surface of graphite boat,the change law of spectral signal in the process of laser cleaning is studied,and the laser cleaning state of the surface of graphite boat is judged in real time.Experiment summarizes the typical graphite boat surface state type,through the detection and analysis of laser induced spectral curve,screen out the key characteristic line,study the variation of characteristic line intensity rule,compared to the conventional lim-it learning machine(ELM)and beluga optimization algorithm(BWO-ELM)classification accuracy,automatic classification of graphite boat surface laser cleaning status.The findings suggest that the selected feature lines include Si(1)390.49 nm,N(1)396.02 nm,C(Ⅱ)588.84 nm.The variation law of the peak intensity of the characteristic spectral lines fully reflects the element composi-tion and content changes of the matrix graphite boat and the surface silicon nitride.The Beluga algorithm optimizes the conventional ELM algorithm,which effectively eliminates the inherent randomness in the ELM classification algorithm,and greatly improves the classification accuracy.When the number of iterations increases to 100,the classification accuracy of the beluga whale optimization algorithm can reach more than 95％.The results are crucial to ensure the laser cleaning effect on the surface of graphite boat,and it is also the key technology to realize the automation of laser cleaning.
维普
万方数据