Tunable simultaneous Bayesian optimization of hole taper and processing time in QCW laser drilling

Abstract

Laser drilling (LD) finds extensive applications in the manufacturing industry, and the improvement of LD quality has been actively pursued by researchers and engineers. However, the large number of tunable processing parameters makes the optimization of LD extremely difficult. In this study, we introduce an innovative approach by applying Bayesian optimization (BO) to tackle the long-standing challenge of multi-dimensional optimization in the field of LD. Moreover, the drilling efficiency was integrated into the optimization objective, representing a crucial metric for practical manufacturing that has frequently been overlooked. As a proof-of-principle demonstration, the drilled-hole taper and processing time were simultaneously optimized relative to four processing parameters in quasi-continuous-wave laser spiral drilling of stainless steel. Results reveal that BO can effectively and efficiently optimize the LD process in a four-dimensional parameter space. Furthermore, we show that the balance between drilling quality and efficiency can be tuned by customizing the objective function. By employing a quality-oriented objective function, the hole taper was significantly improved from 4.3° to 0.8° within just 20 iterations. Meanwhile, with an efficiency-oriented objective function, the processing time was reduced from 13.5 s to 1.6 s. The physical mechanisms underlying how different parameters influenced the drilling process were also discussed. The results in this study reveal the power of BO in multi-dimensional optimization of LD, and the methodology can readily be applied to other laser machining applications.