Optimization of ultrashort pulse laser micromachining process based on machine learning and genetic algorithm

Chenchong Wang*, Zhen Zhang, Wei Xu

Northeastern University, Shenyang, 110819, China

EXTENDED ABSTRACT: As an advanced turbine blade cooling technology, film cooling holes are widely studied and applied. However, many defects in the machining process of film cooling holes have become an important factor for blade failure. Recently, the rapid development and application of ultrashort pulse laser provide a solution to this problem. The extremely short time interval of the interaction of a ultrashort pulse laser and a material can enable processing without thermal damage. However, due to the complexity of the machining process, it is very difficult to optimize the ultrashort pulse micromachining process through experiments or physical models. Therefore, based on IC10 directionally solidified superalloy, 54 groups of experimental databases of ultrashort pulse laser trepan drilling are established in this work. The key laser parameters and non-thermal parameters are selected as inputs, and the taper and machining efficiency of micro-holes are selected as prediction targets. Then, the prediction models of ultrashort pulse laser micromachining in IC10 superalloy are successfully established by using a variety of machine learning algorithms. The prediction model is connected with multi-objective high-throughput genetic algorithm (NSGAⅡ), and the process better than the original data set is obtained. The experimental verification of the designed process shows that the optimized process has the characteristics of no thermal damage, small taper and high machining efficiency. Based machine learning and high-throughput optimization algorithm, this study realizes the effective process optimization of ultrashort pulse laser trepan drilling with low cost and high efficiency using a small experimental data set. The process optimization method of machine learning coupled with high-throughput optimization algorithm proposed in this study can further promote the application of ultrashort pulse laser processing in related fields.