An Optimization Method for Laser Cutting Process Parameters Based on GPR and Seq2Seq

The setting of the laser cutting parameter is struggled for its sensitivity to environment. In this paper, a systematic method for optimizing parameters in laser cutting process is proposed, which combines gaussian process regression (GPR) and Sequence-to-Sequence (Seq2Seq) networks. Firstly, a laser cutting process simulator is developed based on GPR. The simulator is used to expand the dataset for machining parameters, enabling the handling of raw machining data with uneven distribution and limited work samples. Finally, a Seq2Seq-based model is proposed to forecast the laser cutting processing parameters under high-power (20,000 W), with the input of low-power machine operation parameter. The experimental results demonstrate the ability of the method to predict the parameters of high-power machines. The method's ability to predict the parameters of high-power machines is demonstrated by the experimental results, with a mean squared error of prediction for parameters of processed stainless steel less than 0.5 and for processed carbon steel less than 2.6. The proposed method can handle the optimization of the processing parameters for new power machines.