Process optimization of laser melting deposition for DZ125L based on machine learning

Authors:

Abstract
Purpose: The purpose of this paper is to accurately predict and optimize the process parameters of laser metal deposited Ni-based superalloys using machine learning (ML). Design/methodology/approach: Orthogonal experiments were carried out to obtain the raw data and determine the effects of parameters on the geometric characteristics of the deposition layer. A support vector regression (SVR) model based on the Crested Porcupine Optimizer (CPO) algorithm for predicting the geometric characteristics of the deposition layer is proposed. The Multi-objective Red-billed Blue Magpie Optimizer (MORBMO) algorithm was adopted to optimize the process parameters. Findings: Laser power has the greatest effect on depth and width, and scanning speed has the most pronounced impact on height. Both scanning speed and powder feeding rate have a significant impact on the wetting angle. Compared with the classical SVR model, the CPO-SVR improves the R2 values of height, depth, width and wetting angle by 20.72%, 20.16%, 11.27% and 17.78%, respectively, and reduces the MAPE by 36.43%, 8.53%, 3.95% and 9.83%, respectively. The optimized parameters were determined to be a laser power of 1165 W, a scanning speed of 9 mm/s and a powder feed rate of 0.8 r/min. Then, the relative density of the bulk specimen fabricated using these parameters was increased by 1.03% compared to that prepared with the process parameters optimized by previous process experiments. Originality/value: The findings of this study demonstrate the potential of the proposed ML method based on CPO-SVR and MORBMO algorithms for the prediction and optimization of laser-melting-deposited Ni-based superalloys.

This publication has 39 references indexed in Scilit: