研究目的
To develop an efficient machine learning-based method to assist in process design and optimization of the laser powder bed fusion (L-PBF) process for stainless steels.
研究成果
The GP model developed in this study effectively predicts the remelted depth in L-PBF processes for 316L and 17-4 PH stainless steels, with a small prediction error. The process design maps and revised normalized enthalpy criteria provided can assist in optimizing processing parameters to avoid defect formation.
研究不足
The study is limited to two types of stainless steels (316L and 17-4 PH) and does not explore the effects of other materials or processing parameters beyond laser power and scan speed. The computational cost of CFD simulations is high, and the GP model's predictions outside the training range may have larger uncertainties.
1：Experimental Design and Method Selection:
A Gaussian process (GP) regression model is developed to predict the remelted depth of single tracks in L-PBF. The model is trained using both simulation and experimental data.
2：Sample Selection and Data Sources:
Data from computational fluid dynamics (CFD) simulations and experimental observations of 316L and 17-4 PH stainless steels are used.
3：List of Experimental Equipment and Materials:
Laser powder bed fusion setup, 316L and 17-4 PH stainless steel powders.
4：Experimental Procedures and Operational Workflow:
Single track simulations are conducted using a CFD model to generate training data for the GP model. The GP model then predicts remelted depths at various laser power and scan speed combinations.
5：Data Analysis Methods:
The mean absolute prediction error (MAPE) is used to assess the performance of the GP model.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
