过程(计算)
材料科学
统计模型
工程制图
统计分析
几何学
机械工程
过程建模
计算机科学
算法
统计学习
作者
Michael Hofmann,Thomas D Mayer,Francesca Friso,Rene Radis,Christian Kontermann,Falk Müller,Matthias Oechsner
标识
DOI:10.1016/j.matdes.2026.115459
摘要
Melt pool geometry strongly affects process stability and part quality in Laser Powder Bed Fusion (PBF-LB/M), shaping microstructure evolution, porosity, and mechanical properties. Although analytical models, physics-based simulations, and data-driven methods have been used to predict melt pool dimensions, most available datasets combine numerous literature sources that differ in machine configuration, parameter ranges, and measurement techniques. Such inconsistency introduces noise and bias that can degrade model accuracy. This work presents a consistent single-source dataset of categorized melt pool geometries from 316L single-track experiments performed under controlled conditions while systematically varying laser power, scan speed, spot size, and powder layer thickness. Using 677 tracks, multilinear regression models were developed to predict melt pool dimensions and generate process maps that quantify how processing parameters govern melt pool shape and the onset of instability. Model performance and generalizability were assessed through comparison with published experiments, analytical scaling laws, and finite-element simulations. The proposed models outperform existing analytical and numerical approaches across broad conditions, though prediction of literature data remains constrained by variability among PBF-LB/M systems and measurement practices. The dataset and models provide a consistent foundation to support future research in melt pool modeling, process simulation and optimization for PBF-LB/M.
科研通智能强力驱动
Strongly Powered by AbleSci AI