目的 构建一种分层孔隙率预测模型,通过预测单层孔隙率为SLM成形前的缺陷预判以及工艺参数优化提供决策支持。方法 首先,以GH4169高温合金为材料设计正交实验并制备SLM试样,使用CT扫描部分试样获取内部分层图像,提出一种以实测致密度为目标的BlackHat-Otsu算法,从CT图像中精准提取每一层的孔隙特征;其次,基于工艺参数构建反映单层状态的一维特征向量,以及融合当前层与上下层工艺信息的二维特征矩阵;最后,分别针对一维特征和二维特征,建立了SVR模型、BP神经网络模型,并设计了融合2类特征的CNN模型。结果 采用所提出的BlackHat-Otsu算法提取到的分层孔隙率与整体实测值误差小于0.5%;在3种预测模型中,SVR模型在测试集上的最佳预测R2为0.840 5,BP神经网络经参数寻优后,测试集R2提升至0.883 1,双通道CNN模型的测试集R2达到0.904 5。结果表明,引入层间工艺参数交互信息构建的二维特征能有效提升模型预测能力。结论 通过融合工艺参数与CT图像数据,并利用能够捕捉层间时空关联的深度学习模型,可以实现对SLM成形件分层孔隙率的更高精度预测;该方法将孔隙率预测从构件整体细化至单层尺度,为实现SLM成形质量的精细化、前瞻性控制提供了有效的技术途径。
Abstract
The work aims to establish a layer-wise porosity prediction model which can provide decision support for defect pre-judgment before SLM forming and process parameter optimization by predicting the porosity of single layers. Firstly, GH4169 superalloy was used as the material to design orthogonal experiments and fabricate SLM specimens. Partial specimens were scanned by CT to obtain internal layer-wise images, and a BlackHat-Otsu algorithm targeting measured relative density was proposed to accurately extract the porosity features of each layer from the CT images. Secondly, based on process parameters, a one-dimensional feature vector reflecting the state of a single layer and a two-dimensional feature matrix integrating the process information of the current layer, upper and lower layers were constructed. Finally, a Support Vector Regression (SVR) model and a Backpropagation (BP) Neural Network model were established for the one-dimensional and two-dimensional features respectively, and a Convolutional Neural Network (CNN) model fusing the two types of features was designed. The error between the layer-wise porosity extracted by the proposed BlackHat-Otsu algorithm and the overall measured value was less than 0.5%. Among the three prediction models, the SVR model achieved an optimal R2 of 0.840 5 on the test set. After hyperparameter optimization, the R2 of the BP Neural Network model increased to 0.883 1 and the dual-channel CNN model reached a test set R2 of 0.904 5. The results indicated that the two-dimensional features constructed by introducing inter-layer process parameter interaction information could effectively improve the model's prediction capability. By fusing process parameter data and CT image data, and utilizing deep learning models capable of capturing inter-layer spatiotemporal correlations, higher-precision prediction of layer-wise porosity in SLM parts can be achieved. This method refines porosity prediction from the overall component to the single-layer scale, providing an effective technical approach for realizing refined and proactive control of SLM forming quality.
关键词
选区激光熔化（SLM） /
GH4169 /
工艺参数 /
CT图像 /
孔隙率 /
分层预测
Key words
selective laser melting (SLM) /
GH4169 /
process parameters /
CT image /
porosity /
layer-wise prediction
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 卢秉恒. 增材制造技术——现状与未来[J]. 中国机械工程, 2020, 31(1): 19-23.
LU B H.Additive Manufacturing—Current Situation and Future[J]. China Mechanical Engineering, 2020, 31(1): 19-23.
[2] 李涤尘, 贺健康, 田小永, 等. 增材制造:实现宏微结构一体化制造[J]. 机械工程学报, 2013, 49(6): 129-135.
LI D C, HE J K, TIAN X Y, et al.Additive Manufacturing: Integrated Fabrication of Macro/Microstructures[J]. Journal of Mechanical Engineering, 2013, 49(6): 129-135.
[3] DU C B, ZHAO Y H, JIANG J C, et al.Pore Defects in Laser Powder Bed Fusion: Formation Mechanism, Control Method, and Perspectives[J]. Journal of Alloys and Compounds, 2023, 944: 169215.
[4] 求晓玲. 不锈钢选区激光熔化在线监测及缺陷预测研究[D]. 大连: 大连理工大学, 2022: 3-5.
QIU X L.Online Monitoring and Defect Prediction of Selective Laser Melting of Stainless Steel[D]. Dalian: Dalian University of Technology, 2022: 3-5.
[5] 周运龙, 马毅, 管迎春. 面向航空发动机高性能制造的激光选区熔化技术研究进展[J]. 航空学报, 2024, 45(13): 629508.
ZHOU Y L, MA Y, GUAN Y C.Research Progress on Laser Selective Melting Technology for High-Performance Manufacturing of Aero-Engines[J]. Acta Aeronautica et Astronautica Sinica, 2024, 45(13): 629508.
[6] 谢文浩, 李俊, 周建新, 等. 激光选区熔化IN718合金的熔池演化过程数值模拟[J]. 特种铸造及有色合金, 2024, 44(4): 490-495.
XIE W H, LI J, ZHOU J X, et al.Numerical Simulation of Melt Pool Evolution in Selective Laser Melting of IN718 Alloy[J]. Special Casting & Nonferrous Alloys, 2024, 44(4): 490-495.
[7] WU B, JI X Y, ZHOU J X, et al.In Situ Monitoring Methods for Selective Laser Melting Additive Manufacturing Process Based on Images—A Review[J]. China Foundry, 2021, 18(4): 265-285.
[8] 王昊, 吕俊霞, 刘兴迪, 等. 激光选区熔化GH4169合金低周疲劳小裂纹行为研究[J]. 特种铸造及有色合金, 2025, 45(12): 1802-1807.
WANG H, LYU J X, LIU X D, et al.Low-Cycle Fatigue Minor Crack Behavior of GH4169 Alloy by Selective Laser Melting[J]. Special Casting & Nonferrous Alloys, 2025, 45(12): 1802-1807.
[9] KARNA S, YUAN L, ZHANG T Y, et al.Microstructural Analysis and Defect Characterization of Additively Manufactured AA6061 Aluminum Alloy via Laser Powder Bed Fusion[J]. Journal of Materials Science & Technology, 2025, 219: 288-306.
[10] 代俊林, 吴世品, 张宇, 等. 增材制造金属材料的疲劳性能研究进展[J]. 精密成形工程, 2024, 16(1): 1-13.
DAI J L, WU S P, ZHANG Y, et al.Research Progress on Fatigue Properties of Additive Manufactured Metal Materials[J]. Journal of Netshape Forming Engineering, 2024, 16(1): 1-13.
[11] 周圣智, 涂先猛, 计效园, 等. 可见光监测与CT检测融合驱动的高温合金SLM成形孔隙演变研究[J]. 铸造技术, 2023, 44(7): 640-648.
ZHOU S Z, TU X M, JI X Y, et al.Research on the Evolution of Pores in Superalloy SLM Forming Driven by Fusion of Visual Light Monitoring and CT Detection[J]. Foundry Technology, 2023, 44(7): 640-648.
[12] KOELBLIN J, HASTIE J C, KARTAL M E, et al.Deformation of AlSi10Mg Parts Manufactured by Laser Powder Bed Fusion: In-Situ Measurements Incorporating X-Ray Micro Computed Tomography and a Micro Testing Stage[J]. Procedia Structural Integrity, 2022, 35: 168-172.
[13] GENG L C, WU W W, SUN L J, et al. Damage Characterizations and Simulation of Selective Laser Melting Fabricated 3D Re-Entrant Lattices Based on In-Situ CT Testing and Geometric Reconstruction[J]. International Journal of Mechanical Sciences, 2019, 157/158: 231-242.
[14] ZHANG Z J, WU Y J, WANG Z M, et al.Recommendation of SLM Process Parameters Based on Analytic Hierarchy Process and Weighted Particle Swarm Optimization for High-Temperature Alloys[J]. Materials, 2023, 16(16): 5656.
[15] 涂先猛, 彭东剑, 陈威, 等. 数据驱动选区激光熔化成形性能优化与智能工艺研究进展[J]. 铸造, 2024, 73(11): 1487-1505.
TU X M, PENG D J, CHEN W, et al.Research Progress on Data-Driven Selective Laser Melting Forming Performance Optimization and Intelligent Process Recommendation System[J]. Foundry, 2024, 73(11): 1487-1505.
[16] YAN W T, QIAN Y, GE W J, et al.Meso-Scale Modeling of Multiple-Layer Fabrication Process in Selective Electron Beam Melting: Inter-Layer/Track Voids Formation[J]. Materials & Design, 2018, 141: 210-219.
[17] WEI P, WEI Z Y, CHEN Z, et al.Thermal Behavior in Single Track during Selective Laser Melting of AlSi10Mg Powder[J]. Applied Physics A, 2017, 123(9): 604.
[18] 伍诗伟, 贾岳飞, Hou Yi Chia, 等. 抗开裂高温高熵合金激光粉末床熔融工艺参数优化[J]. 精密成形工程, 2025, 17(12): 45-53.
WU S W, JIA Y F, CHIA H Y, et al.Processing Parameter Optimization in Laser Powder-Bed Fusion of a Cracking-Resistant High-Temperature High-Entropy Alloy[J]. Journal of Netshape Forming Engineering, 2025, 17(12): 45-53.
[19] CHEN Y Y, WANG H Z, WU Y, et al.Predicting the Printability in Selective Laser Melting with a Supervised Machine Learning Method[J]. Materials, 2020, 13(22): 5063.
[20] LIU B Q, FANG G, LEI L P, et al.Predicting the Porosity Defects in Selective Laser Melting (SLM) by Molten Pool Geometry[J]. International Journal of Mechanical Sciences, 2022, 228: 107478.
[21] CHEN X Y, LIAO W H, YUE J S, et al.Unveiling the Layer-Wise Dynamics of Defect Evolution in Laser Powder Bed Fusion: Insights for In-Situ Monitoring and Control[J]. Additive Manufacturing, 2024, 94: 104414.
[22] 宋剑锋, 宋有年, 王文武, 等. 金属粉末选区激光熔化成形表面粗糙度预测及控制方法研究[J]. 中国激光, 2022, 49(2): 0202008.
SONG J F, SONG Y N, WANG W W, et al.Prediction and Control on the Surface Roughness of Metal Powder Using Selective Laser Melting[J]. Chinese Journal of Lasers, 2022, 49(2): 0202008.
[23] MALASHIN I, MARTYSYUK D, TYNCHENKO V, et al.Predicting Defects in SLM-Produced Parts Based on Melt Pools Clustering Analysis[J]. The International Journal of Advanced Manufacturing Technology, 2024, 134(3): 1169-1178.
[24] FENG F, ZHU T, YANG B, et al.A Physics-Informed Neural Network Approach for Predicting Fatigue Life of SLM 316L Stainless Steel Based on Defect Features[J]. International Journal of Fatigue, 2024, 188: 108486.
[25] TU X M, ZHOU S Z, JI X Y, et al.A Universal Relative Density Equation: Exploring the Physical Mechanisms of Porosity Entrapment in Laser Powder Bed Fusion[J]. Engineered Science, 2026, 39: 2001.
[26] SUN Y T, DING S F, ZHANG Z C, et al.An Improved Grid Search Algorithm to Optimize SVR for Prediction[J]. Soft Computing, 2021, 25(7): 5633-5644.
基金
新材料重大专项项目（2025ZD0609500）; 国家自然科学基金（52275337,52090042,52674498）
PDF(2760 KB)
渝公网安备 50010702501719号 渝ICP备15012534号-4
