1.中北大学 机电工程学院,山西 太原 030051
2.中国兵器装备集团自动化研究所有限公司 智能制造事业部,四川 绵阳 621000
邮箱:yjp123@nuc.edu.cn
收稿:2025-12-20,
修回:2026-08-26,
录用:2026-05-13,
网络首发:2026-09-09,
纸质出版:2026-09-30
移动端阅览
史慧芳,尹建平,伊建亚.基于数据增强与XGBoost-MLP集成学习的炸药散粒体成型密度预测方法[J].兵工学报,2026,47(09):251124.
SHI Huifang,YIN Jianping,YI Jianya.A Method for Predicting the Formed Density of Explosive Granules Based on Data Augmentation and XGBoost-MLP Ensemble Learning[J].ACTA ARMAMENTARII,2026,47(09):251124.
史慧芳,尹建平,伊建亚.基于数据增强与XGBoost-MLP集成学习的炸药散粒体成型密度预测方法[J].兵工学报,2026,47(09):251124. DOI: 10.12382/bgxb.2025.1124.
SHI Huifang,YIN Jianping,YI Jianya.A Method for Predicting the Formed Density of Explosive Granules Based on Data Augmentation and XGBoost-MLP Ensemble Learning[J].ACTA ARMAMENTARII,2026,47(09):251124. DOI: 10.12382/bgxb.2025.1124.
在武器产品压装领域,产品成型密度与武器装备毁伤效能密切相关,是压制成型质量的重要指标之一。炸药散粒体压制成型过程具有多变量、参数不确定、数据难获取等特点,传统的数值仿真方法难以准确得到产品成型密度,数据驱动方法因数据依懒性强亦无法精准预测成型密度。为此,提出基于数据增强与极端梯度提升-多层感知器(Extreme Gradient Boosting-Multi-Layer Perceptron, XGBoost-MLP)集成学习的炸药散粒体成型密度预测方法。采用基于随机森林的数据增强方法对原始数据进行扩增,并通过物理一致性分析确保扩增后的数据符合物理机理;提出XGBoost-MLP集成学习方法,构建炸药散粒体压制成型密度预测模型,并采用
K
折交叉法和早停策略对模型训练过程进行强化;设计并开展炸药散粒体压制成型与密度测试试验,对所提方法进行验证。在此基础上,将所提方法与XGBoost-MLP、数据增强-XGBoost、数据增强-MLP方法进行对比分析。研究结果表明,基于随机森林的数据增强方法能够有效缓解因样本匮乏而导致的模型过拟合问题,在一定程度上克服了压制成型数据的小样本难题;所提方法在平均绝对误差、均方误差和平均绝对百分比误差上均取得最低值,分别为0.017 5、0.000 555和1.03%,实现了对压制成型密度的高精度预测。
In the field of weapon product pressing, the forming density of products is closely related to the damage effectiveness of weapon equipment and is one of the important indicators of press forming quality. The press forming process of granular explosive material is characterized by multiple variables, uncertain parameters, and difficult data acquisition. The traditional numerical simulation methods often fail to accurately determine the forming density of products, and the data-driven methods are also unable to precisely predict the forming density of products due to strong data dependency. To address this issue, a granular explosive material forming density prediction method based on data augmentation and extreme gradient boosting-multi-layer perceptron (XGBoost-MLP) ensemble learning is proposed. Firstly, the original dataset is augmented using a data augmentation method based on the random forest model, and the physical consistency analysis is conducted to ensure that the augmented data conforms to the physical mechanism . Secondly, an XGBoost-MLP ensemble learning method is proposed to construct a granular explosive material press forming density prediction model, and the K-fold cross-validation method and the early stopping strategies are applied to strengthen the model training process. Finally, the experiments on granular explosive material press forming and density testing are designed and conducted to validate the proposed method. The results show that the random forest-based data augmentation method can effectively alleviate the model overfitting caused by insufficient samples and, overcome the small sample challenge in press forming data to some extent. Based on this, the comparation with other methods shows that the proposed method achieves the lowest values in MAE, MSE, and MAPE, which are 0.017 5, 0.000 555 and 1.03%, respectively, realizing the high-precision prediction of press forming density. This study provides a foundation for forming density prediction in the field of weapon product pressing and is of great significance for optimizing the forming process and improving the forming quality.
王娜 . 散粒体压制过程压制力智能调控技术研究 [D]. 沈阳 : 沈阳航空航天大学 , 2024 .
Wang N . Research on intelligent control technology of pressing force in the process of particle compression [D]. Shenyang : Shenyang Aerospace University , 2024 . (in Chinese)
吴晓阳 , 陶俊 , 王晓峰 , 等 . HMX基散粒体模压成型的力学行为 [J]. 含能材料 , 2025 , 33 ( 9 ): 1113 - 1122 .
Wu X Y , Tao J , Wang X F , et al . Mechanical behavior of HMX‑Based granular molding [J]. Chinese Journal of Energetic Materials , 2025 , 33 ( 9 ): 1113 - 1122 . (in Chinese)
陈浩泽 . 基于有限元分析的散粒体压制过程多尺度力学特征分析 [D]. 北京 : 北京化工大学 , 2024 .
Chen H Z . Multiscale mechanical characteristics analysis of granular materials compaction process based on finite element [D]. Beijing : Beijing University of Chemical Technology , 2024 . (in Chinese)
Ganthaler E , Mustafa S , Peer A . Comparison of different methodologies for estimating local density in particle-based simulations [J]. Computational Particle Mechanics , 2025 , 12 : 1217 – 1231 .
Zhang W , Lin C K , Wu W C , et al . Evolution of multiscale mechanical properties and densification during Fe/Al mixed metal powder compaction—MPFEM [J]. Powder Technology , 2026 , 467 : 121518 .
Rajesh R , Vel E K , Priya S B M , et al . Prediction of wind turbine blade deterioration impact on aerodynamic performance using proper orthogonal decomposition and deep reinforcement learning [J]. International Journal of Heat and Fluid Flow , 2026 , 117 : 110125 .
李琳琳 , 张欲立 , 张宏光 , 等 . 分步压装药工艺及装药参数的实验研究 [J]. 兵工自动化 , 2009 , 28 ( 10 ): 33 - 35 .
Li L L , Zhang Y L , Zhang H G , et al . Experiment research on procedure press-charge detonator technology and charge parameter [J]. Ordnance Industry Automation , 2009 , 28 ( 10 ): 33 - 35 . (in Chinese)
田丽燕 , 徐更光 , 王廷增 . 散粒体炸药压装成型过程分析 [J]. 火炸药学报 , 2002 , 25 ( 2 ): 23 - 24 .
Tian L Y , Xu G G , Wang Y Z . Analysis of the pressing process of several granular explosives [J]. Chinese Journal of Explosives & Propellants , 2002 , 25 ( 2 ): 23 - 24 . (in Chinese)
关通 , 张向荣 , 温永昕 , 等 . 工艺参数对惰性代料高压熔铸成型质量的影响 [J]. 火炸药学报 , 2023 , 46 ( 10 ): 920 - 927 .
Guan T , Zhang X R , Wen Y X , et al . Effects of process parameters on the casting quality of high-pressure molding simulant composites [J]. Chinese Journal of Explosives & Propellants , 2023 , 46 ( 10 ): 920 - 927 . (in Chinese)
Han L , Elliott J , Bentham A , et al . A modified drucker-prager cap model for die compaction simulation of pharmaceutical powders [J]. International Journal of Solids and Structures , 2008 , 45 ( 10 ): 3088 - 3106 .
Liu X Q , Gao Y , Li H , et al . Density dependent constitutive model for Bi-2212 powder compression deformation [J]. Physica Scripta , 2024 , 99 : 065950 .
张涛 , 赵北君 , 朱世富 , 等 . PBX粉末成形的数值模拟研究 [J]. 材料工程 , 2009 ( 5 ): 68 - 72 .
Zhang T , Zhao B J , Zhu S F , et al . Numerical simulation study of PBX powder-forming [J]. Journal of Materials Engineering , 2009 ( 5 ): 68 - 72 . (in Chinese)
岳晓媛 , 张会锁 , 黄鸿鑫 , 等 . 压药工艺参数对成型药柱质量的影响研究 [J]. 兵工自动化 , 2021 , 40 ( 1 ): 84 - 87 , 96 .
Yue X Y , Zhang H S , Huang H X , et al . Study on influence of explosive pressing technique parameter on quality of press-forming Grain [J]. Ordnance Industry Automation , 2021 , 40 ( 1 ): 84 - 87, 96 . (in Chinese)
Ku Q , Zhao J D , Mollon G , et al . Compaction of highly deformable cohesive granular powders [J]. Powder Technology , 2023 , 421 : 118455 .
夏焕雄 , 李康 , 高丰 , 等 . 基于凝固前沿演变特征的熔铸装药成型工艺参数智能优化 [J]. 兵工学报 , 2024 , 45 ( 9 ): 2936 - 2950 .
Xia H X , Li K , Gao F , et al . Intelligent optimization for forming quality of melt-cast explosives based on the evolution characteristics of solidification front [J]. Acta Armamentarii , 2024 , 45 ( 9 ): 2936 - 2950 . (in Chinese)
戚长松 , 余忠华 , 侯智 , 等 . 基于CART决策树的复杂生产过程质量预测方法研究 [J]. 组合机床与自动化加工技术 , 2010 ( 3 ): 94 - 97, 102 .
Qi C S , Yu Z H , Hou Z , et al . Study on quality prediction of the complex production based on CART algorithm [J]. Modular Machine Tool & Automatic Manufacturing Technique , 2010 ( 3 ): 94 - 97, 102 . (in Chinese)
张远舸 . PBX造型粉的致密化行为与密度演化规律研究 [D]. 绵阳 : 中国工程物理研究院 , 2018 .
Zhang Y G . Study on the densification behavior and density evolution of PBX modeling powders [D]. Mianyang : China Academy of Engineering Physics , 2018 . (in Chinese)
陶磊 , 刘检华 , 夏焕雄 , 等 . 基于B 样条神经网络的熔铸装药温度场预测 [J]. 兵工学报 , 2023 , 44 ( 5 ): 1339 - 1349 .
Tao L , Liu J H , Xia H X , et al . Temperature field prediction of melt-cast explosives based on a B-spline neural network [J]. Acta Armamentarii , 2023 , 44 ( 5 ): 1339 - 1349 . (in Chinese)
王伟 , 马乾伦 , 白振华 , 等 . 基于梯度提升决策树的冷轧高强钢卷力学性能预测 [J]. 中国机械工程 , 2023 , 34 ( 18 ): 2222 - 2229 .
Wang W , Ma Q L , Bai Z H , et al . Mechanics property prediction of cold rolled high strength steel coils based on GBDT [J]. China Mechanical Engineering , 2023 , 34 ( 18 ): 2222 - 2229 . (in Chinese)
康俊涛 , 韦朝校 . 多层感知机和支持向量机回归算法预测T型接头固有应变对比研究 [J]. 机械工程学报 , 2025 , 61 ( 18 ): 181 - 189 .
Kang J T , Wei C X . Comparative study of multilayer perceptron and support vector machine regression algorithms for predicting inherent strain in T-joints [J]. Journal of Mechanical Engineering , 2025 , 61 ( 18 ): 181 - 189 . (in Chinese)
陈忠杭 , 王舟挺 , 沈加明 , 等 . 基于WOA-Stacking集成学习的注塑产品尺寸预测 [J]. 工程塑料应用 , 2024 , 52 ( 6 ): 135 - 141, 163 .
Chen Z H , Wang Z T , Shen J M , et al . Injection molding product size prediction based on WOA-Stacking ensemble learning method [J]. Engineering Plastics Application , 2024 , 52 ( 6 ): 135 - 141, 163 . (in Chinese)
虞任豪 , 项薇 , 林文文 , 等 . 基于集成学习与改进灰狼算法的铝合金压铸件仿真与优化 [J]. 特种铸造及有色合金 , 2025 , 45 ( 10 ): 1536 - 1542 .
Yu R H , Xiang W , Lin W W , et at . Simulation and optimisation of aluminium alloy die castings based on integrated learning and improved grey wolf algorithm [J]. Special casting & Nonferrous Alloys , 2025 , 45 ( 10 ): 1536 - 1542 . (in Chinese)
Dong Z R , Wang H W , Zhang R T , et al . Ensemble learning-enabled early prediction of dimensional accuracy for complex products during investment casting [J]. Journal of Manufacturing Processes , 2024 , 113 : 291 - 306 .
Zhang H P , Deng Y , Chen F H , et al . Fatigue life prediction for orthotropic steel bridge decks welds using a Gaussian variational Bayes network and small sample experimental data [J]. Reliability Engineering and System Safety , 2025 , 264 (Part B): 111406.
李悦 , 谢恒 , 周公博 , 等 . 基于半监督贝叶斯Transformer的刀具磨损软测量及不确定性分析方法 [J]. 中国机械工程 , 2024 , 35 ( 11 ): 2015 - 2025 .
Li Y , Xie H , Zhou G B , et al . Soft sensor modeling and uncertainty analysis approach of tool wear based on semi-supervised Bayesian Transformer [J]. China Mechanical Engineering , 2024 , 35 ( 11 ): 2015 - 2025 . (in Chinese)
Xia H J , Ding Y Y , Chen W H , et al . Measurement uncertainty evaluation with small samples: a review and prospect [J]. Measurement , 2026 , 258 ( Part A ): 119031 .
曾升 , 游国强 , 温彤 , 等 . 高聚物粘结炸药压制成型工艺实验研究综述 [J]. 兵工自动化 , 2024 , 43 ( 8 ): 86 - 92 .
Zeng S , You G Q , Wen T , et al . Review of experimental research on PBX pressing process [J]. Ordnance industry automation , 2024 , 43 ( 8 ): 86 - 92 . (in Chinese)
Yasen A , Zhou Y X , Gao W H , et al . A MA-SSA-optimized XGBoost-MLP framework using LIBS for rapid classification and quantitative analysis of heavy metals in traditional chinese medicines [J]. Journal of Pharmaceutical and Biomedical Analysis , 2026 , 270 : 117296 .
Lv B , Gong H X , Dong B , et al . An explainable XGBoost model for international roughness index prediction and key factor identification [J]. Applied Sciences , 2025 , 15 ( 4 ): 1893 .
Chen M Y , Xin J Z , Tang Q Z , et al . Explainable machine learning model for load-deformation correlation in long-span suspension bridges using XGBoost-SHAP [J]. Developments in the Built Environment , 2024 , 20 : 100569 .
0
浏览量
73
下载量
0
CNKI被引量
关联资源
相关文章
相关作者
相关机构
京公网安备11010802024360号