基于XGBoost的PC板爆炸损伤评估模型

赵春风; 吴艺秀; 向思麒; 李晓杰

doi:10.11883/bzycj-2025-0250

基于XGBoost的PC板爆炸损伤评估模型

doi: 10.11883/bzycj-2025-0250

赵春风^{1, 2, 3,},
吴艺秀²,
向思麒⁴,
李晓杰³

1.
合肥工业大学安徽省土木工程结构与材料重点实验室，安徽合肥 230009
2.
合肥工业大学土木与水利工程学院，安徽合肥 230009
3.
大连理工大学工业装备结构分析优化与CAE软件全国重点实验室，辽宁大连 116024
4.
新疆农业大学水利与土木工程学院，新疆乌鲁木齐 830052

基金项目: 新疆维吾尔自治区自然科学重点基金（2022D01D33）；工业装备结构分析优化与CAE软件全国重点实验室开放基金项目（GZ24120）

详细信息

作者简介:
赵春风（1983－　），男，博士，教授，Zhaowindy@hfut.edu.cn

中图分类号: T
计量
- 文章访问数: 192
- HTML全文浏览量: 34
- PDF下载量: 59
- 被引次数: 0
出版历程
- 收稿日期: 2025-08-10
- 修回日期: 2025-09-25
- 网络出版日期: 2025-09-30

Blast damage assessment model of PC slabs based on XGBoost

ZHAO Chunfeng^{1, 2, 3
,},
WU Yixiu²,
XIANG Siqi⁴,
LI Xiaojie³

1.
Anhui Provincial Key Laboratory of Civil Engineering Structures and Materials, Hefei University of Technology, Hefei 230009, Anhui, China
2.
College of Civil Engineering, Hefei University of Technology, Hefei 230009, China
3.
State Key Laboratory of Structural Analysis, Optimization and CAE Software for Industrial Equipment, Dalian University of Technology, Dalian 116024, Liaoning, China
4.
College of Water Resources and Civil Engineering, Xinjiang Agricultural University, Urumqi 830052, Xinjiang, China

摘要

摘要: 装配式建筑结构因其节能环保、质量可控及施工高效快捷等优点，在土木工程中得到了广泛应用。作为装配式建筑结构的核心受力构件，预制钢筋混凝土（precast reinforced concrete，PC）板易受燃气爆炸、工业爆炸和恐怖袭击等威胁。为了准确评估PC板在爆炸作用下的损伤状态，提升结构抗爆能力，并降低人员伤亡风险，通过构建PC板爆炸响应数据集，选取6项几何结构参数和2项爆炸荷载参数作为输入特征，采用3种不同的机器学习算法（GPR、RF和XGBoost）对PC板的最大位移进行预测，采用均方根误差、决定系数、平均绝对误差、散射系数及综合性能目标函数值5项回归评价指标，对3种模型的预测精度进行对比分析；提出了基于支座转角损伤准则的损伤分类评估模型，利用混淆矩阵和5项分类指标（准确率、精确率、召回率、F₁分数和Kappa系数）分析3种准则下模型的性能差异，并与简化后的模型及经验预测方法进行对比。结果表明：在最大位移预测方面，3种机器学习模型中表现最佳的为XGBoost模型，其拟合性优于GPR和RF模型且综合性能最优；在损伤分类预测方面，基于准则Ⅱ的XGBoost损伤分类模型性能最优，损伤识别准确率达92.5%，显示出其高效的损伤类型识别能力。基于XGBoost算法的爆炸作用下PC板损伤分类评估模型具有强大的性能，对结构抗爆和爆后快速损伤评定具有参考价值。
- 预制钢筋混凝土板 /
- 机器学习 /
- 极限梯度增强 /
- 损伤评估
Abstract: Prefabricated building structures have been widely applied in civil engineering due to their advantages of energy conservation, environmental protection, controllable quality, and efficient construction. As the core load-bearing components of prefabricated building structures, precast reinforced concrete (PC) slabs are vulnerable to threats from gas explosions, industrial explosions, and terrorist attacks. To accurately assess the damage state of PC slabs under explosion, enhance structural blast resistance, and reduce casualties, an explosion response dataset of PC slabs was constructed. Six geometric parameters (slab thickness/length/width, steel reinforcement ratio, compressive strength of concrete, etc.) and two explosion load parameters (explosive weight and explosive distance) were selected as input features. Three machine learning algorithms (GPR, RF, and XGBoost) were used to predict the maximum displacement of PC slabs, and their prediction accuracies are compared by root mean square error, coefficient of determination, mean absolute error, scattering index, and comprehensive performance objective function. Furthermore, a damage classification evaluation model based on the support rotation angle damage criterion is proposed. The performance differences of the model under three criteria are analyzed by confusion matrix and five classification indices (accuracy, precision, recall, F₁-score, and Kappa coefficient), and compared with simplified models and empirical prediction methods. The research results indicate that in terms of maximum displacement prediction for PC slabs under explosion loads, the XGBoost model demonstrates the best performance among the three machine learning models (GPR、RF and XGBoost). Specifically, the fitting degree of XGBoost is superior to those of GPR and RF models. Meanwhile, and the XGBoost shows the most outstanding comprehensive performance, with a damage recognition accuracy of 92.5%, which demonstrates its high-efficiency in identifying different damage types. The XGBoost-based damage classification evaluation model for PC slabs under explosion loads exhibits powerful performance, providing important references for structural blast resistance design and rapid post-blast damage assessment.
- precast reinforced concrete slab /
- machine learning /
- extreme gradient boosting /
- damage assessment

HTML全文

图 1 基于XGBoost的PC板爆炸作用下损伤评估模型流程图

Figure 1. Flow chart of damage assessment model under explosion on PC board based on XGBoost

下载: 全尺寸图片幻灯片

图 2 不同算法示意图

Figure 2. Schematic diagrams of different algorithm

下载: 全尺寸图片幻灯片

图 3 3种模型实际值与预测值之间比较

Figure 3. Comparison of actual and predicted values among three models

下载: 全尺寸图片幻灯片

图 4 3种模型的预测性能与残差对比

Figure 4. Comparison of predictive performance and residuals among three models

下载: 全尺寸图片幻灯片

图 5 基于XGBoost模型的SHAP均值特征重要性分析

Figure 5. Analyses of SHAP mean value and feature importance based on XGBoost model

下载: 全尺寸图片幻灯片

图 6 PC板爆炸作用下的支座转角示意图

Figure 6. Schematic diagram of support rotation in PC slab under blast loading

下载: 全尺寸图片幻灯片

图 7 混淆矩阵示意图

Figure 7. Confusion matrix diagram

下载: 全尺寸图片幻灯片

图 8 3种损伤评估准则的模型预测结果

Figure 8. Model prediction results of three damage criteria

下载: 全尺寸图片幻灯片

图 9 基于3种损伤评估准则的混淆矩阵

Figure 9. Confusion matrix based on three damage criteria

下载: 全尺寸图片幻灯片

图 10 SMOTE和粒子群优化算法示意图

Figure 10. Schematic diagrams of SMOTE and PSO algorithm

下载: 全尺寸图片幻灯片

图 11 损伤评估简化模型流程图

Figure 11. Flow chart of damage assessment simplified model

下载: 全尺寸图片幻灯片

图 12 简化模型损伤评估数据集测试集混淆矩阵

Figure 12. Test set confusion matrix for damage assessment database of simplified model

下载: 全尺寸图片幻灯片

图 13 简化模型混合数据集测试集混淆矩阵

Figure 13. Test set confusion matrix for hybrid database of simplified model

下载: 全尺寸图片幻灯片

图 14 3种模型测试集混淆矩阵

Figure 14. Test set confusion matrix of three models

下载: 全尺寸图片幻灯片

图 15 基于经验预测方法的结果

Figure 15. Results based on empirical prediction methods

下载: 全尺寸图片幻灯片

表 1 本研究数值来源及类型

Table 1. Data source and type of this study

作者	数据类型	数量	增扩数据
Zhao^[13]	模拟	14	137
杜永峰^[14]	模拟	6	46
周兆鹏^[15]	试验	3	36
Tian^[16]	试验	3	69
董刚^[17]	模拟	6	46
李建武^[18]	模拟	14	125
Choi^[19]	试验+模拟	1+10	80
Wang^[20]	试验+模拟	5	78

下载: 导出CSV

表 2 位移预测数据集参数范围

Table 2. Parameter range of displacement prediction database

数据	l/m	b/m	D/m	f_c/MPa	f_v/MPa	ρ/%	R/m	W/kg	d/mm
平均值	2.320	1.577	0.158	40.025	417.179	0.474	1.019	3.064	21.126
范围	1.3～4	1～3.1	0.07～0.3	20～80	235～600	0.15～1.77	0.01～2.1	0.05～10	0.44～137

下载: 导出CSV

表 3 损伤程度预测数据集参数范围

Table 3. Parameter range of damage assessment database

数据	l/m	b/m	D/m	f_c/MPa	f_v/MPa	ρ/%	R/m	W/kg
平均值	2.63	1.86	0.17	46.5	458.33	0.42	1.38	4.55
范围	1.4～4	1～3.1	0.08～0.3	30～70	300～600	0.17～1.13	0.5～2.1	0.5～12.5

下载: 导出CSV

表 4 3种模型性能的回归评价指标

Table 4. Regression evaluation indicators for the performance of three models

模型	数据集形式	R²	σ_RMSE/mm	σ_MAE/mm	ξ	f_OBJ
GPR	训练集	0.994	1.357	0.583	0.064	1.325
GPR	测试集	0.965	3.025	2.350	0.143	1.325
RF	训练集	0.972	2.570	0.703	0.121	2.330
RF	测试集	0.891	5.659	3.837	0.268	2.330
XGBoost	训练集	0.998	0.863	0.205	0.041	0.939
XGBoost	测试集	0.975	2.874	2.185	0.136	0.939

下载: 导出CSV

表 5 基于支座转角的损伤评估划分

Table 5. Damage assessment division based on support rotation

损伤程度	准则Ⅰ^[28]	准则Ⅱ^[29]	准则Ⅲ^[30]
轻度损伤	θ≤1.7°	θ≤2°	θ≤2°
中度损伤	1.7°＜θ≤4.6°	2°＜θ≤5°	2°＜θ≤6°
严重损伤	4.6°＜θ≤6.8°	5°＜θ≤10°	6°＜θ≤12°
倒塌破坏	θ＞6.8°	θ＞10°	θ＞12°

下载: 导出CSV

表 6 Kappa系数值的含义

Table 6. Meaning of Kappa coefficient values

Kappa系数	一致性强度
＜0.20	较差
0.21-0.40	一般
0.41-0.60	中等
0.61-0.80	较强
0.81-1.0	强

下载: 导出CSV

表 7 基于3种损伤准则模型的分类指标

Table 7. Classification indicators based on three damage models

准则	A_cc/%	P/%	R_c/%	F₁	Ka
Ⅰ	59.17	60.70	48.20	0.54	0.331
Ⅱ	95.00	92.92	90.10	0.92	0.925
Ⅲ	74.17	76.51	71.65	0.74	0.625

下载: 导出CSV

表 8 基于损伤评估数据集的简化模型分类指标

Table 8. Classification index of simplified model based on damage assessment database

A_cc/%	P/%	R_c/%	F₁	Ka
90.91	91.23	90.84	0.91	0.878

下载: 导出CSV

表 9 基于混合数据集的简化模型分类指标

Table 9. Classification index for simplified models based on hybrid databases

A_cc/%	P/%	R_c/%	F₁	Ka
91.48	91.05	92.39	0.917	0.889

下载: 导出CSV

表 10 基于XGBoost的损伤评估模型和简化模型综合对比

Table 10. Comprehensive comparison between XGBoost model and simplified model

模型	准确率/%	平均计算耗时/min	物理逻辑	可解释性
基于XGBoost的损伤评估模型	95.0	2.1	位移预测+损伤评估	强（分步解释响应与损伤）
简化模型	91.5	1.3	直接损伤评估分类	较弱（黑箱特征明显）

下载: 导出CSV

表 11 对比经验预测方法的样本数据点

Table 11. Sample data points for comparing empirical prediction methods

序号	板厚度/ m	爆炸距离/ m	TNT当量/ kg	比例距离/ (m·kg^−1/3)	比例厚度/ (m·kg^−1/3)	实际损伤	分类预测
1	0.14	0.5	1.8	0.41	0.96	倒塌	倒塌
2	0.14	0.5	1.4	0.45	0.96	倒塌	倒塌
3	0.14	0.5	0.9	0.52	0.96	严重	严重
4	0.14	0.5	0.6	0.59	0.96	严重	严重
5	0.22	0.6	1.8	0.49	0.99	倒塌	倒塌
6	0.22	0.6	1.3	0.55	0.99	中度	中度
7	0.22	0.6	0.9	0.62	0.99	轻度	中度

下载: 导出CSV

参考文献(34)

[1]	YE X, ZHAO C F, HE K C, et al. Blast behaviors of precast concrete sandwich EPS panels: FEM and theoretical analysis [J]. Engineering Structures, 2021, 226: 111345. DOI: 10.1016/j.engstruct.2020.111345.
[2]	ZHAO C F, YE X, HE K C, et al. Numerical study and theoretical analysis on blast resistance of fabricated concrete slab [J]. Journal of Building Engineering, 2020, 32: 101760. DOI: 10.1016/j.jobe.2020.101760.
[3]	李圣童, 汪维, 梁仕发, 等. 长持时爆炸冲击波荷载作用下梁板组合结构的动力响应 [J]. 爆炸与冲击, 2022, 42(7): 075103. DOI: 10.11883/bzycj-2021-0495. LI S T, WANG W, LIANG S F, et al. Dynamic response of beam-slab composite structures under long-lasting explosion shock wave load [J]. Explosion and Shock Waves, 2022, 42(7): 075103. DOI: 10.11883/bzycj-2021-0495.
[4]	NORRIS C H. Structural design for dynamic loads [M]. New York: McGraw-Hill, 1959.
[5]	苏琼, 程月华, 吴昊. 爆炸荷载作用下UHPC板的弯曲损伤评估 [J]. 爆炸与冲击, 2023, 43(12): 125103. DOI: 10.11883/bzycj-2023-0160. SU Q, CHENG Y H, WU H. Flexural damage assessment for UHPC panels under blast loadings [J]. Explosion and Shock Waves, 2023, 43(12): 125103. DOI: 10.11883/bzycj-2023-0160.
[6]	LI Q M, MENG H. Pressure-impulse diagram for blast loads based on dimensional analysis and single-degree-of-freedom model [J]. Journal of Engineering Mechanics, 2002, 128(1): 87–92. DOI: 10.1061/(ASCE)0733-9399(2002)128:1(87).
[7]	WANG W, ZHANG D, LU F Y, et al. Pressure-impulse diagram with multiple failure modes of one-way reinforced concrete slab under blast loading using SDOF method [J]. Journal of Central South University, 2013, 20(2): 510–519. DOI: 10.1007/s11771-013-1513-z.
[8]	PARK G K, KWAK H G, FILIPPOU F C. Evaluation of nonlinear behavior and resisting capacity of reinforced concrete columns subjected to blast loads [J]. Engineering Failure Analysis, 2018, 93: 268–288. DOI: 10.1016/j.engfailanal.2018.07.024.
[9]	NADERPOUR H, MIRRASHID M, PARSA P. Failure mode prediction of reinforced concrete columns using machine learning methods [J]. Engineering Structures, 2021, 248: 113263. DOI: 10.1016/j.engstruct.2021.113263.
[10]	SANDEEP M S, TIPRAK K, KAEWUNRUEN S, et al. Shear strength prediction of reinforced concrete beams using machine learning [J]. Structures, 2023, 47: 1196–1211. DOI: 10.1016/j.istruc.2022.11.140.
[11]	李般若, 霍璞, 喻君. 基于GNN的爆炸压力时空分布预测模型 [J]. 爆炸与冲击, 2025: 1–19. https://www.bzycj.cn/article/doi/ 10.11883/bzycj-2024-0503. DOI: 10.11883/bzycj-2024-0503. LI B R, HUO P, YU J. GNN-based predictive model for spatial and temporal distribution of blast overpressure [J]. Explosion and Shock Waves, 2025: 1–19. DOI: 10.11883/bzycj-2024-0503.
[12]	ZHAO C F, ZHU Y F, ZHOU Z H. Machine learning-based approaches for predicting the dynamic response of RC slabs under blast loads [J]. Engineering Structures, 2022, 273: 115104. DOI: 10.1016/j.engstruct.2022.115104.
[13]	ZHAO C, YE X, HE K, et al. Numerical study and theoretical analysis on blast resistance of fabricated concrete slab [J]. Journal of Building Engineering, 2020, 32(8): 101760. DOI: 10.1016/j.jobe.2020.101760.
[14]	杜永峰, 靳振飞. 灌浆套筒连接装配式剪力墙爆炸响应及参数分析 [J]. 世界地震工程, 2020, 36(4): 62–70. DOI: 10.3969/j.issn.1007-6069.2020.04.008. DU Y F, JIN Z F. Explosive response and parameters analysis of prefabricated shear wall connected by grouting sleeve [J]. World Earthquake Engineering, 2020, 36(4): 62–70. DOI: 10.3969/j.issn.1007-6069.2020.04.008.
[15]	周兆鹏, 闫秋实, 田栓柱, 等. 爆炸荷载作用下装配式钢筋混凝土板抗爆性能试验研究 [J]. 天津大学学报(自然科学与工程技术版), 2022, 55(6): 611–620. DOI: 10.11784/tdxbz202105056. ZHOU Z P, YAN Q S, TIAN S Z, et al. Experimental study on antiexplosion performance of precast concrete slabs under explosion load [J]. Journal of Tianjin University (Science and Technology), 2022, 55(6): 611–620. DOI: 10.11784/tdxbz202105056.
[16]	TIAN S Z, YAN Q S, DU X L, et al. Experimental and numerical studies on the dynamic response of precast concrete slabs under blast load [J]. Journal of Building Engineering, 2023, 70: 106425. DOI: 10.1016/j.jobe.2023.106425.
[17]	董刚. 爆炸作用下装配式钢筋混凝土叠合板动态响应研究 [D]. 长春: 吉林建筑大学, 2023. DOI: 10.27714/d.cnki.gjljs.2023.000411. DONG G. Study on dynamic response of prefabricated reinforced concrete laminate slab under blast loading [D]. Changchun: Jilin Jianzhu University, 2023. DOI: 10.27714/d.cnki.gjljs.2023.000411.
[18]	李建武, 闫秋实, 李述涛, 等. 装配式钢筋混凝土叠合板抗爆性能影响参数分析 [J]. 防护工程, 2022, 44(2): 17–21. DOI: 10.3969/j.issn.1674-1854.2022.02.003. LI J W, YAN Q S, LI S T, et al. Analysis of parameters affecting anti-blast performance of prefabricated reinforced concrete laminated slabs [J]. Protective Engineering, 2022, 44(2): 17–21. DOI: 10.3969/j.issn.1674-1854.2022.02.003.
[19]	CHOI J H, CHOI S J, KIM J H J, et al. Evaluation of blast resistance and failure behavior of prestressed concrete under blast loading [J]. Construction and Building Materials, 2018, 173: 550–572. DOI: 10.1016/j.conbuildmat.2018.04.047.
[20]	WANG W, HUO Q, YANG J C, et al. Damage analysis of POZD coated square reinforced concrete slab under contact blast [J]. Defence Technology, 2022, 18(9): 1715–1726. DOI: 10.1016/j.dt.2021.07.005.
[21]	赵春风, 向思麒, 朱玉富. 基于机器学习的RC板爆炸两阶段损伤评估模型 [J]. 建筑结构学报, 2025, 46(1): 212–222. DOI: 10.14006/j.jzjgxb.2023.0727. ZHAO C F, XIANG S Q, ZHU Y F. A two-stage damage assessment model for RC slab under explosion based on machine learning [J]. Journal of Building Structures, 2025, 46(1): 212–222. DOI: 10.14006/j.jzjgxb.2023.0727.
[22]	李忠献, 师燕超. 建筑结构抗爆分析理论 [M]. 北京: 科学出版社, 2015. LI Z X, SHI Y C. Blast analysis of building structures [M]. Beijing: Science Press, 2015.
[23]	LI K, LONG Y P, WANG H, et al. Modeling and sensitivity analysis of concrete creep with machine learning methods [J]. Journal of Materials in Civil Engineering, 2021, 33(8): 04021206. DOI: 10.1061/(ASCE)MT.1943-5533.0003843.
[24]	YU J B. State of health prediction of lithium-ion batteries: multiscale logic regression and Gaussian process regression ensemble [J]. Reliability Engineering & System Safety, 2018, 174: 82–95. DOI: 10.1016/j.ress.2018.02.022.
[25]	BREIMAN L. Random forests [J]. Machine Learning, 2001, 45(1): 5–32. DOI: 10.1023/A:1010933404324.
[26]	张成玺. 爆炸载荷下钢筋混凝土构件的动态响应及损伤评估 [D]. 济南: 山东建筑大学, 2024. DOI: 10.27273/d.cnki.gsajc.2024.000353. ZHANG C X. Dynamic responseand damage assessment of reinforced concrete members under explosive loading [D]. Ji’an: Shandong Jianzhu University, 2024. DOI: 10.27273/d.cnki.gsajc.2024.000353.
[27]	LIAO Y, SHI S Q, CHEN S, et al. Numerical evaluation of the retrofit effectiveness for polyurea-woven glass fiber mesh composite retrofitted RC slab subjected to blast loading [J]. Structures, 2022, 36: 215–232. DOI: 10.1016/j.istruc.2021.12.012.
[28]	汪维. 钢筋混凝土构件在爆炸载荷作用下的毁伤效应及评估方法研究 [D]. 长沙: 国防科学技术大学, 2012. WANG W. Study on damage effects and assessments method of reinforced concrete structural members under blast loading [D]. Changsha: National University of Defense Technology, 2012.
[29]	李天华, 赵均海, 魏雪英, 等. 爆炸荷载下钢筋混凝土板的动力响应及参数分析 [J]. 建筑结构, 2012, 42(S1): 786–790. DOI: 10.19701/j.jzjg.2012.s1.194. LI T H, ZHAO J H, WEI X Y, et al. Dynamic response and parametric analysis on reinforced concrete slabs under blast loadings [J]. Building Structure, 2012, 42(S1): 786–790. DOI: 10.19701/j.jzjg.2012.s1.194.
[30]	TMS-1300. Structures to Resist the Effects of Accidental Explosions[S]. Washington, D. C: Department of the Army, the Navy, and the Air Force, 1990.
[31]	于晓辉, 王猛, 宁超列. 基于机器学习的钢筋混凝土柱失效模式两阶段判别方法 [J]. 建筑结构学报, 2022, 43(8): 220–231. DOI: 10.14006/j.jzjgxb.2020.0392. YU X H, WANG M, NING C L. A machine-learning-based two-step method for failure mode classification of reinforced concrete columns [J]. Journal of Building Structures, 2022, 43(8): 220–231. DOI: 10.14006/j.jzjgxb.2020.0392.
[32]	COHEN J. A coefficient of agreement for nominal scales [J]. Educational and Psychological Measurement, 1960, 20: 37–46. DOI: 10.1177/001316446002000104.
[33]	MCVAY M K, STATION U S A E W E, ENGINEERS U S A C O. Spall Damage of Concrete Structures [M]. U. S. Army Engineer Waterways Experiment Station, 1988.
[34]	LI Y, CHEN Z Y, REN X B, et al. Experimental and numerical study on damage mode of RC slabs under combined blast and fragment loading [J]. International Journal of Impact Engineering, 2020, 142: 103579. DOI: 10.1016/j.ijimpeng.2020.103579.