Shield attitude prediction and optimization based on CatBoost-NSGA-III algorithm

doi:10.16265/j.cnki.issn1003-3033.2024.08.1360

Abstract

Abstract:

To solve the problems such as forward tilt deformation, serpentine shape, axis deviation and correction during shield tunneling, which affected the safety and efficiency of shield construction, a multi-objective optimization method of shield attitude combining CatBoost and NSGA-Ⅲ was proposed. Taking Guiyang Metro as the background, 22 influencing factors were selected as input parameters, and the nonlinear mapping function relationship between input parameters and shield attitude was established by using CatBoost algorithm. The importance of input parameters was evaluated by random forest (RF) algorithm. A CatBoost-NSGA-Ⅲ multi-objective optimization model was established to minimize the absolute value of the shield attitude, and the applicability and effectiveness of the proposed method were verified by a case study. The results show that the prediction model obtained by using CatBoost algorithm to train engineering measured data has high accuracy, and the R² range of 5 shield attitude targets is 0.916-0.943. By using the CatBoost-NSGA-Ⅲ multi-objective optimization method, the attitude of the shield can be optimized significantly, and the average value of the overall improvement is 53.34%.

Key words: categorical boosting (CatBoost), non-dominated sorting genetic algorithm-Ⅲ (NSGA-Ⅲ), shield attitude, multi-objective optimization, importance ranking

CLC Number:

X948
U455

WU Xianguo, LIU Jun, CAO Yuan, LEI Yu, LI Shifan, QIN Yawei. Shield attitude prediction and optimization based on CatBoost-NSGA-III algorithm[J]. China Safety Science Journal, 2024, 34(8): 69-77.

Figures/Tables 9

Table 1

Table 2

Table 3

Table 4

Fig.1

Table 5

Table 6

Fig.2

Table 7

References 13

[1]	沈翔, 袁大军. 盾构水平偏角变化对盾构-土相互作用影响[J]. 中国公路学报, 2020, 33(3): 132-143. doi: 10.19721/j.cnki.1001-7372.2020.03.011
	SHEN Xiang, YUAN Dajun. Influence of shield yawing angle variation on shield-soil interaction[J]. China Journal of Highway and Transport, 2020, 33(3): 132-143. doi: 10.19721/j.cnki.1001-7372.2020.03.011
[2]	黄威, 任梦, 陈培帅, 等. 盾构水平姿态的理论分析模型[J]. 隧道建设:中英文, 2022, 42(1): 83-89.
	HUANG Wei, REN Meng, CHEN Peishuai, et al. Theoretical analysis model of shield horizontal attitude[J]. Tunnel Construction, 2022, 42(1): 83-89.
[3]	苏栋, 谭毅俊, 沈翔, 等. 软土地层加固对盾构姿态调控及地层变形的影响研究[J]. 现代隧道技术, 2023, 60(2): 138-148,167.
	SU Dong, TAN Yijun, SHEN Xiang, et al. A Study on impact of soft soil stratum reinforcement on the attitude regulation of shield machine and stratum deformation[J]. Modern Tunnelling Technology, 2023, 60(2): 138-148, 167.
[4]	夏汉庸, 尹和军, 徐教煌, 等. 基于机器学习的多施工参数盾构施工姿态预测[J]. 测绘通报, 2021(1): 157-160,164. doi: 10.13474/j.cnki.11-2246.2021.0030
	XIA Hanyong, YIN Hejun, XU Jiaohuang, et al. Multi-construction parameter shield construction attitude prediction based on machine learning[J]. Bulletin of Surveying and Mapping, 2021(1): 157-160, 164. doi: 10.13474/j.cnki.11-2246.2021.0030
[5]	曾铁梅, 李昕懿, 冯宗宝, 等. 基于LightGBM的盾构机姿态预测与控制研究[J/OL]. 铁道标准设计: 1-11. [2024-03-14].https://doi.org/10.13238/j.issn.1004-2954.202212140005.
	ZENG Tiemei, LI Xinyi, FENG Zongbao, et al. Research on attitude prediction and control of shield machine based on LightGBM[J/OL]. Railway Standard Design:1-11. [2024-03-14] https://doi.org/10.13238/j.issn.1004-2954.202212140005.
[6]	吴坚, 曾志全, 张亚鹏, 等. 基于循环神经网络的盾构姿态及掘进参数预测[J]. 浙江工业大学学报, 2023, 51(6): 663-670.
	WU Jian, ZENG Zhiquan, ZHANG Yapeng, et al. Prediction of shield posture and tunneling parameters based on recurrent neural network[J]. Journal of Zhejiang University of Technology, 2023, 51(6): 663-670.
[7]	PROKHORENKOVA L, GUSEV G, VOROBEV A, et al. CatBoost: unbiased boosting with categorical features[C]. Proceedings of the 32^nd International Conference on Neural Information Processing Systems, 2018: 6 639-6 649.
[8]	DEB K, JAIN H. An evolutionary many-objective optimization algorithm using reference-point-based nondominated sorting approach, part I: solving problems with box constraints[J]. IEEE Transactions on Evolutionary Computation, 2014, 18(4): 577-601.
[9]	JUNG J, CHUNG H, KWON Y, et al. An ANN to predict ground condition ahead of tunnel face using TBM operational data[J]. KSCE Journal of Civil Engineering, 2019, 23(7): 3 200-3 206.
[10]	刘茜. 基于智能方法的盾构施工参数预测和优化控制[D]. 武汉: 华中科技大学, 2022.
	LIU Xi. Prediction and optimal control of shield construction parameters based on intelligent methods[D]. Wuhan: Huazhong University of Science and Technology, 2022.
[11]	张利冬, 宋泽阳, 罗振敏, 等. 基于机器学习的煤自然发火期预测[J]. 中国安全科学学报, 2022, 32(12): 118-124. doi: 10.16265/j.cnki.issn1003-3033.2022.12.0134
	ZHANG Lidong, SONG Zeyang, LUO Zhenmin, et al. Prediction of coal spontaneous combustion period based on machine learning[J]. China Safety Science Journal, 2022, 32(12): 118-124. doi: 10.16265/j.cnki.issn1003-3033.2022.12.0134
[12]	吴忠坦, 吴贤国, 刘俊, 等. 基于随机森林-NSGA-Ⅲ的盾构姿态优化控制[J]. 现代隧道技术, 2023, 60(5): 48-57.
	WU Zhongtan, WU Xianguo, LIU Jun, et al. Shield attitude optimization and control based on random forest-NSGA-Ⅲ[J]. Modern Tunnelling Technology, 2023, 60(5): 48-57.
[13]	吴贤国, 冯宗宝, 刘俊, 等. 基于RF-NSGA-Ⅱ的盾构施工地表沉降安全控制多目标优化[J]. 中国安全科学学报, 2022, 32(8): 45-51. doi: 10.16265/j.cnki.issn1003-3033.2022.08.2702
	WU Xianguo, FENG Zongbao, LIU Jun, et al. Multi-objective optimization of surface settlement safety control during shield construction based on RF-NSGA-Ⅱ[J]. China Safety Science Journal, 2022, 32(8): 45-51. doi: 10.16265/j.cnki.issn1003-3033.2022.08.2702

参数类型	变量	数据
参数类型	变量	最小值	最大值	平均值
输入	土仓压力-上 x₁/MPa	0.07	0.21	0.14
	土仓压力-下x₂/MPa	0.04	0.53	0.22
	土仓压力-左 x₃/MPa	0.05	0.29	0.15
	土仓压力-右 x₄/MPa	0.03	0.28	0.15
	螺旋机转速 x₅/ (r·min^-1)	0.91	22.9	12.02
	千斤顶推力-上 x₆/kN	3.69	209.02	77.77
	千斤顶推力-下 x₇/kN	60.19	212.56	123.24
	千斤顶推力-左 x₈/kN	28.84	269.35	132.91
	千斤顶推力-右 x₉/kN	42.20	235.88	109.87
	刀盘转速 x₁₀/(r·min^-1)	1.21	2.04	1.66
	刀盘扭矩 x₁₁/(kN·m)	719.91	2 998.29	2 058.51
	掘进速度 x₁₂/ (mm·min^-1)	7.53	70.29	40.09
	注浆压力 x₁₃/MPa	0.12	0.31	0.25
	总推力 x₁₄/kN	5 441.28	19 086.34	12 249.35
	铰接行程 x₁₅/m	0.87	38.40	13.71
	铰接偏差-水平 x₁₆/mm	1.34	86.68	24.85
	铰接偏差-垂直 x₁₇/mm	0	59.85	21.15
	盾尾间隙-上 x₁₈/mm	29.03	79.03	63.81
	盾尾间隙-下 x₁₉/mm	33.81	83.7	67.81
	盾尾间隙-左 x₂₀/mm	27.03	83.19	67.37
	盾尾间隙-右 x₂₁/mm	11.49	189.87	63.33
	切口行程 x₂₂/m	1.4	1.54	1.5
	泡沫剂 x₂₃/L	25.19	157.79	47.96
输出	切口水平方向位移 f₁/mm	-39.53	32.03	-0.58
	切口垂直方向位移 f₂/mm	-49.47	48.61	-0.32
	尾部水平方向位移 f₃/mm	-52.86	42.36	-25.78
	尾部垂直方向位移 f₄/mm	-46.66	51.94	16.15
	俯仰角 f₅/(°)	-27.11	26.01	-1.45

预测目标	l2_leaf_reg	learning_rate	max_depth
f₁/mm	9	0.026 57	8
f₂/mm	7	0.399 37	8
f₃/mm	4	0.167 00	6
f₄/mm	6	0.034 56	8
f₅/(°)	3	0.065 53	8

折叠序号	f₁/mm			f₂/mm			f₃/mm			f₄/mm			f₅/(°)
折叠序号	R²	RMSE	MAE	R²	RMSE	MAE	R²	RMSE	MAE	R²	RMSE	MAE	R²	RMSE	MAE
1	0.934	3.741	2.922	0.943	3.418	2.705	0.917	3.594	2.827	0.915	4.372	3.429	0.909	0.986	0.833
2	0.943	3.456	2.717	0.949	3.247	2.575	0.923	3.466	2.739	0.920	4.217	3.291	0.920	0.922	0.756
3	0.949	3.292	2.611	0.950	3.216	2.555	0.925	3.431	2.721	0.927	4.037	3.145	0.915	0.953	0.781
4	0.925	3.976	3.198	0.939	3.546	2.851	0.914	3.674	2.876	0.911	4.473	3.555	0.919	0.930	0.766
5	0.922	4.069	3.297	0.933	3.715	3.051	0.912	3.576	2.906	0.912	4.442	3.521	0.920	0.926	0.760
平均值	0.934	3.707	2.950	0.943	3.428	2.748	0.918	3.576	2.814	0.917	4.308	3.388	0.916	0.943	0.779

算法	f₁/mm			f₂/mm			f₃/mm			f₄/mm			f₅/(°)
算法	R²	RMSE	MAE	R²	RMSE	MAE	R²	RMSE	MAE	R²	RMSE	MAE	R²	RMSE	MAE
CatBoost	0.934	3.707	2.950	0.943	3.428	2.748	0.918	3.576	2.814	0.917	4.308	3.388	0.916	0.943	0.779
XGBoost	0.918	4.155	3.321	0.928	3.847	3.257	0.903	3.888	3.514	0.903	4.667	3.757	0.901	1.024	0.820
LightGBM	0.926	3.933	3.047	0.937	3.582	2.872	0.913	3.683	3.092	0.913	4.405	3.530	0.911	0.972	0.794
GRU	0.922	4.047	3.154	0.933	3.699	2.940	0.908	3.787	3.196	0.908	4.546	3.659	0.907	0.994	0.796
LSTM	0.920	4.095	3.228	0.931	3.773	3.094	0.907	3.813	3.263	0.906	4.583	3.697	0.903	1.014	0.813

场景编号	调整参数
1	x₁~x₄,x₆~x₉
2	x₁~x₄,x₆~x₉,x₁₂
3	x₁~x₄,x₆~x₉,x₁₂,x₁₁