改进黏菌算法优化TCN−LSTM−MHSA的巷道锚杆（索）应力预测模型

齐俊艳; 车玉浩; 王磊; 袁瑞甫

doi:10.13272/j.issn.1671-251x.2025010025

改进黏菌算法优化TCN−LSTM−MHSA的巷道锚杆（索）应力预测模型

1.
河南理工大学软件学院，河南焦作　454000
2.
河南理工大学计算机科学与技术学院，河南焦作　454000
3.
河南理工大学能源科学与工程学院，河南焦作　454000

基金项目:

河南省高校科技创新团队支持计划（22IRTSTHN005）。

详细信息

作者简介:
齐俊艳（1978—），女，河南新乡人，副教授，研究方向为计算机网络控制、数据挖掘、工作流等，E-mail：qjywl@hpu.edu.cn

通讯作者:
车玉浩（1997—），男，河南新乡人，硕士研究生，研究方向为机器学习、预测算法、智能信息处理等，E-mail:yuhao_che@163.com。

中图分类号: TD353.6
计量
- 文章访问数: 23
- HTML全文浏览量: 1
- PDF下载量: 5
出版历程
- 收稿日期: 2025-01-09
- 修回日期: 2025-05-18
- 网络出版日期: 2025-05-27
- 刊出日期: 2025-05-14

TCN-LSTM-MHSA model optimized by improved slime mould algorithm for stress prediction of roadway anchor bolts (cables)

1.
School of Software, Henan Polytechnic University, Jiaozuo 454000, China
2.
School of Computer Science and Technology, Henan Polytechnic University, Jiaozuo 454000, China
3.
School of Energy Science and Engineering, Henan Polytechnic University, Jiaozuo 454000, China

摘要

摘要:
锚杆（索）应力的变化过程呈现明显的短期突变与长期时序依赖特征，而传统单一预测模型对长期趋势建模能力有限且对局部突变敏感性不足，往往难以全面捕捉上述复杂特征。针对该问题，提出一种基于改进黏菌算法（ISMA）优化时间卷积网络（TCN）−长短期记忆网络（LSTM）−多头自注意力机制（MHSA）的锚杆（索）应力预测模型。在煤矿巷道锚杆（索）应力预测问题中，模型训练过程通常涉及超参数调整、学习率选择等复杂优化任务，为提升模型的训练效率与预测精度，提出ISMA，引入邻域搜索与动态步长因子增强局部搜索能力，融合人工蜂群搜索机制提升全局搜索效率，有效增强模型跳出局部最优解的能力。TCN−LSTM−MHSA模型采用TCN提取局部时序特征，利用LSTM学习数据的长期依赖关系，通过MHSA强化对全局时序依赖的建模，从而提高模型对锚杆（索）应力的预测能力。在TCN−LSTM−MHSA模型的训练中利用ISMA对学习率进行迭代寻优，以提高模型的预测精度和速度。实验结果表明：① 与黏菌算法（SMA）、遗传算法（GA）、粒子群算法（PSO）、麻雀搜索算法（SSA）相比，ISMA优化策略在多个基准函数测试中表现出更优的收敛速度与寻优能力。② 在应力预测实验中，通过消融实验验证了TCN，LSTM，MHSA模块的必要性。③ ISMA优化TCN−LSTM−MHSA模型在MAE，RMSE及R²等指标上均优于BP，GRU等主流预测模型，具有更高的预测精度和稳定性。
- 锚杆(索)支护 /
- 锚杆(索)应力预测 /
- 黏菌算法 /
- 时间卷积网络 /
- 长短期记忆网络 /
- 多头自注意力机制
Abstract:
The variation process of anchor bolt (cable) stress exhibits distinct short-term fluctuations and long-term temporal dependencies. However, traditional single prediction models have limited capability in modeling long-term trends and insufficient sensitivity to local fluctuations, often making it difficult to fully capture these complex features. To address this problem, an anchor bolt (cable) stress prediction model based on an Improved Slime Mould Algorithm (ISMA) optimized Temporal Convolutional Network (TCN)-Long Short-Term Memory (LSTM)-Multi-Head Self-Attention (MHSA) architecture is proposed. In the problem of anchor bolt (cable) stress prediction in coal mine roadways, model training often involves complex optimization tasks such as hyperparameter tuning and learning rate selection. To improve the training efficiency and prediction accuracy of the model, ISMA was proposed, which enhanced local search capability by introducing neighborhood search and a dynamic step-size factor. Global search efficiency was improved through integrating an Artificial Bee Colony (ABC) search mechanism, thereby effectively improving the model's ability to escape from local optima. The TCN-LSTM-MHSA model was constructed by using TCN to extract local temporal features, employing LSTM to learn long-term dependencies in the data, and strengthening global temporal modeling through MHSA, thereby enhancing the prediction capability for anchor bolt (cable) stress. During training, ISMA was used to iteratively optimize the learning rate of the TCN-LSTM-MHSA model to improve prediction accuracy and speed. Experimental results showed that: ① Compared with the Slime Mould Algorithm (SMA), Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Sparrow Search Algorithm (SSA), the ISMA optimization strategy demonstrated better convergence speed and optimization ability in multiple benchmark function tests. ② In the stress prediction experiment, ablation experiments verified the necessity of TCN, LSTM, and MHSA modules. ③ The ISMA-optimized TCN-LSTM-MHSA model outperformed mainstream prediction models such as BP and GRU in MAE, RMSE, and R² metrics, showing higher prediction accuracy and stability.
- anchor bolt (cable) support /
- anchor bolt (cable) stress prediction /
- slime mould algorithm /
- temporal convolutional network /
- long short-term memory network /
- multi-head self-attention mechanism

HTML全文

图 1 SMA觅食过程

Figure 1. SMA foraging process

下载: 全尺寸图片幻灯片

图 2 ISMA寻优流程

Figure 2. ISMA optimization process

下载: 全尺寸图片幻灯片

图 3 TCN扩张因果卷积

Figure 3. Dilated causal convolution in TCN

下载: 全尺寸图片幻灯片

图 4 LSTM结构

Figure 4. LSTM structure

下载: 全尺寸图片幻灯片

图 5 MHSA输入矩阵的线性变换

Figure 5. Linear transformation of MHSA input matrix

下载: 全尺寸图片幻灯片

图 6 MHSA不同矩阵加权求和

Figure 6. Weighted summation of MHSA matrices

下载: 全尺寸图片幻灯片

图 7 TCN−LSTM−MHSA模型结构

Figure 7. Architectural framework of TCN-LSTM-MHSA

下载: 全尺寸图片幻灯片

图 8 ISMA优化预测模型过程

Figure 8. Process of ISMA-based prediction model optimization

下载: 全尺寸图片幻灯片

图 9 不同基准函数收敛曲线

Figure 9. Convergence curves of different benchmark functions

下载: 全尺寸图片幻灯片

图 10 锚杆（索）应力数据采集方式

Figure 10. Anchor bolts (cables) stress data collection method

下载: 全尺寸图片幻灯片

图 11 不同巷道锚杆（索）应力数据

Figure 11. Stress data of anchor bolts (cables) in different tunnels

下载: 全尺寸图片幻灯片

图 12 SMA与ISMA收敛率对比

Figure 12. Comparison of convergence rates between SMA and ISMA

下载: 全尺寸图片幻灯片

图 13 不同优化器优化学习率损失对比

Figure 13. Comparison of learning rate loss optimized by different optimizers

下载: 全尺寸图片幻灯片

图 14 不同模型预测结果

Figure 14. Predictions of different models

下载: 全尺寸图片幻灯片

表 1 基准测试函数

Table 1 Benchmarki functions

表达式	搜索空间	理论最优解
${A_1}({\boldsymbol{a}}) = \displaystyle\sum\limits_{n = 1}^D {{\boldsymbol{a}}_n^2} $	[−100，100]	0
${A_2}({\boldsymbol{a} }) = \displaystyle\sum\limits_{n = 1}^D \| { {\boldsymbol{a} }_n}\| + \displaystyle\prod\limits_{n = 1}^D \| { {\boldsymbol{a} }_n}\|$	[−10，10]	0
${A_3}({\boldsymbol{a}}) =\displaystyle\sum\limits_{n = 1}^D {\bigg(\displaystyle\sum\limits_{l = 1}^n {{{\boldsymbol{a}}_l}} \bigg)^2}$	[−100,100]	0

下载: 导出CSV

表 2 测试函数结果统计

Table 2 Statistical results of test functions

基准测试函数	算法	最优值
${A_1}$	PSO	2.31×10⁻¹⁰
	SSA	6.36×10⁻⁹
	GA	3.05×10⁻⁹
	SMA	2.03×10⁻¹⁰
	ISMA	1.35×10⁻¹⁰
${A_2}$	PSO	1.59×10⁻⁹
	SSA	4.13×10⁻¹¹
	GA	9.48×10⁻¹¹
	SMA	9.01×10⁻¹²
	ISMA	8.31×10⁻¹²
${A_3}$	PSO	1.34×10⁻⁹
	SSA	1.43×10⁻⁸
	GA	4.86×10⁻⁸
	SMA	1.20×10⁻⁹
	ISMA	1.05×10⁻⁹

下载: 导出CSV

表 3 锚杆（索）部分应力数据

Table 3 Partial stress data of anchor bolts (cables)

应力值/kN	采集时间（年−月−日T时∶分∶秒）
63.8	2024−09−03T 15:15:27
63.4	2024−09−03T 15:38:58
62.9	2024−09−03T 16:02:28
62.7	2024−09−03T 16:25:58
62.3	2024−09−03T 16:49:28
62.7	2024−09−03T 17:12:59
62.9	2024−09−03T 17:36:30
62.9	2024−09−03T 18:00:02
63.1	2024−09−03T 18:23:33
63.4	2024−09−03T 18:47:03

下载: 导出CSV

表 4 各模型相应参数

Table 4 Parameters of different models

模型层次	名称	参数值
TCN层	卷积核大小	3
	残差块	3
	空间丢失因子	2
LSTM层	隐含层神经元	128
MHSA层	注意力头	4
MHSA层	维度	32

下载: 导出CSV

表 5 黏菌算法相应参数

Table 5 Parameters of SMA

名称	参数值
n_agents	20
n_variables	1
lower_bound	0.000 01
upper_bound	0.1
max_iter	10

下载: 导出CSV

表 6 消融实验结果

Table 6 Ablation experiment results

地点	模型结构	MAE	RMSE	${R^2}$
1800−B1	LSTM+MHSA	0.779	0.929	0.989
	TCN+MHSA	2.058	2.119	0.946
	TCN+LSTM	0.629	0.805	0.992
	TCN−LSTM−MHSA	0.414	0.569	0.996
3302−D1	LSTM+MHSA	2.054	3.264	0.948
	TCN+MHSA	1.611	2.654	0.954
	TCN+LSTM	2.368	3.657	0.935
	TCN−LSTM−MHSA	1.521	2.608	0.967

下载: 导出CSV

表 7 各模型性能表现评价

Table 7 Performance evaluation of each model

设备	模型	MAE	RMSE	${R^2}$
运输巷1800−B1	BP	0.565	0.744	0.993
	GRU	0.867	1.108	0.985
	ISMA−TCN−LSTM−MHSA	0.414	0.569	0.996
运输巷中5个设备平均	BP	0.581	0.756	0.991
	GRU	0.869	1.112	0.981
	ISMA−TCN−LSTM−MHSA	0.417	0.571	0.994
回风巷3302−D1	BP	2.026	3.305	0.947
	GRU	2.132	3.549	0.939
	ISMA−TCN−LSTM−MHSA	1.521	2.608	0.967
回风巷中5个设备平均	BP	2.035	3.311	0.932
	GRU	2.143	3.557	0.928
	ISMA−TCN−LSTM−MHSA	1.530	2.616	0.961

下载: 导出CSV

参考文献(28)

[1]	康红普,姜鹏飞,刘畅,等. 煤巷锚杆支护施工装备现状及发展趋势[J]. 工矿自动化,2023,49(1):1-18. KANG Hongpu,JIANG Pengfei,LIU Chang,et al. Current situation and development trend of rock bolting construction equipment in coal roadway[J]. Journal of Mine Automation,2023,49(1):1-18.
[2]	张农,魏群,吴建生. 煤矿巷道喷涂柔膜技术及适用性[J]. 煤炭科学技术,2022,50(1):78-85. DOI: 10.3969/j.issn.0253-2336.2022.1.mtkxjs202201006 ZHANG Nong,WEI Qun,WU Jiansheng. Spray-on membrane technology and its applicability in coal mine roadways[J]. Coal Science and Technology,2022,50(1):78-85. DOI: 10.3969/j.issn.0253-2336.2022.1.mtkxjs202201006
[3]	李永亮,杨仁树,温明睿,等. 煤矿巷道顶板锚索受力特征与分区锚固机理[J]. 煤炭科学技术,2022,50(5):73-83. LI Yongliang,YANG Renshu,WEN Mingrui,et al. Stressed characteristics and regional anchoring mechanism of cable bolts in coal mine roadway roof[J]. Coal Science and Technology,2022,50(5):73-83.
[4]	张哲诚,张向东,刘源浩,等. 横向简谐荷载作用下端锚黏结式锚杆黏结性试验研究[J]. 煤炭学报,2016,41(6):1407-1415. ZHANG Zhecheng,ZHANG Xiangdong,LIU Yuanhao,et al. Experimental research of anchors bonding mechanical properties under transverse harmonic loads[J]. Journal of China Coal Society,2016,41(6):1407-1415.
[5]	董建军,谢郑权,杨嫡,等. 基于FBG传感器的回采巷道锚杆支护监测分析[J]. 安全与环境学报,2021,21(5):2013-2021. DONG Jianjun,XIE Zhengquan,YANG Di,et al. Monitoring and analysis of the bolt supporting for the mining roadways based on the FBG sensor[J]. Journal of Safety and Environment,2021,21(5):2013-2021.
[6]	原钢,刘杰. 基于多参数输入与输出高斯过程回归的锚杆支护状态预测[J]. 液压气动与密封,2023,43(11):47-50. DOI: 10.3969/j.issn.1008-0813.2023.11.009 YUAN Gang,LIU Jie. Prediction of anchor bolt support status with multi-parameter input and output gaussian process regression[J]. Hydraulics Pneumatics & Seals,2023,43(11):47-50. DOI: 10.3969/j.issn.1008-0813.2023.11.009
[7]	徐毅青,邓绍玉,葛琦. 锚索预应力初期与长期损失的预测模型研究[J]. 岩土力学,2020,41(5):1663-1669. XU Yiqing,DENG Shaoyu,GE Qi. Prediction models for short-term and long-term pre-stress loss of anchor cable[J]. Rock and Soil Mechanics,2020,41(5):1663-1669.
[8]	彭泓,刘亚飞. 基于光纤光栅技术的巷道支护锚杆受力监测[J]. 煤炭科学技术,2022,50(6):61-67. PENG Hong,LIU Yafei. Stress monitoring of roadway supporting bolt based on fiber bragg grating technology[J]. Coal Science and Technology,2022,50(6):61-67.
[9]	张涵,赵建利. 巴基斯坦某水电站工程高边坡支护锚杆应力监测分析[J]. 水利科学与寒区工程,2024,7(7):101-107. ZHANG Han,ZHAO Jianli. Analysis on stress monitoring and high slope support of a hydropower project in Pakistan[J]. Hydro Science and Cold Zone Engineering,2024,7(7):101-107.
[10]	林东凤,黄汉明,沈俏. 基于改进遗传算法的广度架构搜索算法[J]. 计算机工程与设计,2024,45(12):3667-3673. LIN Dongfeng,HUANG Hanming,SHEN Qiao. Wide architecture search algorithm based on improved genetic algorithm[J]. Computer Engineering and Design,2024,45(12):3667-3673.
[11]	易云飞,王志勇,施运应. 基于蜣螂优化的改进粒子群算法[J]. 重庆邮电大学学报(自然科学版),2024,36(3):533-542. DOI: 10.3979/j.issn.1673-825X.202306020178 YI Yunfei,WANG Zhiyong,SHI Yunying. The improved particle swarm optimization algorithm based on dung beetle optimization[J]. Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition),2024,36(3):533-542. DOI: 10.3979/j.issn.1673-825X.202306020178
[12]	卢磊,贺智明,黄志成. 基于多策略改进的麻雀搜索算法[J]. 计算机与现代化,2023(10):23-31. DOI: 10.3969/j.issn.1006-2475.2023.10.004 LU Lei,HE Zhiming,HUANG Zhicheng. An improved sparrow search algorithm based on multi-strategy[J]. Computer and Modernization,2023(10):23-31. DOI: 10.3969/j.issn.1006-2475.2023.10.004
[13]	黄元春,张凌波. 改进的鲸鱼优化算法及其应用[J]. 计算机工程与应用,2019,55(21):220-226,270. DOI: 10.3778/j.issn.1002-8331.1901-0296 HUANG Yuanchun,ZHANG Lingbo. Improved whale optimization algorithm and its application[J]. Computer Engineering and Applications,2019,55(21):220-226,270. DOI: 10.3778/j.issn.1002-8331.1901-0296
[14]	MONISMITH D R,MAYFIELD B E. Slime Mold as a model for numerical optimization[C]. Swarm Intelligence Symposium,St. Louis,2008. DOI: 10.1109/SIS.2008.4668295.
[15]	LI Shimin,CHEN Huiling,WANG Mingjing,et al. Slime mould algorithm:a new method for stochastic optimization[J]. Future Generation Computer Systems,2020,111:300-323. DOI: 10.1016/j.future.2020.03.055
[16]	HOUSSEIN E H,MAHDY M A,BLONDIN M J,et al. Hybrid slime mould algorithm with adaptive guided differential evolution algorithm for combinatorial and global optimization problems[J]. Expert Systems with Applications,2021,174. DOI: 10.1016/j.eswa.2021.114689.
[17]	HUSAIN S,AKHMETOV M,KANYMKULOV D,et al. Optimization of behavioral model of VO₂ switches using slime mould algorithm[C]. International Symposium on Networks,Computers and Communications,Doha,2023. DOI: 10.1109/ISNCC58260.2023.10323871.
[18]	DUAN Zaixin,QIAN Xuezhong,SONG Wei. Multi-strategy enhanced slime mould algorithm for optimization problems[J]. IEEE Access,2025,13:7850-7871. DOI: 10.1109/ACCESS.2025.3527509
[19]	薛贵军,赵广昊,史彩娟. 基于改进黏菌算法优化BiLSTM的短期供热负荷控制预测[J]. 沈阳工业大学学报,2024,46(4):434-441. DOI: 10.7688/j.issn.1000-1646.2024.04.12 XUE Guijun,ZHAO Guanghao,SHI Caijuan. Short-term heating load prediction based on improved slime mould algorithm optimized BiLSTM[J]. Journal of Shenyang University of Technology,2024,46(4):434-441. DOI: 10.7688/j.issn.1000-1646.2024.04.12
[20]	王岩,王聪英,申艳梅. 改进的蜂群优化聚类集成联合相似度推荐算法[J]. 计算机工程,2020,46(10):88-94,102. WANG Yan,WANG Congying,SHEN Yanmei. Clustering ensemble joint similarity recommendation algorithm optimized by improved bee colony[J]. Computer Engineering,2020,46(10):88-94,102.
[21]	陶子君,陆芷,蒙炳金. 大规模符号网络划分的学习驱动型扩展变邻域搜索算法[J]. 计算机应用研究,2025,42(3):770-776. TAO Zijun,LU Zhi,MENG Bingjin. Learning-driven extended variable neighborhood search for signed graph partitioning[J]. Application Research of Computers,2025,42(3):770-776.
[22]	杜晓昕,牛丽明,王波,等. 基于邻域搜索策略的蜣螂优化算法及应用[J]. 广西师范大学学报(自然科学版),2025,43(2):149-167. DU Xiaoxin,NIU Liming,WANG Bo,et al. Dung beetle optimization algorithm based on neighborhood search strategy and application[J]. Journal of Guangxi Normal University(Natural Science Edition) ,2025,43(2):149-167.
[23]	冯腾飞,刘小生,钟钰,等. 矿区边坡变形预测的IGM−LSSVM模型[J]. 金属矿山,2019(3):168-172. FENG Tengfei,LIU Xiaosheng,ZHONG Yu,et al. Slope deformation prediction in mining area based on IGM-LSSVM model[J]. Metal Mine,2019(3):168-172.
[24]	IBRAHIM E A,VAN DEN DOOL B,DE S,et al. Dilate-invariant temporal convolutional network for real-time edge applications[J]. IEEE Transactions on Circuits and Systems I:Regular Papers,2022,69(3):1210-1220. DOI: 10.1109/TCSI.2021.3124219
[25]	刘辉,凌宁青,罗志强,等. 基于TCN−LSTM和气象相似日集的电网短期负荷预测方法[J]. 智慧电力,2022,50(8):30-37. DOI: 10.3969/j.issn.1673-7598.2022.08.007 LIU Hui,LING Ningqing,LUO Zhiqiang,et al. Power grid short-term load forecasting method based on TCN-LSTM and meteorological similar day sets[J]. Smart Power,2022,50(8):30-37. DOI: 10.3969/j.issn.1673-7598.2022.08.007
[26]	王珺,王然风,魏凯,等. 基于时间序列对齐和TCNformer的重介精煤灰分多步预测[J]. 工矿自动化,2024,50(5):60-66. WANG Jun,WANG Ranfeng,WEI Kai,et al. Multi step prediction of dense medium clean coal ash content based on time series alignment and TCNformer[J]. Journal of Mine Automation,2024,50(5):60-66.
[27]	HOCHREITER S,SCHMIDHUBER J. Long short-term memory[J]. Neural Computation,1997,9(8):1735-1780. DOI: 10.1162/neco.1997.9.8.1735
[28]	DE JESUS D A R,MANDAL P,CHAKRABORTY S,et al. Solar PV power prediction using a new approach based on hybrid deep neural network[C]. IEEE Power & Energy Society General Meeting,Atlanta,2019. DOI: 10.1109/PESGM40551.2019.8974091.

施引文献(9)

期刊类型引用(4)

1.	王羿帆，杨维，戴蒙天. 一种基于三轴正交接收线圈的磁感应透地通信方案. 中南大学学报(自然科学版). 2024(03): 973-981 . 百度学术
2.	付天晖，刘宝衡，王永斌. 透地无线通信研究综述. 科学技术与工程. 2021(17): 6993-7001 . 百度学术
3.	赵健. 大数据通信仪表的自动化传输控制模型研究. 制造业自动化. 2021(08): 104-108 . 百度学术
4.	邓羽捷，付天晖，王永斌，修梦雷. 透地通信研究概况及其主流技术. 舰船电子工程. 2021(11): 13-17 . 百度学术