[1]刘琳,杨玉晗,梅勇,等.基于改进Transformer的木质粉尘质量浓度预测模型研究*[J].中国安全生产科学技术,2025,21(8):110-118.[doi:10.11731/j.issn.1673-193x.2025.08.014]
LIU Lin,YANG Yuhan,MEI Yong,et al.Research on prediction model of wood dust mass concentration based on improved Transformer[J].Journal of Safety Science and Technology,2025,21(8):110-118.[doi:10.11731/j.issn.1673-193x.2025.08.014]
点击复制
基于改进Transformer的木质粉尘质量浓度预测模型研究*
LIU Lin,YANG Yuhan,MEI Yong,et al.Research on prediction model of wood dust mass concentration based on improved Transformer[J].Journal of Safety Science and Technology,2025,21(8):110-118.[doi:10.11731/j.issn.1673-193x.2025.08.014]
《中国安全生产科学技术》[ISSN:1673-193X/CN:11-5335/TB]
- 卷:
- 21
- 期数:
- 2025年8期
- 页码:
- 110-118
- 栏目:
- 职业安全卫生管理与技术
- 出版日期:
- 2025-08-30
文章信息/Info
- Title:
- Research on prediction model of wood dust mass concentration based on improved Transformer
- 文章编号:
- 1673-193X(2025)-08-0110-09
- 关键词:
- 粉尘质量浓度; Transformer模型; 长时间序列预测; 机器学习; 注意力机制
- Keywords:
- dust mass concentration; Transformer model; long-term time series prediction; machine learning; attention mechanism
- 分类号:
- X913
- 文献标志码:
- A
- 摘要:
- 为解决现有非平稳时间序列预测模型（如LSTM）对粉尘质量浓度数据的长期依赖关系表征不足、以及多步预测准确率低等问题，提出1种改进的Transformer模型并用于木质粉尘质量浓度多变量多步预测。该模型首先引入多时间分辨率模块，以增强对时间序列中短期与长期依赖关系的捕捉能力；其次，采用稀疏图自注意力机制替代原有注意力机制，用于减少计算量，同时使模型能够充分学习序列之间的隐式空间依赖关系。研究结果表明：改进的Transformer模型在短期（1 h和2 h）、中期（6 h和12 h）和长期（18 h和24 h）预测任务中均表现出优异的性能，在中长期预测任务中与原始模型相比在预测精度上提升了50%以上。研究结果可为提高粉尘质量浓度预测精度、实现职业健康预警与安全生产管理提供理论与方法参考。
- Abstract:
- In order to address the limitations of existing non-stationary time series prediction models (e.g.,LSTM) in capturing long-term dependencies and their low accuracy in multi-step forecasting of dust mass concentration data,this study proposed an improved Transformer model for multi-variable multi-step prediction of wood dust mass concentration.The model first introduces a multi-resolution temporal module to enhance the capture ability of both short-term and long-term dependencies in time series.Secondly,it replaces the original attention mechanism with a sparse graph self-attention mechanism to reduce computational complexity while enabling the model to fully learn implicit spatial dependencies between sequences.The results demonstrate that the improved Transformer model exhibits excellent performance in short-term (1 h and 2 h),mid-term (6 h and 12 h),and long-term (18 h and 24 h) prediction tasks,improving prediction accuracy by over 50% compared to the original model in med-term and long-term prediction tasks.These findings provide theoretical and methodological references for enhancing dust mass concentration prediction accuracy,enabling occupational health warnings and safety production management.
参考文献/References:
［1］王爱红,冷朋波,李晓海,等.“用人单位职业病危害综合风险评估方法”在木质家具企业职业卫生分类管理中的应用［J］.环境与职业医学,2023,40(6):634-640,647.
WANG Aihong,LENG Pengbo,LI Xiaohai,et al.Application of a comprehensive risk assessment method for occupational disease hazards of employers in occupational health classification management of wooden furniture manufacturing factories ［J］.Journal of Environmental and Occupational Medicine,2023,40(6):634-640,647.
［2］聂志超,田甜,李婷婷,等.新加坡模型在小型家具制造企业接木粉尘岗位职业健康风险评估中的应用［J］.环境与职业医学,2020,37(11):1100-1105.
NIE Zhichao,TIAN Tian,LI Tingting,et al.Application of Singapore model in occupational health risk assessment of workstations exposed to wood dust in small furniture manufacturing enterprises ［J］.Journal of Environmental and Occupational Medicine,2020,37(11):1100-1105.
［3］周福宝,袁亮,程卫民,等.矿井粉尘职业健康防护技术2013—2023年研究进展［J］.中国安全生产科学技术,2023,19(12):5-15.
ZHOU Fubao,YUAN Liang,CHENG Weimin,et al.Research progress on occupational health protection technology of mine dust from 2013 to 2023 ［J］.Journal of Safety Science and Technology,2023,19(12):5-15.
［4］冯恒原,李治刚,朱芷涵,等.高瓦斯矿井综掘工作面粉尘运移规律及富集特征研究［J］.中国安全生产科学技术,2023,19(10):59-65.
FENG Hengyuan,LI Zhigang,ZHU Zhihan,et al.Study on dust migration law and enrichment characteristics of fully-mechanized heading face in high gas mines ［J］.Journal of Safety Science and Technology,2023,19(10):59-65.
［5］杨秋月,王海椒.木粉尘职业暴露与肺癌关系病例对照研究的Meta分析［J］.环境与职业医学,2018,35(3):266-271.
YANG Qiuyue,WANG Haijiao.Meta analysis of case-control studies on relationship between occupational exposure to wood dust and lung cancer ［J］.Journal of Environmental and Occupational Medicine,2018,35(3):266-271.
［6］赵耀忠,严俊龙,任吉凯,等.基于机器学习的露天煤矿粉尘浓度预测［J］.煤炭工程,2022,54(增刊1):157-161.
ZHAO Yaozhong，YAN Junlong，REN Jikai，et al.Dust concentration prediction of open-pit coal mine based on machine learning［J］.Coal Engineering,2022,54(Supplement 1):157-161.
［7］CUI B,LIU M,LI S,et al.Deep learning methods for atmospheric PM2.5 prediction:a comparative study of transformer and CNN-LSTM-attention［J］.Atmospheric Pollution Research,2023,14(9):101833.
［8］谭聪,谢艳芝,任金瑶,等.打磨车间粉尘控制技术数值模拟［J］.安全与环境学报,2023,23(7):2352-2359.
TAN Cong,XIE Yanzhi,REN Jinyao,et al.Numerical simulation of dust control technology in grinding workshops ［J］.Journal of Safety and Environment,2023,23(7):2352-2359.
［9］赵洪宝,翟汝鹏,戈海宾,等.基于统计学模型的矿区粉尘污染特征及影响因素的定量分析［J］.矿业科学学报,2024,9(2):243-257.
ZHAO Hongbao,ZHAI Rupeng,GE Haibin,et al.Quantitative analysis of dust pollution characteristics and influencing factors in mining areas based on statistical modelling ［J］.Journal of Mining Science and Technology,2024,9(2):243-257.
［10］GUO Q,HE Z,WANG Z.Prediction of hourly PM2.5 and PM10 concentrations in Chongqing City in China based on artificial neural network［J］.Aerosol and Air Quality Research,2023,23(6):220448.
［11］霍文,栾博钰,周伟,等.基于环境因素的露天煤矿粉尘质量浓度预测［J］.辽宁工程技术大学学报(自然科学版),2021,40(5):409-414.
HUO Wen,LUAN Boyu,ZHOU Wei,et al.Prediction of dust mass concentration in open-pit coal mine based on environmental factors ［J］.Journal of Liaoning Technical University (Natural Science Edition),2021,40(5):409-414.
［12］LIU Z,ZHANG R,MA J,et al.Analysis and prediction of the meteorological characteristics of dust concentrations in open-pit mines［J］.Sustainability,2023,15(6):4837.
［13］王栋,张江石,张琪,等.基于HPO-BP神经网络的隧道粉尘处理预测模型［J］.工业建筑,2023,53(增刊1):561-564，574.
WANG Dong,ZHANG Jiangshi,ZHANG Qi,et al.Tunnel dust treatment prediction model based on HPO-BP neural network ［J］.Industrial Construction,2023,53(Supplement 1):561-564，574.
［14］VASWANI A,SHAZEER N,PAEMAR N,et al.Attention is all you need［EB/OL］.（2017-06-12）［2025-08-01］.https://arxiv.org/abs/1706.03762.
［15］YANG Y,LU J.Foreformer:an enhanced transformer-based framework for multivariate time series forecasting［J］.Applied Intelligence,2023,53(10):12521-12540.
［16］WANG Y,LONG H,ZHENG L,et al.Graphformer:adaptive graph correlation transformer for multivariate long sequence time series forecasting［J］.Knowledge-Based Systems,2024,285:111321.
［17］ZHOU H,ZHANG S,PENG J,et al.Informer:beyond efficient transformer for long sequence time-series forecasting［C］//Proceedings of the AAAI Conference on Artificial Intelligence.2021,35(12):11106-11115.
［18］董浩,孙琳,欧阳峰.基于Informer的PM2.5浓度预测［J］.环境工程,2022,40(6):48-54，62.
DONG Hao,SUN Lin,OUYANG Feng.Prediction of PM2.5 concentration based on Informer ［J］.Environmental Engineering,2022,40(6):48-54，62.
［19］赵生昊,李家启,刘俊,等.基于多变量时间序列预测和XGBoost的O3浓度短期预报研究［J］.热带气象学报,2025,41(1):125-135.
ZHAO Shenghao,LI Jiaqi,LIU Jun,et al.Short-Term forecasting of ozone concentration using multivariate time series and XGBoost ［J］.Journal of Tropical Meteorology,2025,41(1):125-135.
［20］汤健,梁永琪,夏恒,等.基于TS模糊森林回归的MSWI过程粉尘浓度预测［J/OL］.北京工业大学学报,1-11［2025-08-07］.https://link.cnki.net/urlid/11.2286.t.20250529.1910.020.
相似文献/References:
[1]时训先,邓权龙,蒋仲安,等.石棉破碎车间粉尘质量浓度分布的数值模拟及实测[J].中国安全生产科学技术,2017,13(6):179.[doi:10.11731/j.issn.1673-193x.2017.06.030]
SHI Xunxian,DENG Quanlong,JIANG Zhongan,et al.Numerical simulation and measurement on mass concentration distribution of dust in asbestos crushing plant[J].Journal of Safety Science and Technology,2017,13(8):179.[doi:10.11731/j.issn.1673-193x.2017.06.030]
[2]王志荣.T形掘进巷粉尘污染规律及司机接尘特征研究[J].中国安全生产科学技术,2025,21(7):78.[doi:10.11731/j.issn.1673-193x.2025.07.010]
WANG Zhirong.Research on dust pollution patterns in T-shape excavation roadways and dust exposure characteristics for roadheader operators[J].Journal of Safety Science and Technology,2025,21(8):78.[doi:10.11731/j.issn.1673-193x.2025.07.010]
[3]刘丹丹,王威廉,李德文,等.CCD散射图样法测量粉尘质量浓度影响因素研究*[J].中国安全生产科学技术,2024,20(9):89.[doi:10.11731/j.issn.1673-193x.2024.09.010]
LIU Dandan,WANG Weilian,LI Dewen,et al.Study on influencing factors of dust mass concentration measurement based on CCD scattering pattern method[J].Journal of Safety Science and Technology,2024,20(8):89.[doi:10.11731/j.issn.1673-193x.2024.09.010]
备注/Memo
- 备注/Memo:
-
收稿日期： 2024-12-03
* 基金项目： 国家自然科学基金项目（51805382）；湖北省科技创新人才计划项目（2024DJC043）；湖北省应急管理厅应急能力与安全生产专项资金项目（2023016）
作者简介： 刘琳，硕士研究生，主要研究方向为职业与环境危害因素检测与评价。
通信作者： 易灿灿，博士，副教授，主要研究方向为机械设备故障诊断与信号处理。
