Page 101 - 《振动工程学报》2026年第5期
P. 101
第 5 期 赵 玲,等:大语言模型在跨座式单轨列车齿轮箱振动信号预测中的应用 1305
有一定局限性:一方面,它对高质量训练数据的依赖 Systems and Signal Processing,2019,115:213-237.
使得在故障样本稀缺的情况下性能可能不理想;另 [12] YIN S K,FU C Y,ZHAO S R,et al. A survey on multi-
一方面,模型的高计算复杂度限制了其在实时预测 modal large language models[J]. National Science Review,
2024,11(12):277-296.
中的部署。此外,模型在不同机械系统和工况下的
[13] VASWANI A, SHAZEER N, PARMAR N, et al. Atten-
泛化能力仍需进一步验证和提升。未来,团队将优
tion is all you need[J]. Advances in Neural Information
化 VSP-LLM 架 构 , 减 少 模 型 参 数 量 并 保 证 预 测 性
Processing Systems,2017,30:1-11.
能,实现模型轻量化,以满足实时应用需求。同时, [14] GRAVES A. Long short-term memory[M]. Supervised
探索迁移学习和小样本学习技术,增强模型的泛化 Sequence Labelling with Recurrent Neural Networks. Berlin,
能力,并扩展多模态数据融合方法,为更广泛的机械 Heidelberg:Springer,2012:37-45.
系统振动信号预测提供高效可靠的解决方案。 [15] DEY R,SALEM F M. Gate-variants of gated recurrent unit
(GRU) neural networks[C]//Proceedings of IEEE 60th Inter-
national Midwest Symposium on Circuits and Systems
参考文献:
(MWSCAS). 2017:1597-1600.
[16] VAN DEN OORD A, DIELEMAN S, ZEN H G, et al.
[1] HE Q L,CAI C B,ZHU S Y,et al. An improved dynamic WaveNet:a generative model for raw audio[EB/OL]. (2016-
model of suspended monorail train-bridge system considering 09-12). https://arXiv.org/abs/1609.03499v2.
a tyre model with patch contact[J]. Mechanical Systems and [17] ZHOU H Y, ZHANG S H, PENG J Q, et al. Informer:
Signal Processing,2020,144:106865. beyond efficient transformer for long sequence time-series
[2] WEN X X, NIE Y, DU Z X, et al. Operational safety forecasting[J]. Proceedings of the AAAI Conference on Artifi-
assessment of straddle-type monorail vehicle system based on cial Intelligence,2021,35(12):11106-11115.
cloud model and improved CRITIC method[J]. Engineering [18] JIN M, WANG S Y, MA L T, et al. Time-LLM: time
Failure Analysis,2022,139:106463. series forecasting by reprogramming large language models
[3] PALKA R,WORONOWICZ K. Linear induction motors in [EB/OL]. (2023-10-03). https://arXiv.org/abs/2310.01728v2.
transportation systems[J]. Energies,2021,14(9):2549. [19] XU X Y, LU H P, HUA X, et al. An investigation on
[4] GAO Q F,CUI K M,LI Z L,et al. Numerical investiga- dynamic characteristics of the straddle-type monorail system
tion of the dynamic performance and riding comfort of a strad- under spatial alignment[J]. Machines,2022,10(9):724.
dle-type monorail subjected to moving trains[J]. Applied [20] HORNIK K. Approximation capabilities of multilayer feedfor-
Sciences,2020,10(15):5258. ward networks[J]. Neural Networks,1991,4(2):251-257.
[5] ZHU B,ZHANG C R,ZHANG Z,et al. Dynamic analysis [21] KRIZHEVSKY A, SUTSKEVER I, HINTON G E.
of suspension-type monorail long-span cable-stayed bridge in a ImageNet classification with deep convolutional neural
wind-vehicle-bridge system[J]. Urban Rail Transit, 2023, networks[J]. Communications of the ACM, 2017, 60( 6) :
9(1):19-30. 84-90.
[6] BOKRANTZ J, SKOOGH A, BERLIN C, et al. Smart [22] HUANG X, BELONGIE S. Arbitrary style transfer in real-
Maintenance: a research agenda for industrial maintenance time with adaptive instance normalization[C]//Proceedings of
management[J]. International Journal of Production IEEE International Conference on Computer Vision (ICCV).
Economics,2020,224:107547. 2017:1510-1519.
[7] GUO J W,LAO Z P,HOU M,et al. Mechanical fault time [23] SUN C X,LI H Y,LI Y L,et al. TEST:text prototype
series prediction by using EFMSAE-LSTM neural network[J]. aligned embedding to activate LLM’s ability for time
Measurement,2021,173:108566. series[EB/OL]. (2023-08-16). https://arXiv.org/abs/2308.08241.
[8] YAN R Q,GAO R X,CHEN X F. Wavelets for fault diag- [24] WU H X, XU J H, WANG J M, et al. Autoformer:
nosis of rotary machines: a review with applications[J]. decomposition transformers with auto-correlation for long-
Signal Processing,2014,96:1-15. term series forecasting[C]//Proceedings of Neural Information
[9] ZHENG J D,CHENG J S,YANG Y. A rolling bearing fault Processing Systems,2021,34:22419-22430.
[25] ZENG A L,CHEN M X,ZHANG L,et al. Are transform-
diagnosis approach based on LCD and fuzzy entropy[J]. Mech-
anism and Machine Theory,2013,70:441-453. ers effective for time series forecasting?[J]. Proceedings of the
[10] WU Z H,HUANG N E. Ensemble empirical mode decompo- AAAI Conference on Artificial Intelligence,2023,37(9):
sition:a noise-assisted data analysis method[J]. Advances in 11121-11128.
Adaptive Data Analysis,2009,1(1):1-41.
[11] ZHAO R,YAN R Q,CHEN Z H,et al. Deep learning and 通信作者:赵 玲(1979—),女,博士,教授。
its applications to machine health monitoring[J]. Mechanical E-mail:zhao.ling@163.com

