结构地震响应预测大语言模型

doi:10.3778/j.issn.1002-8331.2403-0074

摘要/Abstract

摘要： 建筑结构的地震响应预测是基于性能的地震工程中建筑评估的重要组成部分。聚焦于地震响应预测中的少样本场景，提出了一种融合了大语言模型（large language model，LLM）和提示学习的震动响应预测方法LLM-PaP。该方法将LLM对时序数据的通用分析能力迁移到地震响应预测中，以克服一般模型在小样本条件下的性能缺陷。在模型中引入“PaP（prompt-as-prefix）”思想，为时间序列增加自然语言任务指令和地震输入序列数据的统计信息，以增强模型对输入序列的理解能力并引导推理预测过程。在两个数据算例的实验中验证了所提方法的有效性。结果表明：LLM-PaP在数据集上的预测性能显著优于基于MLP、频域和Transformer等先进预测方法。进一步的泛化性实验结果揭示了LLM-PaP在跨数据集适应上的卓越性能。LLM-PaP为地震响应预测任务提供了一种创新性的解决方案，为未来大模型与震动响应预测领域的交叉性研究提供了新的思路和方法。

关键词: 地震响应预测, 大语言模型（LLM）, 提示学习, 时间序列

Abstract: The prediction of seismic responses in structural engineering is a critical component of building assessment, particularly in performance-based seismic engineering. This paper focuses on scenarios with limited data samples in seismic response prediction and proposes a method termed LLM-PaP (large language model with prompt-as-prefix). This method leverages the general analytical capabilities of large language models (LLMs) for time-series data to address performance deficiencies typically encountered by conventional models under small sample conditions. The model integrates the “PaP (prompt-as-prefix)” concept, enhancing the understanding of input sequences by incorporating natural language task instructions and statistical information of seismic input sequence data to guide the reasoning and prediction process. Experimental validation conducted on two datasets demonstrates the effectiveness of the proposed method. The results indicate that LLM-PaP significantly outperforms advanced prediction methods based on MLP, frequency domain analysis, and Transformer models in terms of predictive performance on the datasets. Additional experiments on generalization reveal the superior adaptability of LLM-PaP across datasets. LLM-PaP presents an innovative solution for seismic response prediction tasks, offering new insights and methods for the interdisciplinary research at the large models and seismic response prediction in the future.

Key words: seismic response prediction, large language model (LLM), prompt learning, time series

郭茂祖, 张欣欣, 赵玲玲, 张庆宇. 结构地震响应预测大语言模型[J]. 计算机工程与应用, 2025, 61(16): 132-145.

GUO Maozu, ZHANG Xinxin, ZHAO Lingling, ZHANG Qingyu. Seismic Response Prediction of Structures Using Large Language Models[J]. Computer Engineering and Applications, 2025, 61(16): 132-145.

参考文献

[1] WEN W P, ZHANG C Y, ZHAI C H. Rapid seismic response prediction of RC frames based on deep learning and limited building information[J]. Engineering Structures, 2022, 267: 114638.
[2] BEHMANESH I, YOUSEFIANMOGHADAM S, NOZARI A, et al. Uncertainty quantification and propagation in dynamic models using ambient vibration measurements, application to a 10-story building[J]. Mechanical Systems and Signal Processing, 2018, 107: 502-514.
[3] CELIK O C, ELLINGWOOD B R. Seismic fragilities for non-ductile reinforced concrete frames-Role of aleatoric and epistemic uncertainties[J]. Structural Safety, 2010, 32(1): 1-12.
[4] VAMVATSIKOS D, FRAGIADAKIS M. Incremental dynamic analysis for estimating seismic performance sensitivity and uncertainty[J]. Earthquake Engineering & Structural Dynamics, 2010, 39(2): 141-163.
[5] YIN Y J, LI Y. Seismic collapse risk of light-frame wood construction considering aleatoric and epistemic uncertainties[J]. Structural Safety, 2010, 32(4): 250-261.
[6] DE LAUTOUR O R, OMENZETTER P. Prediction of seismic-induced structural damage using artificial neural networks[J]. Engineering Structures, 2009, 31(2): 600-606.
[7] 赵煜东, 许卫晓, 于德湖, 等. 基于人工神经网络的RC框架结构地震响应预测方法[J]. 地震研究, 2024, 47(1): 123-134.
ZHAO Y D, XU W X, YU D H, et al. Response prediction method of the RC frame structure based on the artificial neural network[J]. Journal of Seismological Research, 2024, 47(1): 123-134.
[8] NGUYEN H D, DAO N D, SHIN M. Prediction of seismic drift responses of planar steel moment frames using artificial neural network and extreme gradient boosting[J]. Engineering Structures, 2021, 242: 112518.
[9] MENON U A, NAIR D S. Seismic response prediction of RC buildings using artificial neural network[C]//Proceedings of SECON2022: Structural Engineering and Construction Management. Cham: Springer, 2022: 403-413.
[10] HWANG S H, MANGALATHU S, SHIN J, et al. Machine learning-based approaches for seismic demand and collapse of ductile reinforced concrete building frames[J]. Journal of Building Engineering, 2021, 34: 101905.
[11] KIM T, SONG J, KWON O S. Probabilistic evaluation of seismic responses using deep learning method[J]. Structural Safety, 2020, 84: 101913.
[12] OH B K, GLISIC B, PARK S W, et al. Neural network-based seismic response prediction model for building structures using artificial earthquakes[J]. Journal of Sound and Vibration, 2020, 468: 115109.
[13] XU Y J, LU X Z, CETINER B, et al. Real-time regional seismic damage assessment framework based on long short-term memory neural network[J]. Computer-Aided Civil and Infrastructure Engineering, 2021, 36(4): 504-521.
[14] LIN K Y, LIN T K, LIN Y. Real-time seismic structural response prediction system based on support vector machine[J]. Earthquakes and Structures, 2020, 18(2): 163-170.
[15] ZHANG X Y, CHEN J, WU Y, et al. Predicting the maximum seismic response of the soil-pile-superstructure system using random forests[J]. Journal of Earthquake Engineering, 2022, 26(15): 8120-8141.
[16] CHOU J S, NGO N T, PHAM A D. Shear strength prediction in reinforced concrete deep beams using nature-inspired metaheuristic support vector regression[J]. Journal of Computing in Civil Engineering, 2016, 30(1): 04015002.
[17] MORFIDIS K, KOSTINAKIS K. Rapid prediction of seismic incident angle??s influence on the damage level of RC buildings using artificial neural networks[J]. Applied Sciences, 2022, 12(3): 1055.
[18] ZHANG R Y, CHEN Z, CHEN S, et al. Deep long short-term memory networks for nonlinear structural seismic response prediction[J]. Computers & Structures, 2019, 220: 55-68.
[19] OH B K, PARK Y, PARK H S. Seismic response prediction method for building structures using convolutional neural network[J]. Structural Control and Health Monitoring, 2020, 27(5): e2519.
[20] TORKY A A, OHNO S. Deep learning techniques for predicting nonlinear multi-component seismic responses of structural buildings[J]. Computers & Structures, 2021, 252: 106570.
[21] ZAMANI A A, ETEDALI S. Seismic response prediction of open-and closed-loop structural control systems using a multi-state-dependent parameter estimation approach[J]. International Journal of Computational Methods, 2022, 19(5): 2250006.
[22] PEREZ-RAMIREZ C A, AMEZQUITA-SANCHEZ J P, VALTIERRA-RODRIGUEZ M, et al. Recurrent neural network model with Bayesian training and mutual information for response prediction of large buildings[J]. Engineering Structures, 2019, 178: 603-615.
[23] LI T, PAN Y X, TONG K T, et al. A multi-scale attention neural network for sensor location selection and nonlinear structural seismic response prediction[J]. Computers & Structures, 2021, 248: 106507.
[24] XU Z K, CHEN J, SHEN J X, et al. Recursive long short-term memory network for predicting nonlinear structural seismic response[J]. Engineering Structures, 2022, 250: 113406.
[25] 许泽坤, 陈隽. 非线性结构地震响应的神经网络算法[J]. 工程力学, 2021, 38(9): 133-145.
XU Z K, CHEN J. Neural network algorithm for nonlinear structural seismic response[J]. Engineering Mechanics, 2021, 38(9): 133-145.
[26] MENG C Z, SEO S, CAO D F, et al. When physics meets machine learning: a survey of physics-informed machine learning[J]. arXiv:2203.16797, 2022.
[27] HAO Z K, LIU S M, ZHANG Y C, et al. Physics-informed machine learning: a survey on problems, methods and applications[J]. arXiv:2211.08064, 2022.
[28] KARNIADAKIS G E, KEVREKIDIS I G, LU L, et al. Physics-informed machine learning[J]. Nature Reviews Physics, 2021, 3(6): 422-440.
[29] XUE H, SALIM F D. PromptCast: a new prompt-based learning paradigm for time series forecasting[J]. IEEE Transactions on Knowledge and Data Engineering, 2024, 36(11): 6851-6864.
[30] SUN C X, LI H Y, LI Y L, et al. TEST: text prototype aligned embedding to activate LLM??s ability for time series[J]. arXiv:2308.08241, 2023.
[31] ZHOU T, NIU P, SUN L, et al. One fits all: power general time series analysis by pretrained LM[C]//Advances in Neural Information Processing Systems 36, 2023: 43322-43355.
[32] CHANG C, PENG W C, CHEN T F. LLM4TS: two-stage fine-tuning for time-series forecasting with pre-trained LLMs[J] arXiv:2308.08469, 2023.
[33] LESTER B, AL-RFOU R, CONSTANT N, et al. The power of scale for parameter-efficient prompt tuning[J]. arXiv:2104.08691, 2021.
[34] LI X L, LIANG P. Prefix-tuning: optimizing continuous prompts for generation[J]. arXiv:2101.00190, 2021.
[35] LI L H, ZHANG P C, ZHANG H T, et al. Grounded language-image pre-training[C]//Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2022: 10955-10965.
[36] ZHANG H, ZHANG P, HU X, et al. GLIPv2: unifying localization and vision-language understanding[C]//Advances in Neural Information Processing Systems 35, 2022: 36067-36080.
[37] RADFORD A, KIM J W, HALLACY C, et al. Learning transferable visual models from natural language supervision[C]//Proceedings of the 38th International Conference on Machine Learning, 2021: 8748-8763.
[38] WANG Z F, ZHANG Z Z, LEE C Y, et al. Learning to prompt for continual learning[C]//Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2022: 139-149.
[39] WANG Z F, ZHANG Z Z, EBRAHIMI S, et al. DualPrompt: complementary prompting for rehearsal-free continual learning[C]//Proceedings of the 17th European Conference on Computer Vision. Cham: Springer, 2022: 631-648.
[40] RADFORD A, WU J, CHILD R, et al. Language models are unsupervised multitask learners[J]. OpenAI Blog, 2019, 1(8): 9.
[41] ZHANG R Y, LIU Y, SUN H. Physics-guided convolutional neural network (PhyCNN) for data-driven seismic response modeling[J]. Engineering Structures, 2020, 215: 110704.
[42] ZHANG R Y, LIU Y, SUN H. Physics-informed multi-LSTM networks for metamodeling of nonlinear structures[J]. Computer Methods in Applied Mechanics and Engineering, 2020, 369: 113226.
[43] ZENG A L, CHEN M X, ZHANG L, et al. Are transformers effective for time series forecasting?[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2023, 37(9): 11121-11128.
[44] ZHOU T, MA Z, WEN Q, et al. FILM: frequency improved Legendre memory model for long-term time series forecasting[C]//Advances in Neural Information Processing Systems, 2022: 12677-12690.
[45] YI K, ZHANG Q, FAN W, et al. Frequency-domain MLPs are more effective learners in time series forecasting[C]//Advances in Neural Information Processing Systems 36, 2024.
[46] LIU Y, HU T G, ZHANG H R, et al. iTransformer: inverted transformers are effective for time series forecasting[J]. arXiv:2310.06625, 2023.
[47] NIE Y Q, NGUYEN N H, SINTHONG P, et al. A time series is worth 64 words: long-term forecasting with transformers[J]. arXiv:2211.14730, 2022.
[48] WOO G, LIU C H, SAHOO D, et al. ETSformer: exponential smoothing transformers for time-series forecasting[J]. arXiv:2202.01381, 2022.
[49] ZHOU T, MA Z Q, WEN Q S, et al. FEDformer: frequency enhanced decomposed transformer for long-term series forecasting[C]//Proceedings of the 39th International Conference on Machine Learning, 2022: 27268-27286.