| 苏哲,刘宗显,余红玲,佟大威,余佳,王晓玲.基于虚拟传感器的坝区多输出自由场地震时程长序列预测模型研究[J].水利学报,2024,55(8):966-976,989 |
| 基于虚拟传感器的坝区多输出自由场地震时程长序列预测模型研究 |
| Research on multi-output seismic time-history long-term sequences prediction model for free field of dam based on virtual sensors |
| 投稿时间：2024-01-06 |
| DOI：10.13243/j.cnki.slxb.20240014 |
| 中文关键词: 自由场地震 虚拟传感器 多输出长时序预测 TFA-Seq2Seq 多任务学习 |
| 英文关键词: free field seismic virtual sensors multi-output long-term sequences prediction TFA-Seq2Seq multi-task learning |
| 基金项目:天津市自然科学基金项目(22JCQNJC01150) |
|
| 摘要点击次数: 1148 |
| 全文下载次数: 1370 |
| 中文摘要: |
| 坝区自由场地震时程的多维长时序预测对于震害快速分析具有重要意义。虚拟传感器是地震物理传感器的补充感知手段，可实现地震时程的预测，然而现有虚拟传感器难以对多个信号做长时序预测，导致大坝震害分析较为滞后。针对上述问题，提出基于TFA-Seq2Seq虚拟传感器的坝区多输出自由场地震时程长序列预测模型。其中，基于多任务学习将Seq2Seq的虚拟传感器改进为“Encoder-3 Decoder”结构，以建立多个坝体物理传感器信号与自由场三个方向长时序地震时程的映射关系，并添加注意力机制捕获多个输入信号的时序依赖关系，以解决同步多输出预测问题及提升预测精度。进一步，引入可逆的时频变换层和其逆变换层改进编码器和解码器，以缩短地震信号的时域长度，提取频域特征，并提出对应的随机强制学习的模型训练策略，从而克服了现有虚拟传感器难以对长序列进行有效预测的缺陷。案例分析表明，该方法实现了坝区自由场三个方向地震信号的超前10s虚拟感知，且相较于未添加注意力机制和单输出的模型，预测精度分别提高了6.88%和3.32%，研究为震时地震信息的超前感知提供了新思路和新途径。 |
| 英文摘要: |
| The multidimensional long-term prediction of seismic time-history in dam areas holds significant importance for rapid damage analysis.Virtual sensors，as complementary sensing mechanisms to seismic physical sensors，facilitate seismic time-history predictions.However，existing virtual sensors face challenges in effectively predicting long-term sequences for multiple signals，leading to delays in analyzing dam seismic damage.Addressing the aforementioned issue，a multi-output seismic time-history long-term sequences prediction model based on TFA-Seq2Seq virtual sensors is proposed.This model enhances the Seq2Seq virtual sensors using multi-task learning，restructuring them into an “Encoder-3 Decoder” architecture.This structure establishes the mapping relationship between multiple dam physical sensor signals and long-term seismic time-history in three free-field directions.Additionally，an attention mechanism is integrated to capture temporal dependencies among multiple input signals，resolving synchronous multi-output prediction issues and enhancing prediction accuracy.Furthermore，Time-Frequency transform（TF）layers and their inverse transformation layers are introduced to improve the Encoder and Decoder，shortening the temporal length of seismic signals and extracting frequency domain features.Correspondingly，a model training strategy involving stochastic forced learning is proposed to overcome the limitations of existing virtual sensors in effectively predicting long sequences.Case studies demonstrate that the proposed method achieves a virtual sense of 10 seconds ahead for seismic signals in three directions within dam free field.Compared to models without attention mechanisms and single outputs，the proposed method exhibits an enhanced prediction accuracy of 6.88% and 3.32%，respectively.This research presents novel insights and approaches for advancing the anticipatory sense of seismic information during seismic events. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |