To address the frequent issue of missing data in practical monitoring scenarios, high-accuracy imputation methods were required to ensure the reliability of monitoring and subsequent analysis. Therefore, a novel TPE Double Attention-Recurrent Neural Network (TDA-RNN) model was proposed, which integrated the Tree-structured Parzen Estimator (TPE) and attention mechanisms, aiming at imputing continuous missing wind speed data. By introducing the TPE optimization, the key hyperparameters of the model could be automatically optimized and selected. Meanwhile, the attention mechanisms could effectively capture feature information from the missing data. The model was validated using downburst wind speed data collected from field measurements at Texas Tech University. The results showed that, compared to traditional neural networks, the TDA-RNN model significantly improved imputation accuracy. When multiple segments of data were continuously missing, the root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) of the data interpolation error index reached 0.72, 0.64 and 5.65%, respectively, demonstrating the high accuracy and robustness of the TDA-RNN model for data imputation.
DENGYue. A machine learning based method for multivariate data anomaly detection in structural health monitoring[D]. Harbin: Harbin Institute of Technology, 2021.
[3]
MAYafei, ZHANGJianren, WANGLei, et al. Probabilistic prediction with Bayesian updating for strength degradation of RC bridge beams[J]. Structural Safety, 2013, 44: 102-109.
LIUXiaoyong, FANGHuajing, XIONGZhonggang, et al. Multi-step-ahead time series prediction based on LSSVR using UKF with sliding-windows[J]. Journal of Xinyang Normal University(Natural Science Edition), 2019, 32(2): 320-326.
LEIYing, LIULijun, ZHENGZhupeng. Review on the developments of some methods and techniques in structural health monitoring[J]. Journal of Xiamen University(Natural Science), 2021, 60(3): 630-640.
LIUDaohua, FENGChen, ZHAOYansong, et al. A fast MRI reconstruction method based on multi-scale feature fusion network[J]. Journal of Xinyang Normal University(Natural Science Edition), 2023, 36(1): 127-131.
[10]
LIChenzhao, MAHADEVANS, LINGYou, et al. Dynamic Bayesian network for aircraft wing health monitoring digital twin[J]. AIAA Journal, 2017, 55(3): 930-941.
[11]
LIUYongqi, QINHui, ZHANGZhendong, et al. Probabilistic spatiotemporal wind speed forecasting based on a variational Bayesian deep learning model[J]. Applied Energy, 2020, 260: 114259.
[12]
WANGZhiwei, LIAndong, ZHANGWenming, et al. Long-term missing wind data recovery using free access databases and deep learning for bridge health monitoring[J]. Journal of Wind Engineering and Industrial Aerodynamics, 2022, 230: 105201.
[13]
ZHANGZeyu, LUOYaozhi. Restoring method for missing data of spatial structural stress monitoring based on correlation[J]. Mechanical Systems and Signal Processing, 2017, 91: 266-277.
LINQiushuang, ZHOUHuan, LIUJintao. On-line damage detection under unknown external loadings[J]. Journal of Xinyang Normal University(Natural Science Edition), 2015, 28(4): 612-615.
[16]
KULLAAJ. Sensor validation using minimum mean square error estimation[J]. Mechanical Systems and Signal Processing, 2010, 24(5): 1444-1457.
[17]
PASPARAKISG D, DOS SANTOSK R M, KOUGIOUMTZOGLOUI A, et al. Wind data extrapolation and stochastic field statistics estimation via compressive sampling and low rank matrix recovery methods[J]. Mechanical Systems and Signal Processing, 2022, 162: 107975.
[18]
BAOYuequan, SHIZuoqiang, WANGXiaoyu, et al. Compressive sensing of wireless sensors based on group sparse optimization for structural health monitoring[J]. Structural Health Monitoring, 2017, 17(4): 823-836.
[19]
WANHuaping, DONGGuansen, LUOYaozhi. Compressive sensing of wind speed data of large-scale spatial structures with dedicated dictionary using time-shift strategy[J]. Mechanical Systems and Signal Processing, 2021, 157: 107685.
[20]
LINQiushuang, LIChunxiang. Kriging-based sequence interpolation and probability distribution correction for Gaussian wind field data reconstruction[J]. Journal of Wind Engineering and Industrial Aerodynamics, 2020, 205: 104340.
[21]
ZAPATA-SIERRAA J, CAMA-PINTOA, MONTOYAF G, et al. Wind missing data arrangement using wavelet based techniques for getting maximum likelihood[J]. Energy Conversion and Management, 2019, 185: 552-561.
[22]
MARTINEZ-LUENGOM, SHAFIEEM, KOLIOSA. Data management for structural integrity assessment of offshore wind turbine support structures: Data cleansing and missing data imputation[J]. Ocean Engineering, 2019, 173: 867-883.
[23]
ORWIGK D, SCHROEDERJ L. Near-surface wind characteristics of extreme thunderstorm outflows[J]. Journal of Wind Engineering and Industrial Aerodynamics, 2007, 95(7): 565-584.
[24]
QINYao, SONGDongjin, CHENHaifeng, et al. A dual-stage attention-based recurrent neural network for time series prediction[EB/OL]. (2017-08-14) [2024-12-01].
[25]
BAHDANAUD, CHOK, BENGIOY. Neural machine translation by jointly learning to align and translate[EB/OL]. (2016-05-19) [2024-12-01].