Real-time prediction of mechanical behaviors of underwater shield tunnel structure using machine learning method based on structural health monitoring data

Predicting the mechanical behaviors of structure and perceiving the anomalies in advance are essential to ensuring the safe operation of infrastructures in the long run.In addition to the incomplete consideration of influencing factors,the prediction time scale of existing studies is rough.Therefore,this study focuses on the development of a real-time prediction model by coupling the spatio-temporal correlation with external load through autoencoder network(ATENet)based on structural health monitoring(SHM)data.An autoencoder mechanism is performed to acquire the high-level representation of raw monitoring data at different spatial positions,and the recurrent neural network is applied to understanding the temporal correlation from the time series.Then,the obtained temporal-spatial information is coupled with dynamic loads through a fully connected layer to predict structural performance in next 12 h.As a case study,the proposed model is formulated on the SHM data collected from a representative underwater shield tunnel.The robustness study is carried out to verify the reliability and the prediction capability of the proposed model.Finally,the ATENet model is compared with some typical models,and the results indicate that it has the best performance.ATENet model is of great value to predict the realtime evolution trend of tunnel structure.

Real-Time Prediction Algorithm for Intelligent Edge Networks with Federated Learning-Based Modeling

Intelligent healthcare networks represent a significant component in digital applications,where the requirements hold within quality-of-service(QoS)reliability and safeguarding privacy.This paper addresses these requirements through the integration of enabler paradigms,including federated learning(FL),cloud/edge computing,softwaredefined/virtualized networking infrastructure,and converged prediction algorithms.The study focuses on achieving reliability and efficiency in real-time prediction models,which depend on the interaction flows and network topology.In response to these challenges,we introduce a modified version of federated logistic regression(FLR)that takes into account convergence latencies and the accuracy of the final FL model within healthcare networks.To establish the FLR framework for mission-critical healthcare applications,we provide a comprehensive workflow in this paper,introducing framework setup,iterative round communications,and model evaluation/deployment.Our optimization process delves into the formulation of loss functions and gradients within the domain of federated optimization,which concludes with the generation of service experience batches for model deployment.To assess the practicality of our approach,we conducted experiments using a hypertension prediction model with data sourced from the 2019 annual dataset(Version 2.0.1)of the Korea Medical Panel Survey.Performance metrics,including end-to-end execution delays,model drop/delivery ratios,and final model accuracies,are captured and compared between the proposed FLR framework and other baseline schemes.Our study offers an FLR framework setup for the enhancement of real-time prediction modeling within intelligent healthcare networks,addressing the critical demands of QoS reliability and privacy preservation.

Superiority of a Convolutional Neural Network Model over Dynamical Models in Predicting Central Pacific ENSO

The application of deep learning is fast developing in climate prediction,in which El Ni?o–Southern Oscillation(ENSO),as the most dominant disaster-causing climate event,is a key target.Previous studies have shown that deep learning methods possess a certain level of superiority in predicting ENSO indices.The present study develops a deep learning model for predicting the spatial pattern of sea surface temperature anomalies(SSTAs)in the equatorial Pacific by training a convolutional neural network(CNN)model with historical simulations from CMIP6 models.Compared with dynamical models,the CNN model has higher skill in predicting the SSTAs in the equatorial western-central Pacific,but not in the eastern Pacific.The CNN model can successfully capture the small-scale precursors in the initial SSTAs for the development of central Pacific ENSO to distinguish the spatial mode up to a lead time of seven months.A fusion model combining the predictions of the CNN model and the dynamical models achieves higher skill than each of them for both central and eastern Pacific ENSO.

Prediction of ENSO using multivariable deep learning

本文基于残差神经网络和观测数据构建了一套深度学习多因子预报测模型,以改进厄尔尼诺-南方涛动(ENSO)的预报.该模型基于最大信息系数进行因子时空特征提取,并根据泰勒图的评估标准可自动确定关键预报因子进行预报.该模型在超前8个月以内的预报性能要优于当前传统的业务预报模式.2011–2018年间,该模型的预报性能优于多模式集成预报的结果.在超前6个月预报时效上,模型预报相关性可达0.82,标准化后的均方根误差仅为0.58°C,多模式集成预报的相关性和标准化后的均方根误差分别为0.70和0.73°C.该模型春季预报障碍问题有所缓解,并且自动选取的关键预报因子可用于解释热带和副热带热动力过程对于ENSO变化的影响.

A spatiotemporal 3D convolutional neural network model for ENSO predictions: A test case for the 2020/21 La Niña conditions

2020–22年间热带太平洋经历了持续性多年的拉尼娜事件,多数耦合模式都难以准确预测其演变过程,这为厄尔尼诺-南方涛动(ENSO)的实时预测带来了很大的挑战.同时,目前学术界对此次持续性双拉尼娜事件的发展仍缺乏合理的物理解释,其所涉及的物理过程和机制有待于进一步分析.本研究利用再分析数据产品分析了热带东南太平洋东南风异常及其引起的次表层海温异常在此次热带太平洋海表温度(SST)异常演变中的作用,并构建了一个时空分离(Time-Space)的三维(3D)卷积神经网络模型(TS-3DCNN)对此次双拉尼娜事件进行实时预测和过程分析.通过将TS-3DCNN与中国科学院海洋研究所(IOCAS)中等复杂程度海气耦合模式(IOCAS ICM)的预测结果对比,表明TS-3DCNN模型对2020–22年双重拉尼娜现象的预测能力与IOCAS ICM相当,二者均能够从2021年初的初始场开始较好地预测2021年末El Niño3.4区SST的演变.此外,基于TS-3DCNN和IOCAS ICM的敏感性试验也验证了赤道外风场异常和次表层海温异常在2021年末赤道中东太平洋海表二次变冷过程中的关键作用.未来将神经网络与动力模式模式间的有效结合,进一步发展神经网络与物理过程相结合的混合建模是进一步提高ENSO事件预测能力的有效途径.

A real-time prediction method for tunnel boring machine cutter-head torque using bidirectional long short-term memory networks optimized by multi-algorithm

Based on data from the Jilin Water Diversion Tunnels from the Songhua River(China),an improved and real-time prediction method optimized by multi-algorithm for tunnel boring machine(TBM)cutter-head torque is presented.Firstly,a function excluding invalid and abnormal data is established to distinguish TBM operating state,and a feature selection method based on the SelectKBest algorithm is proposed.Accordingly,ten features that are most closely related to the cutter-head torque are selected as input variables,which,in descending order of influence,include the sum of motor torque,cutter-head power,sum of motor power,sum of motor current,advance rate,cutter-head pressure,total thrust force,penetration rate,cutter-head rotational velocity,and field penetration index.Secondly,a real-time cutterhead torque prediction model’s structure is developed,based on the bidirectional long short-term memory(BLSTM)network integrating the dropout algorithm to prevent overfitting.Then,an algorithm to optimize hyperparameters of model based on Bayesian and cross-validation is proposed.Early stopping and checkpoint algorithms are integrated to optimize the training process.Finally,a BLSTMbased real-time cutter-head torque prediction model is developed,which fully utilizes the previous time-series tunneling information.The mean absolute percentage error(MAPE)of the model in the verification section is 7.3%,implying that the presented model is suitable for real-time cutter-head torque prediction.Furthermore,an incremental learning method based on the above base model is introduced to improve the adaptability of the model during the TBM tunneling.Comparison of the prediction performance between the base and incremental learning models in the same tunneling section shows that:(1)the MAPE of the predicted results of the BLSTM-based real-time cutter-head torque prediction model remains below 10%,and both the coefficient of determination(R^(2))and correlation coefficient(r)between measured and predicted values exceed 0.95;and(2)the incremental learning method is suitable for realtime cutter-head torque prediction and can effectively improve the prediction accuracy and generalization capacity of the model during the excavation process.

Real-time 3-D space numerical shake prediction for earthquake early warning

In earthquake early warning systems, real-time shake prediction through wave propagation simulation is a promising approach. Compared with traditional methods, it does not suffer from the inaccurate estimation of source parameters. For computation efficiency, wave direction is assumed to propagate on the 2-D surface of the earth in these methods. In fact, since the seismic wave propagates in the 3-D sphere of the earth, the 2-D space modeling of wave direction results in inaccurate wave estimation. In this paper, we propose a 3-D space numerical shake pre- diction method, which simulates the wave propagation in 3-D space using radiative transfer theory, and incorporate data assimilation technique to estimate the distribution of wave energy. 2011 Tohoku earthquake is studied as an example to show the validity of the proposed model. 2-D space model and 3-D space model are compared in this article, and the prediction results show that numerical shake prediction based on 3-D space model can estimate the real-time ground motion precisely, and overprediction is alleviated when using 3-D space model.

Efficient Harmonic Analysis Technique for Prediction of IGS Real-Time Satellite Clock Corrections

Real-time satellite orbit and clock corrections obtained from the broadcast ephemerides can be improved using IGS real-time service (RTS) products. Recent research showed that applying such corrections for broadcast ephemerides can significantly improve the RMS of the estimated coordinates. However, unintentional streaming interruption may happen for many reasons such as software or hardware failure. Streaming interruption, if happened, will cause sudden degradation of the obtained solution if only the broadcast ephemerides are used. A better solution can be obtained in real-time if the predicted part of the ultra-rapid products is used. In this paper, Harmonic analysis technique is used to predict the IGS RTS corrections using historical broadcasted data. It is shown that using the predicted clock corrections improves the RMS of the estimated coordinates by about 72%, 58%, and 72% in latitude, longitude, and height directions, respectively and reduces the 2D and 3D errors by about 80% compared with the predicted part of the IGS ultra-rapid clock corrections.

Real-time numerical shake prediction and updating for earthquake early warning

Ground motion prediction is important for earthquake early warning systems, because the region＇s peak ground motion indicates the potential disaster. In order to predict the peak ground motion quickly and pre- cisely with limited station wave records, we propose a real- time numerical shake prediction and updating method. Our method first predicts the ground motion based on the ground motion prediction equation after P waves detection of several stations, denoted as the initial prediction. In order to correct the prediction error of the initial prediction, an updating scheme based on real-time simulation of wave propagation is designed. Data assimilation technique is incorporated to predict the distribution of seismic wave energy precisely. Radiative transfer theory and Monte Carlo simulation are used for modeling wave propagation in 2-D space, and the peak ground motion is calculated as quickly as possible. Our method has potential to predict shakemap, making the potential disaster be predicted before the real disaster happens. 2008 Ms8.0 Wenchuan earthquake is studied as an example to show the validity of the proposed method.

ENSOMIM:一种新型ENSO时空预测模型

为了提高厄尔尼诺南方涛动(El Ni?o-southern oscillation,ENSO)预测的准确性,解决卷积核难以捕获ENSO的长距离前兆的问题,将ENSO预测视为一个时空序列预测问题,并提出一种基于注意力机制和循环神经网络的ENSO非稳态时空预测深度学习模型,称为ENSOMIM。该模型通过提出的新型注意力机制BGAM来局部和全局交互地学习空间特征,并使用高阶非线性时空网络对长期的时间序列特征进行编码。由于ENSO观测数据集样本数量少,为了更充分地训练模型,采用迁移学习的方法,使用历史模式模拟数据进行预训练再利用观测数据校正模型。实验结果表明,ENSOMIM更适合于大区域和长期的预测。在1984-2014年验证期间,ENSOMIM的Ni?o3.4指数的全季节相关性技巧比经典的卷积神经网络提高16%,均方误差降低17%,它可以为长达18个月的提前期提供有效预测,并且在23个月的提前期内相关技巧达到0.45。因此,ENSOMIM可以作为预测ENSO事件的有力工具。

Seasonal Prediction Skill and Biases in GloSea5 Relating to the East Asia Winter Monsoon

The simulation and prediction of the climatology and interannual variability of the East Asia winter monsoon(EAWM),as well as the associated atmospheric circulation,was investigated using the hindcast data from Global Seasonal Forecast System version 5(GloSea5),with a focus on the evolution of model bias among different forecast lead times.While GloSea5 reproduces the climatological means of large-scale circulation systems related to the EAWM well,systematic biases exist,including a cold bias for most of China’s mainland,especially for North and Northeast China.GloSea5 shows robust skill in predicting the EAWM intensity index two months ahead,which can be attributed to the performance in representing the leading modes of surface air temperature and associated background circulation.GloSea5 realistically reproduces the synergistic effect of El Niño–Southern Oscillation(ENSO)and the Arctic Oscillation(AO)on the EAWM,especially for the western North Pacific anticyclone(WNPAC).Compared with the North Pacific and North America,the representation of circulation anomalies over Eurasia is poor,especially for sea level pressure(SLP),which limits the prediction skill for surface air temperature over East Asia.The representation of SLP anomalies might be associated with the model performance in simulating the interaction between atmospheric circulations and underlying surface conditions.

Testing a Four-Dimensional Variational Data Assimilation Method Using an Improved Intermediate Coupled Model for ENSO Analysis and Prediction

A four-dimensional variational （4D-Var） data assimilation method is implemented in an improved intermediate coupled model （ICM） of the tropical Pacific. A twin experiment is designed to evaluate the impact of the 4D-Var data assimilation algorithm on ENSO analysis and prediction based on the ICM. The model error is assumed to arise only from the parameter uncertainty. The ＂observation＂ of the SST anomaly, which is sampled from a ＂truth＂ model simulation that takes default parameter values and has Gaussian noise added, is directly assimilated into the assimilation model with its parameters set erroneously. Results show that 4D-Var effectively reduces the error of ENSO analysis and therefore improves the prediction skill of ENSO events compared with the non-assimilation case. These results provide a promising way for the ICM to achieve better real-time ENSO prediction.

Impact of observational MJO forcing on ENSO predictability in the Zebiak-Cane model: PartⅠ.Effect on the maximum prediction error

With the observational wind data and the Zebiak-Cane model, the impact of Madden-Iulian Oscillation （MJO） as external forcing on El Nino-Southern Oscillation （ENSO） predictability is studied. The observational data are analyzed with Continuous Wavelet Transform （CWT） and then used to extract MJO signals, which are added into the model to get a new model. After the Conditional Nonlinear Optimal Perturbation （CNOP） method has been used, the initial errors which can evolve into maximum prediction error, model errors and their join errors are gained and then the Nifio 3 indices and spatial structures of three kinds of errors are investigated. The results mainly show that the observational MJO has little impact on the maximum prediction error of ENSO events and the initial error affects much greater than model error caused by MJO forcing. These demonstrate that the initial error might be the main error source that produces uncertainty in ENSO prediction, which could provide a theoretical foundation for the adaptive data assimilation of the ENSO forecast and contribute to the ENSO target observation.

Diagnosing SST Error Growth during ENSO Developing Phase in the BCC_CSM1.1(m) Prediction System

In this study, the predictability of the El Nino-South Oscillation(ENSO) in an operational prediction model from the perspective of initial errors is diagnosed using the seasonal hindcasts of the Beijing Climate Center System Model,BCC;SM1.1(m). Forecast skills during the different ENSO phases are analyzed and it is shown that the ENSO forecasts appear to be more challenging during the developing phase, compared to the decay phase. During ENSO development, the SST prediction errors are significantly negative and cover a large area in the central and eastern tropical Pacific, thus limiting the model skill in predicting the intensity of El Nino. The large-scale SST errors, at their early stage, are generated gradually in terms of negative anomalies in the subsurface ocean temperature over the central-western equatorial Pacific,featuring an error evolutionary process similar to that of El Nino decay and the transition to the La Nina growth phase.Meanwhile, for short lead-time ENSO predictions, the initial wind errors begin to play an increasing role, particularly in linking with the subsurface heat content errors in the central-western Pacific. By comparing the multiple samples of initial fields in the model, it is clearly found that poor SST predictions of the Nino-3.4 region are largely due to contributions of the initial errors in certain specific locations in the tropical Pacific. This demonstrates that those sensitive areas for initial fields in ENSO prediction are fairly consistent in both previous ideal experiments and our operational predictions,indicating the need for targeted observations to further improve operational forecasts of ENSO.

ENSO组合模态可预测性的季节-年代际变化

厄尔尼诺-南方涛动组合模态(C-mode)是厄尔尼诺衰减期西北太平洋上空异常反气旋发展的主要驱动力,通常可以形成水汽输送通道,导致我国南部地区降水增多并发生严重的洪涝灾害,而在拉尼娜期间则情况相反。利用1981—2020年美国国家环境预报中心和美国国家大气研究中心再分析数据集的表面风场资料分析了C-mode可预测性的季节-年代际变化。结果表明:C-mode在2000年以后的可预测性明显下降,主要原因是其变率减小、强度减弱、信噪比降低等。在季节尺度上,C-mode存在“秋季预报障碍”,这与信号的季节循环密切关联,当C-mode在秋季进入衰退期时,信号强度最弱、变率最小,因此其秋季的可预测性降低。

Application of the Analogue-Based Correction of Errors Method in ENSO Prediction

In this study, a method of analogue-based correction of errors(ACE) was introduced to improve El Ni?o-Southern Oscillation(ENSO) prediction produced by climate models. The ACE method is based on the hypothesis that the flow-dependent model prediction errors are to some degree similar under analogous historical climate states, and so the historical errors can be used to effectively reduce such flow-dependent errors. With this method, the unknown errors in current ENSO predictions can be empirically estimated by using the known prediction errors which are diagnosed by the same model based on historical analogue states. The authors first propose the basic idea for applying the ACE method to ENSO prediction and then establish an analogue-dynamical ENSO prediction system based on an operational climate prediction model. The authors present some experimental results which clearly show the possibility of correcting the flow-dependent errors in ENSO prediction, and thus the potential of applying the ACE method to operational ENSO prediction based on climate models.

Analysis of Precipitation Anomaly and a Failed Prediction During the Dragon-boat Rain Period in 2022

This study investigates the possible causes for the precipitation of Guangdong during dragon-boat rain period(DBRP) in 2022 that is remarkably more than the climate state and reviews the successes and failures of the prediction in2022. Features of atmospheric circulation and sea surface temperature(SST) are analyzed based on several observational datasets for nearly 60 years from meteorological stations and the NCEP/NCAR Global Reanalysis Data. Results show that fluctuation of the 200-h Pa westerly wind as well as the westerly jet is strengthened due to the propagation of wave energy, leading to strong updraft over southern China. Activities of a subtropical high and a shear line provide favorable conditions for the transport of moisture to Guangdong. With the support of powerful southwest winds, extreme precipitation is induced. ENSO is a good indicator of atmospheric circulation at mid-and high-levels during the DBRP in2022 but it performs badly at low levels. During recent years, the influence of ENSO on precipitation during the DBRP has decreased obviously. The SSTA of tropical southeast Atlantic(SEA) in spring may become the key indicator. During the years with warm SEA, wave trains propagate from northwest to southeast over Eurasia with energy enhancing the westerly jet, conducive to updraft over southern China and the occurrence of heavy precipitation. Meanwhile, the Rossby wave is triggered over Maritime Continent by heat sources of southern Atlantic-western Indian Ocean through the Gill response. Thus, strong transport of moisture and heavy rainfall occur.

A Hybrid Neural Network Model for ENSO Prediction in Combination with Principal Oscillation Pattern Analyses

El Niño-Southern Oscillation(ENSO)can be currently predicted reasonably well six months and longer,but large biases and uncertainties remain in its real-time prediction.Various approaches have been taken to improve understanding of ENSO processes,and different models for ENSO predictions have been developed,including linear statistical models based on principal oscillation pattern(POP)analyses,convolutional neural networks(CNNs),and so on.Here,we develop a novel hybrid model,named as POP-Net,by combining the POP analysis procedure with CNN-long short-term memory(LSTM)algorithm to predict the Niño-3.4 sea surface temperature(SST)index.ENSO predictions are compared with each other from the corresponding three models:POP model,CNN-LSTM model,and POP-Net,respectively.The POP-based pre-processing acts to enhance ENSO-related signals of interest while filtering unrelated noise.Consequently,an improved prediction is achieved in the POP-Net relative to others.The POP-Net shows a high-correlation skill for 17-month lead time prediction(correlation coefficients exceeding 0.5)during the 1994-2020 validation period.The POP-Net also alleviates the spring predictability barrier(SPB).It is concluded that value-added artificial neural networks for improved ENSO predictions are possible by including the process-oriented analyses to enhance signal representations.

The“Spring Predictability Barrier” Phenomenon of ENSO Predictions Generated with the FGOALS-g Model

Using the sea surface temperature (SST) predicted for the equatorial Pacific Ocean by the Flexible Global Ocean-Atmosphere-Land System Model-gamil (FGOALS-g),an analysis of the prediction errors was performed for the seasonally dependent predictability of SST anomalies both for neutral years and for the growth/decay phase of El Ni o/La Ni a events.The study results indicated that for the SST predictions relating to the growth phase and the decay phase of El Ni o events,the prediction errors have a seasonally dependent evolution.The largest increase in errors occurred in the spring season,which indicates that a prominent spring predictability barrier (SPB) occurs during an El Ni o-Southern Oscillation (ENSO) warming episode.Furthermore,the SPB associated with the growth-phase prediction is more prominent than that associated with the decay-phase prediction.However,for the neutral years and for the growth and decay phases of La Ni a events,the SPB phenomenon was less prominent.These results indicate that the SPB phenomenon depends extensively on the ENSO events themselves.In particular,the SPB depends on the phases of the ENSO events.These results may provide useful knowledge for improving ENSO forecasting.

Preliminary Studies on Predicting the Tropical Indian Ocean Sea Surface Temperature through Combined Statistical Methods and Dynamic ENSO Prediction

The sea surface temperature(SST) in the Indian Ocean affects the regional climate over the Asian continent mostly through a modulation of the monsoon system.It is still difficult to provide an a priori indication of the seasonal variability over the Indian Ocean.It is widely recognized that the warm and cold events of SST over the tropical Indian Ocean are strongly linked to those of the equatorial eastern Pacific.In this study,a statistical prediction model has been developed to predict the monthly SST over the tropical Indian Ocean.This model is a linear regression model based on the lag relationship between the SST over the tropical Indian Ocean and the Ni o3.4(5°S-5°N,170°W-120°W) SST Index.The predictor(i.e.,Ni o3.4 SST Index) has been operationally predicted by a large size ensemble El Ni o and the Southern Oscillation(ENSO) forecast system with coupled data assimilation(Leefs_CDA),which achieves a high predictive skill of up to a 24-month lead time for the equatorial eastern Pacific SST.As a result,the prediction skill of the present statistical model over the tropical Indian Ocean is better than that of persistence prediction for January 1982 through December 2009.