The objective of this article is to develop methods for extracting trends from long-term structural health monitoring data and try to set an early warning threshold level based on the results of analyses. The long-term monitoring data in this study is the continuous monitoring of the dam static deformation. Two different approaches were applied to extract features of the long-term structural health monitoring data of the static deformation of the Fei-Tsui Arch Dam (Taiwan). The methods include the singular spectrum analysis with auto regressive model (SSA-AR) and the nonlinear principal component analysis (NPCA) using auto-associative neural network method (AANN). The singular spectrum analysis is a novel nonparametric technique based on principles of multi-variance statistics. An AR model is optimized for each of the principal components obtained from SSA, and the multi step predicted values are recombined to make the time series. Different from SSA method the NPCA-AANN method is also used to extract the underlying features of static deformation of the dam. By using these two different methods, the residual deformation between the estimated and the recorded data was generated, through statistical analysis, the threshold level of the dam static deformation can be determined. Discussion on the two proposed methods to the static deformation monitoring data of Fei-Tsui Arch Dam (Taiwan) is discussed.

On-site earthquake early warning techniques, which issue alerts based on seismic waves measured at a single station, are promising, and have performed quite successfully during some damaging earthquakes. Conventionally, most existing techniques extract several P-wave features from the first few seconds of seismic waves after the trigger to predict the intensity or destructiveness of an incoming earthquake. This type of technique neglects the behavior of temporal varying features within P waves. In other words, the characteristics of data sequences are not considered. In this study, a long short-term memory (LSTM) neural network, which is capable of learning order dependence in seismic waves, is employed to predict the peak ground acceleration (PGA) of the coming earthquake. A dense LSTM architecture is proposed and a large data set of earthquakes is used to train the LSTM model. The general performance of the LSTM model indicated that the predicted PGA values are quite promising but are generally overestimated. However, the predicted PGA of the Chi-Chi earthquake data set, whose fault rupture is complex and long, using the proposed LSTM model is more accurate than the PGA predicted in a previous study using a support vector regression approach. In addition, an alternative alert criterion, which issues alerts when the predicted PGA exceeds the threshold in successive time windows, is presented, and the performance of the proposed LSTM model when different PGA thresholds are considered is also discussed.

The National Center for Research on Earthquake Engineering, Taiwan, has developed an Earthquake Early Warning System (NEEWS). The NEEWS predicts peak ground acceleration (PGA) using an on‐site approach, whereas the Central Weather Bureau (CWB), Taiwan, uses a regional approach. Earthquake alerts are issued at the NEEWS stations once PGA reaches a preassigned PGA threshold, regardless of the approach used. An earthquake with a magnitude of 6.2 and a focal depth of 10.0 km struck Hualien, in eastern Taiwan, on 6 February 2018. It resulted in 17 fatalities and 285 injuries, 4 collapsed buildings, and damage to more than 175 buildings. During the earthquake, the system performance of 28 NEEWS stations was documented. In this study, we compare and discuss the accuracy of the PGA predictions, lead times, and classification performance of both approaches.

Although convolutional neural networks (CNN) have been applied successfully to many fields, the onsite earthquake early warning by CNN remains unexplored. This study aims to predict the peak ground acceleration (PGA) of the incoming seismic waves using CNN, which is achieved by analyzing the first 3 s of P-wave data collected from a single site. Because the amplitude of P-wave data of large and small earthquakes can differ, the multi-scale input of P-wave data is proposed in this study in order to let the CNN observe the input data in different scales. Both the time and frequency domains of the P-wave data are combined into multi-domain input, and therefore the CNN can observe the data from different aspects. Since only the maximum absolute acceleration value of the time history of seismic waves is the target output of the CNN, the absolute value of the P-wave time history data is used instead of the raw value. The proposed arrangement of the input data shows its superiority to the one directly inputting the raw P-wave data into the CNN. Moreover, the predicted PGA accuracy using the proposed CNN approach is higher than the one using the support vector regression approach that employed the extracted P-wave features as its input. The proposed CNN approach also shows that the accuracy of the predicted PGA and the alert performances are acceptable based on data from two independent and damaging earthquakes.

The National Center for Research on Earthquake Engineering in Taiwan has developed an on‐site earthquake early warning system (NEEWS). The Meinong earthquake with a moment magnitude of 6.53 and a focal depth of 14.6 km occurred on 5 February 2016 in southern Taiwan. It caused 117 deaths, injured 551, caused the collapse of six buildings, and serious damage to 247 buildings. During the Meinong earthquake, the system performance of 16 NEEWS stations was recorded. Based on a preassigned peak ground acceleration (PGA) threshold to issue alarms at different stations, no false alarms or missed alarms were issued during the earthquake. About 4 s to 33 s of lead time were provided by the NEEWS depending on the epicenter distance. In addition, the directivity of the earthquake source characteristic and also possibly the site effects were observed in the diagram of the distribution of PGA difference between the predicted PGA and the measured PGA.