A method for non-parametric damage detection through the use of neural networks

A neural network-based approach is presented for the detection of changes in the characteristics of structure-unknown systems. The approach relies on the use of vibration measurements from a ‘healthy’ system to train a neural network for identification purposes. Subsequently, the trained network is fed comparable vibration measurements from the same structure under different episodes of response in order to monitor the health of the structure. The methodology is applied to actual data obtained from ambient vibration measurements on a steel building structure that was damaged under strong seismic motion during the Hyogo-Ken Nanbu Earthquake of 17 January 1995. The measurements were done before and after repairs to the damaged frame were made. A neural network is trained with data after the repairs, which represents ‘healthy’ condition of the building. The trained network, which is subsequently fed data before the repairs, successfully identified the difference between the damaged storey and the undamaged storey. Through this study, it is shown that the proposed approach has the potential of being a practical tool for a damage detection methodology applied to smart civil structures. © 1998 John Wiley & Sons, Ltd.     

地震应急信息自动分类方法研究

地震应急信息的高效处理为地震应急救援工作提供了重要支撑。本文根据地震应急信息分类的需求,构建了一种高效便捷的地震信息分类处理方法。以震前、震时、震后为时间主线,将地震应急信息分为震前基础背景信息、地震震情灾情信息及震后应急救援信息,并采用“关键词分类”的方法,在计算机语言的支持下,将多渠道汇集的应急信息进行自动分类,在一定程度上缩短了应急信息加工处理与服务的时间,能快速高效地为应急指挥提供信息服务。     

全卷积神经网络在建筑物震害遥感提取中的应用研究

为解决建筑物震害信息提取自动化程度不高的问题,本文将全卷积神经网络应用于建筑物震害遥感信息提取。以玉树地震后获取的玉树县城区0.2m分辨率航空影像作为建筑物震害信息提取试验数据源,将试验区地物划分为倒塌建筑物、未倒塌建筑物和背景3类。对427个500×500像素的子影像进行人工分类与标注,选取393个组成训练样本集,34个用于验证。利用训练样本集对全卷积神经网络进行训练,采用训练后的网络对验证样本进行建筑物震害信息提取及精度评价。研究结果表明:建筑物震害遥感信息提取总体分类精度为82.3%,全卷积神经网络方法能提高信息提取自动化程度,具有较好的建筑物震害信息提取能力。     

Generating multiple spectrum compatible accelerograms using stochastic neural networks

A new neural‐network‐based methodology for generating artificial earthquake spectrum compatible accelerograms from response spectra was proposed in 1997, in which, the learning capabilities of neural networks were used to develop the knowledge of the inverse mapping from the response spectra to earthquake accelerograms. Recently, this methodology has been further extended and enhanced. This paper presents a new stochastic neural network that is capable of generating multiple earthquake accelerograms from a single‐response spectrum. A new stochastic feature to the neural network has been combined with a new scheme for data compression using the replicator neural networks developed in the original method. A benefit of this extended methodology is gaining efficiency in compressing the earthquake accelerograms and extracting their characteristics. The proposed method produces a stochastic ensemble of earthquake accelerograms from any response spectra or design spectra. An example is presented that used 100 recorded accelerograms to train the neural network and several design spectra and response spectra to test this improved methodology. Copyright © 2001 John Wiley & Sons, Ltd.     

Seismic response estimation method for earthquake‐damaged RC buildings

This study proposes a procedure for identifying spectral response curves for earthquake‐damaged areas in developing countries without seismic records. An earthquake‐damaged reinforced concrete building located in Padang, Indonesia was selected to illustrate the identification of the maximum seismic response during the 2009 West Sumatra earthquake. This paper summarizes the damage incurred by the building; the majority of the damage was observed in the third story in the span direction. The damage was quantitatively evaluated using the damage index R according to the Japanese guidelines for post‐earthquake damage evaluation. The damage index was also applied to the proposed spectral response identification method. The seismic performance of the building was evaluated by a nonlinear static analysis. The analytical results reproduced a drift concentration in the third story. The R‐index decreased with an increase in the story drift, which provided an estimation of the maximum response of the building during the earthquake. The estimation was verified via an earthquake response analysis of the building using ground acceleration data, which were simulated based on acceleration records of engineering bedrock that considered site amplification. The maximum response estimated by the R‐index was consistent with the maximum response obtained from the earthquake response analysis. Therefore, the proposed method enables the construction of spectral response curves by integrating the identification results for the maximum responses in a number of earthquake‐damaged buildings despite a lack of seismic records. Copyright © 2016 The Authors. Earthquake Engineering & Structural Dynamics published by John Wiley & Sons Ltd.     

Optimum design of structures against earthquake by wavelet neural network and filter banks

Optimum design of structures for earthquake is achieved by simulated annealing. To reduce the computational work, a fast wavelet transform is used by means of which the number of points in the earthquake record is decreased. The record is decomposed into two parts. One part contains the low frequency of the record, and the other contains the high frequency of the record. The low‐frequency content is the effective part, since most of the energy of the record is contained in this part of the record. Thus, the low‐frequency part of the record is used for dynamic analysis. Then, using a wavelet neural network, the dynamic responses of the structures are approximated. By such approximation, the dynamic analysis of the structure becomes unnecessary in the process of optimization. The wavelet neural networks have been employed as a general approximation tool for the time history dynamic analysis. A number of structures are designed for optimal weight and the results are compared to those corresponding to the exact dynamic analysis. Copyright © 2004 John Wiley & Sons, Ltd.     

Unsupervised fuzzy neural network structural active pulse controller

Applying active control systems to civil engineering structures subjected to dynamic loading has received increasing interest. This study proposes an active pulse control model, termed unsupervised fuzzy neural network structural active pulse controller (UFN‐SAP controller), for controlling civil engineering structures under dynamic loading. The proposed controller combines an unsupervised neural network classification (UNC) model, an unsupervised fuzzy neural network (UFN) reasoning model, and an active pulse control strategy. The UFN‐SAP controller minimizes structural cumulative responses during earthquakes by applying active pulse control forces determined via the UFN model based on the clusters, classified through the UNC model, with their corresponding control forces. Herein, we assume that the effect of the pulses on structure is delayed until just before the next sampling time so that the control force can be calculated in time, and applied. The UFN‐SAP controller also averts the difficulty of obtaining system parameters for a real structure for the algorithm to allow active structural control. Illustrative examples reveal significant reductions in cumulative structural responses, proving the feasibility of applying the adaptive unsupervised neural network with the fuzzy classification approach to control civil engineering structures under dynamic loading. Copyright © 2001 John Wiley & Sons, Ltd.     

Neural network for earthquake selection in structural time history analysis

A neural network is employed to select earthquake waves in a time history approach for structural dynamics. The neural network is a preferable alternative to an expert system because knowledge can easily be renewed. It involves a back propagation model having three layers (one input, one hidden and one output layer) and is used to avoid inappropriate earthquake input prior to practical numerical computations. Knowledge to categorize the earthquake waves is acquired through network training with earthquake response spectra and structural responses. The trained network is tested by categorizing the responses of three types of unknown structures caused by 50 previously recorded earthquakes. Comparisons are made with analogous data from the traditional site dominant period method. Results demonstrate that, unlike the latter method, a neural network is generally more successful as the number of training patterns increases.     

Dynamic characteristics and seismic behavior of prefabricated steel stairs in a full‐scale five‐story building shake table test program

This paper investigates the dynamic characteristics and seismic behavior of prefabricated steel stairs in a full‐scale five‐story building shake table test program. The test building was subjected to a suite of earthquake input motions and low‐amplitude white noise base excitations first, while the building was isolated at its base, and subsequently while it was fixed to the shake table platen. This paper presents the modal characteristics of the stairs identified using the data recorded from white noise base excitation tests as well as the physical and measured responses of the stairs from the earthquake tests. The observed damage to the stairs is categorized into three distinct damage states and is correlated with the interstory drift demands of the building. These shake table tests highlight the seismic vulnerability of modern designed stair systems and in particular identifies as a key research need the importance of improving the deformability of flight‐to‐building connections. Copyright © 2015 John Wiley & Sons, Ltd.     

Parameter identification of torsionally coupled shear buildings from earthquake response records

This paper presents an efficient procedure to determine the natural frequencies, modal damping ratios and mode shapes for torsionally coupled shear buildings using earthquake response records. It is shown that the responses recorded at the top and first floor levels are sufficient to identify the dominant modal properties of a multistoried torsionally coupled shear building with uniform mass and constant eccentricity even when the input excitation is not known. The procedure applies eigenrealization algorithm to generate the state‐space model of the structure using the cross‐correlations among the measured responses. The dynamic characteristics of the structure are determined from the state‐space realization matrices. Since the mode shapes are obtained only at the instrumented floor (top and first floors) levels, a new mode shape interpolation technique has been proposed to estimate the mode shape coefficients at the remaining floor levels. The application of the procedure has been demonstrated through a numerical experiment on an eight‐storied torsionally coupled shear building subjected to earthquake base excitation. The results show that the proposed parameter identification technique is capable of identifying dominant modal parameters and responses even with significant noise contamination of the response records. Copyright © 2008 John Wiley & Sons, Ltd.     

Aftershock collapse vulnerability assessment of reinforced concrete frame structures

In a seismically active region, structures may be subjected to multiple earthquakes, due to mainshock–aftershock phenomena or other sequences, leaving no time for repair or retrofit between the events. This study quantifies the aftershock vulnerability of four modern ductile reinforced concrete (RC) framed buildings in California by conducting incremental dynamic analysis of nonlinear MDOF analytical models. Based on the nonlinear dynamic analysis results, collapse and damage fragility curves are generated for intact and damaged buildings. If the building is not severely damaged in the mainshock, its collapse capacity is unaffected in the aftershock. However, if the building is extensively damaged in the mainshock, there is a significant reduction in its collapse capacity in the aftershock. For example, if an RC frame experiences 4% or more interstory drift in the mainshock, the median capacity to resist aftershock shaking is reduced by about 40%. The study also evaluates the effectiveness of different measures of physical damage observed in the mainshock‐damaged buildings for predicting the reduction in collapse capacity of the damaged building in subsequent aftershocks. These physical damage indicators for the building are chosen such that they quantify the qualitative red tagging (unsafe for occupation) criteria employed in post‐earthquake evaluation of RC frames. The results indicated that damage indicators related to the drift experienced by the damaged building best predicted the reduced aftershock collapse capacities for these ductile structures. Copyright © 2014 John Wiley & Sons, Ltd.     

Correlating measured and simulated dynamic responses of a tall building to long‐distance earthquakes

For almost a decade, a 66‐storey, 280m tall building in Singapore has been instrumented to monitor its dynamic responses to wind and seismic excitations. The dynamic characteristics of the tall building have been investigated via both the finite element method and the experimental modal analysis. The properties of the finite element model have been shown to correlate well with those derived from the data recorded during the ambient vibration tests. During the study period, 21 sets of earthquake ground motions have been recorded at the building site. The basement motions may be divided into three categories based on their predominant frequency components with respect to the building's fundamental frequency. The calibrated three‐dimensional finite element model is employed to simulate the seismic response of the tall building. Correlation analysis of the time histories between the recorded data and the simulated results has been carried out. The correlation analysis results show that the simulated dynamic response time histories match well with those of the recorded dynamic responses at the roof level. The results also show that the simulated maximum response at the roof level is close to the peak response recorded during the earthquakes. Copyright © 2004 John Wiley & Sons, Ltd.     

Correlation of dynamic characteristics of a super‐tall building from full‐scale measurements and numerical analysis with various finite element models

The Di Wang Tower located in Shenzhen has 79 storeys and is about 325 m high. Field measurements have been conducted to investigate the dynamic characteristics of the super‐tall building. In parallel with the field measurements, seven finite element models have been established to model the multi‐outrigger‐braced tall building and to analyse the effects of various arrangements of outrigger belts and vertical bracings on the dynamic characteristics and responses of the Di Wang Tower under the design wind load and earthquake action. The distributions of shear forces in vertical structural components along the building height are also presented and discussed. The results from detailed modelling of group shear walls with several types of finite elements are addressed and compared to investigate various modelling assumptions. Finally, the performance of the finite element models is evaluated by correlating the natural frequencies and mode shapes from the numerical analysis with the finite element models and the field measurements. The results generated from this study are expected to be of interest to professionals and researchers involved with the design of tall buildings. Copyright © 2004 John Wiley & Sons, Ltd.     

多尺度分割和深度学习相结合的倾斜摄影三维影像建筑物震害信息提取

采用多尺度分割和深度学习相结合的方法对震后倾斜摄影三维影像建筑物震害信息进行提取,获取建筑物的屋顶和墙体多种破坏信息。以2017年九寨沟M_S7.0地震后倾斜摄影三维影像为例,依据三维影像建筑物顶面和墙体等进行样本的多尺度分割,样本分为完好建筑物面、破坏建筑物面、其它地物和背景等三类,选取211个100×100像素的样本集对卷积神经网络模型进行训练,采用训练后的模型提取灾区千古情风景区和漳扎镇小学的建筑物震害信息,并将提取结果与目视解译结果进行精度对比,结果显示:破坏建筑物面提取精度分别为65.5%和71.1%,总体分类精度分别为82.1%和84.1%,卡帕（Kappa）系数分别为68.7%和64.9%,表明该方法在倾斜摄影三维影像建筑物震害提取方面具有一定的优势。      

Seismic response analysis on the stability of running vehicles

The seismometer network of the Japanese expressway system has been enhanced since the 1995 Kobe earthquake. Using earthquake information from the instruments, the expressways are closed if the peak ground acceleration (PGA) is larger than or equal to 80cm/s². The aim of this regulation is to avoid secondary disasters, e.g. cars running into the collapsed sections. However, recent studies on earthquake damage have revealed that expressway structures are not seriously damaged under such‐level of earthquake motion. Hence, we may think of relaxing the regulation of expressway closure. But before doing this, it is necessary to examine the effects of shaking to automobiles since the drivers may encounter difficulties in controlling their vehicles and traffic accidents may occur. In this study, a vehicle was modelled with a six‐degree‐of‐freedom system and its responses were investigated with respect to PGA, peak ground velocity (PGV) and Japan Meteorological Agency (JMA) seismic intensity using five ground motion records. It was observed that the response of the vehicle shows a larger amplitude for the record that has larger response spectrum in the long period range compared to other records. However, similar response amplitudes of the vehicle were observed for all the records with respect to the JMA seismic intensity. The response characteristics of the vehicle model may be very useful for decision‐making regarding the relaxation of the expressway closure under seismic motion. Copyright © 2002 John Wiley & Sons, Ltd.     

Influence of spatial correlation of strong ground motion on uncertainty in earthquake loss estimation

In addition to the mean values of possible loss during an earthquake, parameters of the probability distribution function for the loss to a portfolio (e.g. fractiles and standard deviation) are very important. Recent studies have shown that the proper treatment of ground‐motion variability and, particularly, the correlation of ground motion are essential for the estimation of the seismic hazard, damage and loss for distributed portfolios. In this study, we compared the effects of variations in the between‐earthquake correlation and in the site‐to‐site correlation on seismic loss and damage estimations for the extended objects (hypothetical portfolio) and critical elements (e.g. bridges) of a network. A scenario earthquake approach and a portfolio containing a set of hypothetical building and bridges were used for the purpose. We showed that the relative influences of the types of correlation on characteristics of loss distribution and the probability of damage are not equal. In some cases, when the median values of loss distribution or the probability that at least one critical element of a lifeline will be damaged are considered and when the spatial correlation of ground motion is used, the possible variations in the between‐earthquake correlation may be neglected. The shape of the site‐to‐site correlation function (i.e. the rate of decrease of the coefficient of spatial correlation with separation distance) seems also to be important when modelling spatially correlated ground‐motion fields. Copyright © 2010 John Wiley & Sons, Ltd.     

Viscoelastic coupling dampers (VCDs) for enhanced wind and seismic performance of high‐rise buildings

As high‐rise buildings are built taller and more slender, their dynamic behavior becomes an increasingly critical design consideration. Wind‐induced vibrations cause an increase in the lateral wind design loads, but more importantly, they can be perceived by building occupants, creating levels of discomfort ranging from minor annoyance to severe motion sickness. The current techniques to address wind vibration perception include stiffening the lateral load‐resisting system, adding mass to the building, reducing the number of stories, or incorporating a vibration absorber at the top of the building; each solution has significant economic consequences for builders. Significant distributed damage is also expected in tall buildings under severe seismic loading, as a result of the ductile seismic design philosophy that is widely used for such structures. In this paper, the viscoelastic coupling damper (VCD) that was developed at the University of Toronto to increase the level of inherent damping of tall coupled shear wall buildings to control wind‐induced and earthquake‐induced dynamic vibrations is introduced. Damping is provided by incorporating VCDs in lieu of coupling beams in common structural configurations and therefore does not occupy any valuable architectural space, while mitigating building tenant vibration perception problems and reducing both the wind and earthquake responses of the structure. This paper provides an overview of this newly proposed system, its development, and its performance benefits as well as the overall seismic and wind design philosophy that it encompasses. Two tall building case studies incorporating VCDs are presented to demonstrate how the system results in more efficient designs. In the examples that are presented, the focus is on the wind and moderate earthquake responses that often govern the design of such tall slender structures while reference is made to other studies where the response of the system under severe seismic loading conditions is examined in more detail and where results from tests conducted on the viscoelastic material and the VCDs in full‐scale are presented. Copyright © 2013 John Wiley & Sons, Ltd.     

Parameter identification of framed structures using an improved finite element model‐updating method—Part II: application to experimental data

In this study, we determine an updated finite element model of a reinforced concrete building—which was damaged from shaking during 1994 Northridge earthquake—using forced‐vibration test data and a novel model‐updating technique. Developed and verified in the companion paper (viz. BVLSrc, Earthquake Eng. Struct. Dyn. 2006; this issue), this iterative technique incorporates novel sensitivity‐based relative constraints to avoid ill conditioning that results from spatial incompleteness of measured data. We used frequency response functions and natural frequencies as input for the model‐updating problem. These data were extracted from measurements obtained during a white‐noise excitation applied at the roof of the building using a linear inertial shaker. Flexural stiffness values of properly grouped structural members, modal damping ratios, and translational and rotational mass values were chosen as the updating parameters, so that the converged results had direct physical interpretations, and thus, comparisons with common parameters used in seismic design and evaluation of buildings could be made. We investigated the veracity of the updated finite element model by comparing the predicted and measured dynamic responses under a second, and different type of forced (sine‐sweep) vibration, test. These results indicate that the updated model replicates the dynamic behaviour of the building reasonably well. Furthermore, the updated stiffness factors appear to be well correlated with the observed building damage patterns (i.e. their location and severity). Copyright © 2006 John Wiley & Sons, Ltd.     

A new model order reduction strategy adapted to nonlinear problems in earthquake engineering

Earthquake dynamic response analysis of large complex structures, especially in the presence of nonlinearities, usually turns out to be computationally expensive. In this paper, the methodical developments of a new model order reduction strategy (MOR) based on the proper orthogonal decomposition (POD) method as well as its practical applicability to a realistic building structure are presented. The seismic performance of the building structure, a medical complex, is to be improved by means of base isolation realized by frictional pendulum bearings. According to the new introduced MOR strategy, a set of deterministic POD modes (transformation matrix) is assembled, which is derived based on the information of parts of the response history, so‐called snapshots, of the structure under a representative earthquake excitation. Subsequently, this transformation matrix is utilized to create reduced‐order models of the structure subjected to different earthquake excitations. These sets of nonlinear low‐order representations are now solved in a fractional amount of time in comparison with the computations of the full (non‐reduced) systems. The results demonstrate accurate approximations of the physical (full) responses by means of this new MOR strategy if the probable behavior of the structure has already been captured in the POD snapshots. Copyright © 2016 The Authors. Earthquake Engineering & Structural Dynamics Published by John Wiley & Sons Ltd.     

Active interaction control of civil structures. Part 2: MDOF systems

In this paper, the performance of active interaction control (AIC) algorithms is assessed within the context of two realistic building models. The AIC control approach is proposed as a semi‐active means of mitigating the structural response during large earthquakes. To implement the AIC control algorithms into MDOF systems, the modal control (MC) approach that directs the control effort to certain dominant response modes is formulated and utilized herein. Two structures, a 3‐storey building and a 9‐storey steel‐framed benchmark building controlled by the AIC algorithms are analysed for two historical earthquake records. The results of numerical simulation verify the efficacy of the AIC control algorithms in controlling vibration of building structures during large earthquakes. Copyright © 2001 John Wiley & Sons, Ltd.