Democritus University of Thrace (DUTH)
Automatic classification of earthquake - induced building damages
Abstract
dc:descriptionEarthquake engineering aims at developing techniques that will prevent damage in a minor quake and avoid serious damage or collapse of structures in a major shake. Earthquake engineering is the scientific field concerned with protecting the society, the natural environment, and the man-made environment from earthquakes by limiting the seismic risk to socio-economically acceptable levels. While no structure can be entirely immune to damage from earthquakes, the goal of earthquake engineering design is to erect structures that fare better during seismic activity than their conventional counterparts. Post-earthquake damage estimation has become important in the last decades. Seismic performance assessment is a powerful tool of earthquake engineering which utilizes detailed modeling of the structure combined with methods of structural analysis to gain a better understanding of seismic performance of buildings. In order to perform structural damage assessment on buildings, reliable damage measures must be defined; indices that describe the earthquake’s damage potential and parameters that describe the seismic intensity. Moreover, these seismic intensity parameters must be related to damage indices. The correlation of the seismic intensity parameters with actual levels of damage is a multi-parameter subject of study. Several seismic parameters have been proposed for the representation of the severity of seismic signals. A great number of researches concerning the seismic intensity correlation to the damages in buildings have revealed the weakness for the description of damage potential with the existing parameters. Furthermore, the lack of adequate seismic record in some areas with limited anti-seismic funding allocation, poses an additional problem in the development of automatic damage assessment methods. The existence of strong motion records in areas of interest could be utilized to create simulation programs with realistic models and thus describe the consequences of the seismic excitation on certain structures of the examined area.The present Ph.D. thesis is an effort to resolve the aforementioned problems and estimate accurately the vulnerability of reinforced concrete buildings based on intelligent techniques. In the first stage of this effort, the interest concentrates on the generation of artificial accelerograms. In the second stage, a new set of seismic parameters is proposed. Finally, all the above findings are combined and tested to certain automatic damage estimation models.Artificial spectrum-compatible accelerograms are widely used in dynamic analysis instead of natural earthquake records. In this work, four different methodologies for generating spectrum-compatible accelerograms are developed. Each of the proposed methods investigates the various aspects of artificial seismic signals composition. The novelty of the proposed methodologies is that they are all based on the recently introduced Hilbert-Huang Transform (HHT), which appears to be an effective technique for analyzing non-stationary and nonlinear signals such as seismic signals. All previously reported methods for generation of artificial accelerograms are based on conventional signal processing techniques such as Fourier series and wavelets. Although Fourier analysis is a strong mathematical tool, its efficacy is related to the type of analyzed data. Fourier analysis requires the fulfilment of two conditions; linearity of the system and non-stationary nature of signals. Therefore, Fourier analysis is an inadequate tool for analyzing non-stationary and nonlinear signals like seismic records. Wavelets take into account the nonlinear and non-stationary characteristics of seismic signals. Yet, this approach still lacks effectiveness due to the definition of the mother wavelet function which results in poor energy-frequency-time distribution. In the HHT the extraction of the components is based on the local features of the signal. In the first method, the HHT is introduced in seismic accelerograms generation. In the second method only the strong motion duration of the seismic signal is analyzed in order to eliminate a reported drawback of the HHT, the end effects. The third method additionally utilizes only the appropriate components related to the eigenfrequency of low-, mid- and high-rise buildings that influence the response of structures and, unlike all other proposed methods, it is not an iterative method. In the last technique the Complementary Empirical Mode Decomposition (CEEMD) is utilized to resolve the mode-mixing phenomenon, a drawback of the HHT which is not considered in any of the previous reported approaches. Moreover, from the generated seismic signals, a set of well-known parameters is extracted. It is shown that all extracted parameters from synthetic accelerograms reveal the same degree of interrelation with damage indices as natural accelerograms. Thus it is proven that the artificially generated seismic signals exhibit the same behaviour as natural ones. To quantify the efficiency of the methods, a comparison with two reference methods, SIMQKE and SeismoArtif, takes place. All methodologies are proved to be efficient as they generate accelerograms that 1) exhibit high compatibility with the desired target spectrum 2) are non uniform in time domain since they do not follow predetermined envelope shapes and are based on the local properties of the signal 2) follow the restrictions of the Eurocode 8 and the simulated spectra match the target spectrum in all periods. After the problem of low disposal in input earthquake signals for dynamic analysis is solved, research is extended in extracting new seismic intensity parameters, which could mirror the seismic damage potential effectively. Up to now, a number of seismic intensity parameters have been proposed. However, seismic signals are complex natural phenomena and the existing seismic parameters are not sufficient for the description and evaluation of their damage potential. The novelty of the new set of seismic parameters proposed is that they are associated with the dissipated energy and the frequency content of the seismic signal. Moreover, a new definition for seismic strong motion duration is introduced, which is directly related to the evolution of the damage occurred to the examined structure. For the extraction of the new set of seismic intensity parameters HHT and Ensemble Empirical Mode Decomposition (EEMD) are employed. As the superiority of HHT is already noted, the introduction of EEMD additionally resolves the mode-mixing phenomenon which is not considered in any of the previous approaches for seismic feature extraction. Thus, a time-frequency analysis is performed to the specific frequency band defined from the introduced strong motion duration of the seismic accelerogram so as to extract the new set of seismic features. The aforementioned analysis is carried out only to those signal components which have a direct impact on the damages of the construction under study. A correlation study is conducted and shows that the extracted seismic intensity parameters are strongly correlated to the structural damage. Therefore, the new seismic intensity parameters constitute an efficient data set, which can be utilized to represent the seismic intensity of signals in an intelligent system for structural damage estimation.The aforementioned research is essentially the background work for creating a model for structural damage assessment. Having artificially generated accelerograms of any desired intensity available, along with an efficient data set of seismic parameters strongly related to structural damage, the research is completed with the implementation of a system suitable for damage estimation. All extracted seismic parameters and damage indices refer to a certain reinforced concrete (RC) frame model utilized throughout the present project. However, the research methodology is structured in such a way as to be general in nature thus, being in a form that renders it ready to be implemented in different engineering design provisions with only minor changes.For the automatic assessment of structural damage, four models are tested. First an Adaptive Neuron-Fuzzy Inference System (ANFIS) is presented. A number of twenty well-known seismic parameters and a set of two hundred artificial accelerograms with known damage effects are utilized. The damage caused by these signals is evaluated and categorized into four classes according to two widely used damage indices, Maximum Inter-Storey Drift Ratio (MISDR) and Global Damage Index after Park/Ang (DIG,PA). The unknown accelerograms are tested and classified with the help of an ANFIS with correct classification rate up to 90% and 87%, when the MISDR and DIG,PA are used respectively as the criterion for structural damage. Furthermore, a feature selection process takes place. Results show that reducing the number of parameters according to their interdependence with the MISDR, does not affect the effectiveness of the method.Classification results are not always reliable indicators of the post-earthquake status of buildings. By estimating the damage through classification, it is common to assign the damage, which belongs to the edges of the predefined intervals, in the wrong category. The aim of the consequent work is to estimate numerically the MISDR structural damage through a time-frequency analysis of the seismic signal. One hundred earthquake records are utilized to test the remaining three proposed models; a Mamdani-type Fuzzy Inference System (FIS), a Sugeno-type FIS and a Neural Network (NN). One hundred natural and artificial earthquake signals are utilized to train and test the proposed models. The models are trained to estimate the MISDR value induced by a seismic signal in a certain structure. Every tested seismic signal is inserted as input to the models in terms of eight new proposed seismic intensity parameters. The FIS is combined with a genetic algorithm (GA) to tune its MFs optimally. Results reveal that the two models perform similarly in estimating the MISDR value, since the average mean square error (MSE) of all experiments is almost the same; 0.253 for Mamdani FIS and 0.289 for Sugeno FIS. When the MISDR estimation values are assigned to damage categories, correct classification of up to 91% is achieved with the Mamdani FIS, while 87% with the Sugeno FIS. Finally, for the same set of earthquake signals, and the same input and output parameters, a NN is developed. Experimental results reveal a total average MSE of MISDR estimation equal to 0.128; results translated to classification rate correspond to 96%. The contribution of the presented techniques is the numerical computation of MISDR using the proposed models, which has not been previously reported in the literature. The proposed models evaluate the post-seismic damage status of buildings in the form of damage indices, avoiding complicated and time consuming nonlinear dynamic analysis. The proposed techniques can effectively detect the structural damage degree quantitatively and rapidly. This report provides a practical solution to post-seismic management as it enables civil authorities to estimate the potential damage from future earthquakes. Thus, the repair and aid actions can be automatically targeted to the affected areas only, immediately after the seismic event. The proposed techniques could be useful tools for an on-line estimation of the structural damage on buildings immediately after the occurrence of an earthquake.
Degree
thesis:*- Grantor dc:publisher
- Democritus University of Thrace (DUTH)
- Year dc:date
- 2016
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Vrochidou, Eleni
- Βροχίδου, Ελένη
Subjects
dc:subject × 11- Επεξεργασία σήματος
- Ανάλυση σήματος στο πεδίο συχνότητας-χρόνου
- Μετασχηματισμός Hilbert - Huang
- Κατασκευή τεχνητών επιταχυνσιογραφημάτων
- Ταξινόμηση δομικών βλαβών σε κτήρια
- seismic signal processing
- Hilbert-Huang transform
- Επιστήμες Μηχανικού και Τεχνολογία
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Engineering and Technology
- Electrical Engineering, Electronic Engineering, Information Engineering
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/38529