Democritus University of Thrace (DUTH)
Ευφυείς τεχνικές επεξεργασίας σεισμογραφημάτων για ταξινόμηση βλαβών σε δομικές κατασκευές
Abstract
dc:descriptionRecently, the field of earthquake engineering has increased the interest of more scientists. Modern structures are based on new earthquake-resistant building codes where more design criteria are taken into account. In addition, the demand for the assessment of seismic vulnerability of existing buildings is arisen in order to minimize or avoid the effects. Although safety against collapse remains the main goal of design, the performance of constructions has a significant role, taking into account functional and economical measures. To date, a major effort for the improvement of earthquake - resistant design has taken place, not only to avoid collapse, but also to reduce the damages after the manifestation of moderate earthquakes. For the implementation of the above ideas is required to define damage measures. The major problem in the creation of earthquake-resistant building codes is the creditable estimation of ground motion's damage potential. If the valuated intensity of designed earthquake ground motion is over-estimated, then the cost of new buildings' construction and existing buildings' reinforcement is increased. In reverse, if the valuated intensity is under-estimated, then massive disasters and human lives loss may occur. To avoid this prospect, a secure definition of the seismic intensity should be linked to the buildings' caused damages. Besides, the estimation of seismic intensity with the use of seismic parameters may contribute to the creation of warning systems in order to avoid the secondary consequences of earthquake, e.g. shut off gas line, activation of security systems in nuclear plant. Until nowadays a great number of seismic parameters have been proposed for the estimation of seismic hazard and the classification of seismic signals. The existing seismic parameters are not sufficient for the description and evaluation of damage potential of a seismic excitation, since earthquake is a complex phenomenon during its manifestation and its consequences. Furthermore, a great deal of research, which has been made for the seismic intensity's correlation to the damages in buildings, has revealed the inability for the description of damage potential with the existing parameters. The present PhD thesis is an effort for the extraction of seismic parameters, which can be considered as quantitative measures of seismic excitations' damage potential. The seismic parameters are the main tool for the characteristics' description of strong earthquakes in a compact and representative form. In a signal's characteristics we distinguish amplitude, frequency content and duration. For the extraction of new ground motion parameters, the earthquake analysis was performed using an innovating transform (Hilbert-Huang). As the supremacy of the new transform in comparison with well-known signal processing methods (Fourier analysis, wavelet) is noted, the search of characteristics derived from the time frequency-energy representation of the signal (Hilbert spectrum). The novelty of the present research is the extraction of seismic parameters with the use of Hilbert-Huang transform and the definition of those signal components, which have a direct impact on the damages of the constructions under study. The advantage of the above method is that it doesn't require an a priori assumption for the basis functions. In reverse, the signal components extract from the characteristics of the signal itself. This is important, since this way signal with physical meaningful are extracted. The choice of the components, which are responsible for the caused damages, was based on the building's eigenfrequency. The new seismic parameters are extracted from the components with frequency content close to the building's eigenfrequency. In contrast to common seismic parameters, which are defined in the field of time and express amplitude or energy, such as PGA, PGV, lA, RMSa, CAV, lFVF, DPAS, P090/ Einp, the new parameters distinguish, because their energy is related to frequency content. They present a great correlation to damage indicators PARK/ANG and MISDR. The Spearman's correlation coefficients for the new parameters are in the range of 0.854-0.901. Moreover, they present greater correlation toward some classic parameters, such as PGA, iA, P090, CAV, RMS, DPAS, EPAMAX, SMDTB and CP. By knowing the signal, which is responsible for the earthquake damages, the research advanced the improvement of certain widespread seismic parameters. The results were quite encouraging. Significant parameters, which were defined in the field of time and not related with frequencies such as PGA, lA, P090, CAV, and RMS, presented noticeable improvement, up to 31% in the case of PGA. In reverse, some parameters, such as SMD and CP, which in general show small correlation to damages, with their connection to the parameters under study showed even smaller correlation. These results lead to the conclusion that with the use of signals close to the building's eigenfrequency a better polarization between parameters with small and high correlation takes place. Besides the extraction of seismic parameters, the present research included also the design and implementation of a system for earthquake damage estimation on reinforced concrete buildings. For automatic assessment of structural, architectural, mechanical and contents damages three different classification systems were used. Firstly, an artificial neural network was examined. Since the artificial neural networks try to map inputs to outputs, in the above system a set of seismic parameters was utilized as an input vector, and the induced damage (damage category) as an output. Secondly, a Neuro-Fuzzy classification system was tested and its fuzzy rules were defined by a set of training accelerograms. The results' comparison was fulfilled with the implementation of a system based on Support Vector Machines. Firstly, the representation of seismic accelerograms in the proposed systems has been made with the use of 20 seismic parameters. Then a Genetic algorithm was used for the reduction of the seismic parameters' number, with main goal the performance improvement of the proposed classification systems. High percentages of correct classification were observed. In the case of structural damages {MISDR) were up to 90%. In structural damages based on damage indicator PARK/ANG, significant improvement was achieved with the percentage of correct classification rising up to 85%. To establish the reliability of the above method, the proposed systems were tested in tow other structures. Furthermore, the classification system based on SVMs was able, also, to classify the seismic events in two main categories with a classification rate up to 100%. In contrast to the existing researches, the proposed systems show the same stability, since their performance does not depend on the structure under study, while in unknown earthquakes they provide better results.
Degree
thesis:*- Grantor dc:publisher
- Democritus University of Thrace (DUTH)
- Year dc:date
- 2011
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Alvanitopoulos, Petros-Fotios
- Αλβανιτόπουλος, Πέτρος-Φώτιος
Subjects
dc:subject × 14- Σεισμικές παράμετροι
- Δείκτες βλάβης
- Μετασχηματισμός Hilbert - Huang
- Ταξινόμηση σεισμικών βλαβών σε δομικές κατασκευές
- Seismic parameters
- Damage indices
- Hilbert - Huang transform
- Earthquake damage classification in RC structures
- Επιστήμες Μηχανικού και Τεχνολογία
- Επιστήμη Πολιτικού Μηχανικού
- Επιστήμη Μηχανολόγου Μηχανικού
- Engineering and Technology
- Civil Engineering
- Mechanical Engineering
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/29631