Geri Dön

Deprem etkisinde taban yalıtımlı bina tasarımı ve makine öğrenmesi algoritmalarıyla deplasman doğruluklarının tespiti

Base isolated building design under the effect of earthquake, and the calculation of displacement accuracy with machine learning algorithms

PDF İndir

  1. Tez No: 731292
  2. Yazar: AHMET KARAKAYA
  3. Danışmanlar: DOÇ. DR. BEYZA TAŞKIN AKGÜL
  4. Tez Türü: Yüksek Lisans
  5. Konular: Deprem Mühendisliği, İnşaat Mühendisliği, Earthquake Engineering, Civil Engineering
  6. Anahtar Kelimeler: Belirtilmemiş.
  7. Yıl: 2022
  8. Dil: Türkçe
  9. Üniversite: İstanbul Teknik Üniversitesi
  10. Enstitü: Lisansüstü Eğitim Enstitüsü
  11. Ana Bilim Dalı: Deprem Mühendisliği Ana Bilim Dalı
  12. Bilim Dalı: Deprem Mühendisliği Bilim Dalı
  13. Sayfa Sayısı: 143

Özet

İnsanoğlu, yüzyıllar boyunca yüksek sayıda can ve mal kaybına neden olan depremin meydana getirdiği dinamik etkilerden korunmanın yollarını aramıştır. Bu etkilerden korumayı düşündükleri ilk alanlar ise içerisinde yaşadıkları konutları olmuştur. Çoğunluğu aktif fay bölgeleri içerisinde bulunan ülkemizde ise can ve mal kaybının önlenmesi adına depreme dayanıklı ve sürekli kullanımı hedefleyen tasarım yöntemlerinin önemi gün geçtikçe artmaktadır. Özellikle, merkez üssü Gölcük olan 17 Ağustos 1999 depreminden sonra, bu konu ile ilgili standartların ve yönetmeliklerin güçlendirilmesi ve uygulamada alınan önlemlerin artması, ülkemiz adına bu konuda atılmış en önemli adımlardan birkaçı olmuştur. Bu çalışma, dört katlı bir binanın, tabanına yerleştirilen on altı adet kurşun çekirdekli kauçuk deprem yalıtım birimiyle birlikte, deprem etkisi altında taban yalıtımlı olarak tasarlanmasını içermektedir. Ayrıca, aynı binaya beş yüz yetmiş beş farklı özelliğe sahip olan kurşun çekirdekli kauçuk deprem yalıtım biriminin uygulanmasıyla birlikte, TBDY 2018'e göre en büyük deprem yer hareketi seviyesi altında, alt ve üst limitlerinin ortalaması ile elde edilen deplasman değerlerinin sonucu olarak oluşturulan veri tabanıyla, altı farklı makine öğrenmesi algoritması çalıştırılmış, başarı oranları tespit edilmiş ve algoritmalar da kendi aralarında karşılaştırılmıştır. Bu çalışmadaki amaç, farklı özelliklere sahip bir yalıtım birimi türünün, aynı özellikteki bir binaya uygulandığında, oluşabilecek deplasman değerinin hızlı bir şekilde tespiti ve yaklaşık bir fikir vermesidir. Bu tez çalışması, dört bölümden oluşmaktadır. Giriş bölümünde, deprem etkisinde taban yalıtımlı bina tasarımı ve makine öğrenmesi kavramlarından bahsedilmiştir. İkinci bölümde, deprem etkisinde taban yalıtımlı bina tasarımının dinamik teorisi ve mekanizmasından detaylı bir şekilde bahsedilmiştir. Ayrıca bu bölümde, SAP2000 v20 yazılımı ile deprem yalıtım birimiyle birlikte binanın modellenmesini, deprem etkisi altında tasarımını, farklı hesap yöntemlerine göre elde edilen deprem yüklerinin ve bu yüklerin elverişsiz olarak uygulanması ile oluşan deplasman değerlerinin TBDY 2018'e göre değerlendirilmesini içermektedir. Üçüncü bölümde, makine öğrenmesi'nin ortaya çıkış süreci ve çalışmada uygulanan altı makine öğrenmesi algoritmasının çalışma mekanizması aktarılmış, python yazılım dili ile edilen sonuçlar gösterilmiş ve değerlendirilmiştir. Dördüncü bölümde ise hem deprem etkisi altında tasarlanan binadan hem de makine öğrenmesi algoritmalarından elde edilen sonuçlar, genel anlamıyla değerlendirilmiştir.

Özet (Çeviri)

Mankind has always tried to find ways to protect itself from the devastating effects of earthquakes causing great loss of life and damage to property since the beginning of time. Our first thought has always been to protect our homes. In our country, which is located in the middle of active fault zones, the importance of earthquake-safe and sustainable structural design to prevent loss of life and property is increasing every day. Particularly, after the earthquake of 17th August 1999, the epicenter of which was Gölcük, the strengthening of the structural standards and regulations governing the matter, and their consequent applications have been the most important steps taken in this regard. This study includes the design of a four story building with sixteen lead rubber bearing units placed on its base, with base isolation, under the effect of an earthquake. In accordance with Building Seismic Code of Turkey (TBDY) 2018, five hundred and seventy five different lead rubber bearing units taken from the Bridgestone lead rubber bearing catalog were added to the building, which was designed with base isolation under the influence of earthquake, together with sixteen lead rubber bearing units placed on its base. The dataset was created using the result of the displacement values obtained by the average of the lower and upper limits below the level and these average displacement values. Simultaneously, six different, supervised machine learning algorithms were run, their accuracies were determined and the algorithms were compared among themselves. The aim of this study is to speedily determine the displacement value that may occur when a certain type of isolation unit with varied characteristics is applied to a building with the same characteristics and to give an approximate idea. From here, it is proven in detail, with the control mechanisms, that the purpose of the study is achieved. This thesis consists of four parts: In the introduction, the concepts of base isolated building design and machine learning under the effect of earthquakes are studied. In the second part, the dynamic theory and mechanism of the base isolated building design under the effect of earthquakes are discussed in detail. Also in this section, we take a detailed look at SAP2000 v20, a software that computes modeling of the building in tandem with the lead rubber bearing units, its design under the influence of earthquakes, the evaluation of seismic loads obtained according to different calculation methods, and the displacement values created by the unfavorable application of these loads according to TBDY 2018. In the third chapter, the inception of machine learning and the working mechanism of six machine learning algorithms applied in the study are explained, and the results with python software language are shown and evaluated. In the fourth chapter, the results obtained from both the building designed under the effect of earthquake and machine learning algorithms are assessed in general terms. This work includes the design of a four story building, whose structural system consists of reinforced concrete frames with high ductility, with sixteen lead rubber bearing units, in accordance with TBDY 2018. The most important standards used creating this design are“Requirements for Design and Construction of Reinforced Concrete Structures”(TS-500) and“Account Values of Loads to be Taken in Sizing of Building Elements”(TS-498). By providing the necessary flexibility of the building, which is designed with base isolation, its natural period increased and its damping ratio under the effect of earthquakes remained at a sufficient level. In addition, the earthquake isolation units used covered all vertical and horizontal loads both under the effect of the earthquake and in use. Concertedly, after the application of the base isolation system, which reduces the earthquake effect on the superstructure of the building, the relative floor drifts and floor shear forces decreased. As a result of the calculations made in accordance with TBDY 2018, no structural irregularity in the building was observed. Four different calculation methods were applied under the effect of earthquakes simultaneously. These are equivalent seismic load, effective seismic load, mode combination method, and nonlinear time history analysis. These methods were examined in detail in accordance with the relevant regulations and the efficacy of the results was demonstrated. For the equivalent seismic load method, a comparative calculation was made both with the program I coded in visual basic based on Excel, and using SAP2000 v20. The calculation with the effective seismic load method was based on the code (TBDY 2018). The mode combination method was calculated using SAP2000 v20 and the displacement and seismic load values obtained as a result of the effective seismic load method, due to the absence of structural irregularity in the building, meet the necessary conditions. As a result of eleven earthquake records that meet the necessary conditions in accordance with TBDY 2018 and their application to the building with the SAP2000 v20, rotated ninety degrees, a nonlinear time history analysis was made. The average of the twenty two displacement values obtained according to the unfavorable seismic load and the average of the seismic loads were compared with the displacement and seismic load values obtained as a result of the effective seismic load method due to the absence of structural irregularity in the building and it was observed that it met the necessary criteria. The results obtained from six different supervised machine learning algorithms The programming was done using python and the machine learning libraries in the program. These six algorithms are as follows; k nearest neighbor, logistic regression, support vector machines, decision trees, random forest, and gradient boosting. Step by step results were also obtained using python based jupyter notebook. These algorithms whose theory and the working mechanism mentioned are explained one by one and their success rates are presented in detail. The r square score, the best measure of how well the experimental data fits a linear curve, is the determination quotient calculated in the regression analysis process. The more data points there are, the higher the reliability of r squared, and it can be stated that the algorithms with values close to one (1) perform better. It is ascertained that the algorithm with the highest r square score supports vector machines with 98.83\%. It is observed that the algorithm with the lowest r square score is the k nearest neighbor with 94.14\%. The resulting values are reasonably close to one (1) and expected ratios are obtained. The mean square error calculates how close a correlation curve is to a set of points. This metric measures the performance of the machine learning algorithm, it is always positive, and it should be stated that algorithms with a value close to zero (0) perform better. It was observed that the algorithm with the highest mean square error is the k nearest neighbor with 8.70\%. It was observed that the algorithm with the lowest mean square error is support vector machines with 1.74\%. It is thus deduced that the values are rationally close to zero and the desired ratios are obtained. The results obtained from these algorithms were first compared within themselves and then with each other. It was observed that the algorithm with the highest success rate is the support vector machine with 98.26\% accuracy. Furthermore, it was observed that the algorithm with the lowest success rate is the k nearest neighbor with 91.30\% accuracy. This result leads us to the deduction that the implementations of all of these six algorithms have created satisfactory outcomes.

Benzer Tezler

  1. Tez No

    802029

    Sismik taban yalıtımı ve enerji sönümleyici sistemlerin mimari tasarıma etkileri

    Effects of seismic base isolation and energy damping systems on architectural design

    HİLAL GÜNAY

    Yüksek Lisans

    Türkçe

    Türkçe

    2023

    Mimarlıkİstanbul Teknik Üniversitesi

    Mimarlık Ana Bilim Dalı

    PROF. DR. NECDET TORUNBALCI

  2. Tez No

    865728

    Mevcut betonarme bir yapının sismik izolasyon ve geleneksel yöntemlerle güçlendirilmesinin sayısal olarak karşılaştırılması

    Numerical comparison of the strengthening of a reinforced concerete building by seismic isolation and conventional method

    AYSU BIYIKLI

    Yüksek Lisans

    Türkçe

    Türkçe

    2024

    İnşaat Mühendisliğiİstanbul Üniversitesi-Cerrahpaşa

    İnşaat Mühendisliği Ana Bilim Dalı

    PROF. DR. TURGAY ÇOŞGUN

  3. Tez No

    641157

    Sismik yalıtımlı yapılarda patlama etkilerinin incelenmesi

    Investigation of blast effects in seismic isolated buildings

    ELİF TOPLU

    Yüksek Lisans

    Türkçe

    Türkçe

    2020

    İnşaat MühendisliğiSakarya Uygulamalı Bilimler Üniversitesi

    İnşaat Mühendisliği Ana Bilim Dalı

    DR. ÖĞR. ÜYESİ OSMAN KIRTEL

  4. Tez No

    867382

    Sismik taban yalıtımlı bir binanın deprem yalıtım birimi ankrajlarının incelenmesi

    Investigation of the seismic isolator anchors of a base-isolated building

    İLKCAN CEYLAN

    Yüksek Lisans

    Türkçe

    Türkçe

    2024

    Deprem Mühendisliğiİstanbul Teknik Üniversitesi

    Deprem Mühendisliği Ana Bilim Dalı

    DR. ÖĞR. ÜYESİ PINAR ÖZDEMİR ÇAĞLAYAN

  5. Tez No

    766083

    Kurşun çekirdekteki ısınmanın sismik izolasyonlu yapıların davranışına etkisi

    Effect of lead core heating on the behavior of seismically isolated structures

    ZAFER KANBİR

    Doktora

    Türkçe

    Türkçe

    2022

    Deprem Mühendisliğiİstanbul Üniversitesi-Cerrahpaşa

    İnşaat Mühendisliği Ana Bilim Dalı

    PROF. DR. CENK ALHAN

    PROF. DR. GÖKHAN ÖZDEMİR

Geri Dön