Development of structural health monitoring and artificial intelligence based damage detection and early warning system for truss like in-water structures

---

Tezin Türü: Yüksek Lisans

Tezin Yürütüldüğü Kurum: Orta Doğu Teknik Üniversitesi, Fen Bilimleri Enstitüsü, Türkiye

Tezin Onay Tarihi: 2019

Tezin Dili: İngilizce

Öğrenci: SERAP KARA

Danışman: Ahmet Türer

Özet:

Continuously growing industry necessitates substantial amount of energy which is frequently harvested in-water regions. Electricity, gas, and oil collected offshore areas generally require platforms constructed in seas and oceans. Inspection of offshore platforms is hard and it is difficult to detect damage in early stages of these structures that perform in harsh salty water environments, which might cause catastrophic failures if unattended. Damage identification of these structures is a challenging issue. Structural damage detection at its earlier stage would prevent possible collapses, economic losses, and environmental disasters. Structural Health Monitoring (SHM) approach for early damage detection and warning for truss-like tower structures in water is investigated in this thesis study. This thesis targeted to set rules about damage detection and warning system using simple and sophisticated methods (Artificial Intelligence – AI) for tower like structures. Wave loading, stored mass changes, and damaged members are considered in the developed monitoring system, which is verified by tests on a small scale laboratory model. Finite Element Model (FEM) of the lab model is calibrated using static and dynamic test results. Damage scenarios are generated using a VBA based program through Application Programming Interface (API) to generate analytical structural response of damage scenarios. An artificial intelligence (AI) based monitoring system is developed using a large number of analytically simulated damage and mass change scenarios. The trained ANN is tested using FEM generated structural response that are not used during training. Total of 22 damage scenarios are also implemented on the physical lab model and tests are conducted to obtain dynamic and static properties which are then fed into the trained ANNs to see if damage locations can be identified. A Graphical User Interface (GUI) is also developed to show the current health status of the structure including specific details such as natural vibration frequencies, current mass in the tank, displacements, etc. A damage level warning system is also implemented on the GUI to alarm for any detected damage.