Development of optimized damage prediction method for health monitoring of ultra high performance fiber-reinforced concrete communication tower

• Main Author: Gatea, Sarah Jabbar
• Format: Thesis
• Language: English
• Published: 2018
• Online Access: http://psasir.upm.edu.my/id/eprint/76403/1/FK%202018%2075%20IR.pdf; http://psasir.upm.edu.my/id/eprint/76403/

The requirement for communication towers increases due to the growing demand for power supply and telecommunication services. Recently, many attempts have been exerted to monitor the tower to ensure its excellent performance during operation. The capability of the tower to detect, localize, and quantify structural damage is the most important factor in maintaining excellent performance, reliability, and costeffectiveness and ensuring its stability and integrity. The dynamic analysis of tall slender towers is commonly performed in the frequency domain. However, the recorded frequencies can be noisy, random, unstable, and with skewed data. The damage, due to uncontrolled noise reciprocating motion in the machines or broadband noise from wind or other sources, is identified based on frequency testing in an operator. Therefore, this study aims to develop a new health monitoring system for communication towers based on AdaBoost, Bagging, and RUSBoost algorithms as hybrid algorithm, which can predict the damage by using noisy, random, unstable, and skewed frequency data with high accuracy. For this purpose, a UHPFRC tower with 30-m height is considered, and the finite element model (FEM) of the tower is developed. The modal frequencies of the tower are evaluated under various conditions of damage in concrete and connection in different parts of the tower by using finite element simulation. The results are used to develop the hybrid learning algorithm based on the AdaBoost, Bagging, and RUSBoost methods to predict the damage in the tower based on dynamic frequency domain. Therefore, 78 damage scenarios have been simulated by using finite element software to generate the frequency of the UHPFRC communication tower with various types of damage. The damage scenarios consist of losing bolts and vertical and horizontal cracks. The frequency before and after damage was set as input training data, whereas the damage types and locations are set as output data (damage index). The verification results indicate that all the structural defects were predicted with high accuracy by the developed hybrid algorithm in cases of healthy and damaged structures. The full-scale UHPFRC communication tower is experimentally tested for dynamic frequencies to verify the numerical analysis results. The frequency response of the tower structure was obtained by exciting with an impact hammer at various points, and the acceleration of the tower structure was gathered through three accelerometer sensors attached at the top, middle, and bottom parts of the structure. Damaging the full-scale tower is not practical; thus, two different parts of the tower segments and their connections (1-2 and 2-3) are considered and tested experimentally with and without damage to validate the capability of the developed hybrid algorithm to detect damage. A dynamic actuator was used to cause damage in the tower segments by applying vibration force. In addition, a simple procedure is proposed to determine the optimal solution and predict the correlation factor and the frequency of the damaged communication tower by using the particle swarm optimization (PSO) method. This technique avoids the exhaustive traditional trial-and-error procedure to obtain the coefficient of the correlation factor of frequency for the damaged communication tower by conducting several analyses. The new assessments on the capability of the indicator to detect and quantify the defects are performed. For this purpose, the FEM is implemented to model three communication towers with a height of 15, 30, and 45m to develop the frequency correlation factor. The verification results indicate that the PSO technique can develop a correlation factor with acceptable accuracy to predict the damage.