چكيده به لاتين
Nowadays, bolted joints are widely used in megastructures and engineering structures such as airplanes, bridges, ships, pipelines, as well as satellite and transportation systems. In practical applications, bolted joints are exposed to various failure modes including self-loosening, slipping, shaking, fatigue cracks, and fractures. This can lead to the failure of the entire structure or negatively affect their performance. Therefore, the development of structural health monitoring systems is one of the essential and important solutions for evaluating the performance of megastructures, ultimately increasing reliability in their performance and transforming the maintenance process. In this research, the concepts in this field are introduced first, and a review of previous studies by researchers on various inspection methods for bolted connections, particularly in cylindrical flanged pipes, is conducted. The aim of this study is to use ultrasonic wave propagation in the cylindrical flanged shell to quantitatively and qualitatively examine the loosening of bolted connections, determine the vibrational behavior of the system, and detect failures in structural connections. Thus, by determining different values of preload forces in the bolts and performing experimental tests related to ultrasonic wave propagation, along with validating the simulated model in the finite element software environment with the test results, a large dataset is prepared, in which the damage caused by bolt loosening is recorded both quantitatively and qualitatively. Then, using machine learning algorithms, the location of the damage (bolt loosening) in the flanged connection is identified, and the amount of preload force reduction relative to the initial state is quantitatively determined. The results showed that the trained neural network, using the time-of-flight index, is capable of detecting bolt loosening with 93% accuracy. Therefore, the results of this study demonstrate the high potential of using Lamb waves combined with artificial neural networks for the accurate and timely detection of bolt loosening.
