Damage detection in gas transmission lines based on natural frequency changes via artificial neural networks

The purpose of this paper is to introduce a new approach for troubleshooting of gas pipelines using mechanical waveforms. In this paper, an attempt has been made to determine the defects by using natural frequency characteristics and changes. Due to the relationship between mass and stiffness in determining the natural frequency, a mass increase has been used instead of the reduction of stiffness. The proposed method involves modeling a 2-inch pipe with a length of 2 m in the Abaqus software and examining the natural frequencies of the pipe in a state of intact and defective (cracking). Then, an equivalent mass value was applied to model the defects. Subsequently, the pipe was subjected to experimental modal analysis and various conditions were tested by creating artificial defects (adding mass). Then, the obtained information was used in the study of the perceptron neural network in MATLAB software, and the program output is the defect location and its severity (mass value).