Application of Artificial Neural Networks to Detect Defects

This article introduces a new approach for troubleshooting gas transmission pipelines using artificial neural network with the help of mechanical waves, which is much cheaper and easier than the ultrasound method. who is currently working. These lines are usually located in harsh environmental conditions and far away and in long distances, and the use of systems that can instantly and accurately report the defects and leaks of this pipe is vital. The presented method includes modeling a piece of ۲-inch pipe with a length of ۵۰ meters is ۶,۱۲۱ in Abacus software. Then, the finite element model is obtained in the finite element software. Then, to confirm and validate the finite element model, the part was subjected to modal test and after confirming the model, the defects simulation, which includes creating ۱۵ holes with a radius of one millimeter at three meter intervals, was performed on the finite element model. Then by taking the vibrations (acceleration) of the pipe in the healthy state and the state with the defect and transferring the acceleration data to the frequency domain, then using the mechanical signature of the defects, the acceleration difference between the two healthy and defective models is calculated. In the next step Using statistical techniques, we reduce the obtained data optimally, and then Radial Basis Artificial Neural Network (RBFN) and Multilayer Perceptron Artificial Neural Network (MLP) are trained to estimate the location of the defect. The designed neural network shows the fault location well

Keywords:

Mechanics , Abacus , Finite Elements , Mechanical Signature , ANN Artificial Neural Network