Crack prediction in pipeline using ANN-PSO based on numerical and experimental modal analysis

In this paper, a crack identification using Artificial Neural Network (ANN) is investigated to predict the crack depth in pipeline structure based on modal analysis technique using Finite Element Method (FEM). In various fields, ANN has become one of the most effective instruments using computational intelligence techniques to solve complex problems. This paper uses Particle Swarm Optimization (PSO) to enhance ANN training parameters (bias and weight) by minimizing the difference between actual and desired outputs and then using these parameters to generate the network. The convergence study during the process proves the advantage of using PSO based on two selected parameters. The data are collected from FEM based on different crack depths and locations. The provided technique is validated after collecting the data from experimental modal analysis. To study the effectiveness of ANN-PSO, different hidden layers values are considered to study the sensitivity of the predicted crack depth. The results demonstrate that ANN combined with PSO (ANN-PSO) is accurate and requires a lower computational time in terms of crack identification based on inverse problem.