Detection of Size and Location of Crack in Pipes Under Fluid Pressure by Neural Networks

In this research crack size and location in pipes under fluid pressure will be detected using pipe’s natural frequencies by neural network. Neural network used in this research is multi-layer perceptron. Comparing different inputs، appropriate inputs are selected. Pipes contain water. Steel and aluminum pipes were used in this research. Pressure condition of the pipes is: 1) without water 2) water with zero pressure 3) water with 0. 498 MPa 4) water with 0. 981 MPa. Crack size range from 0. 19043 to 0. 6346. Crack location range from 0. 199 to 0. 403. Many researches have been done about crack detection based on natural frequencies of structures by neural network. However، as far as authors know، no work has been done for crack detection in pipes containing pressurized water. Also in this paper two structures with different materials have been used for neural network training and testing which is another innovation of this research. Comparison of the results of this method with analytic methods shows that the proposed method is always more accurate in detecting crack size but is not always better in estimating crack location.