Details

Autor(en) / Beteiligte

• Abdullah, Aidie Zeid Muhammad
• Norman, Mohd Arif Mat
• Azman, Nurshahirah
• Idris, Zahirrudin
• Azman, Hazeeq Hazwan
• Roslan, Siti Aisyah
• Narowi, Mumtazah
• Maniyam, Maegala Nallapan

Titel

Predicting the crack location and crack depth of steel rail due to vibration using artificial neural networks (ANN)

Ist Teil von

• AIP Conference Proceedings, 2023, Vol.2625 (1)

Ort / Verlag

Melville: American Institute of Physics

Erscheinungsjahr

2023

Beschreibungen/Notizen

• Rail crack is the most threatening defect in railway industries. The crack can cause the train to derail. Using artificial neural networks (ANN), the paper aims to predict the crack location and depth in a rail track panel under clamp boundary conditions using artificial neural networks (ANN). In this study, the model of rail track was used based on 60E1(European standard EN 13674-1) and has been modelled as a reliable structural using CATIA software. The free vibrations analysis of the rail track with the various crack models had performed using MODAL analysis in ANSYS software. The finite element analysis can observe that the modes shape of vibrations had different values with the various crack models. The results obtained were used as data for training, validation, and test for the ANN model. The acquired results were used for the ANN model's train, validation, and test phases. Cascade-forward Back Propagation (CFBP) was used in the ANN models to predict rail track cracks). furthermore, the regression analysis determination coefficient was used to test the validity of the ANN model (R2). ANN performance indeed found the R2 (coefficient of determination) values both for train and test data to be 0.998 and 0.998 accordingly. The ANN predicted crack prediction analyzed outcomes to the ANSYS simulated results to measure the ANN model's accuracy in predicting crack lengths and locations. The result can observe that the predicted results were in excellent agreement with the finite element analysis results (FEA). In general, this study is beneficial and adds significant knowledge to predict the location and depth of crack using the ANN model.