Foad Nazari - ۱Msc Student, Mechanical engineering department, Bu-Ali Sina University, Hamedan, Iran.
Hossein Goudarzvand Chegini - Msc Student, Mechanical engineering department, Islamic Azad University of Takestan, Takestan, Iran
Mohsen Behzadi۳ - ۳Msc Student, Mechanical engineering department, Bu-Ali Sina University, Hamedan, Iran.
Mahdi Karimi - Assistance Professor, Mechanical engineering department, Bu-Ali Sina University, Hamedan, Iran.

In this paper a method for crack detection in blades is presented. In the suggested method, the process of crack identification is consists of four stages. In first stage, three natural frequencies of a blade for different locations and depths of cracks were calculated using Finite Element Method (FEM). The obtained results were verified with the results of experimental modal analysis. In second stage, two Multi Layer Feed Forward (MLFF) neural networks were created. In third stage, Genetic Algorithm (GA) was used to training the neural network. The inputs of neural networks were the first three natural frequencies and the outputs of first and second neural networks were corresponding locations and depths of cracks, respectively. In forth stage, some of natural frequencies of blade with different crack situations as inputs applied to trained neural networks. Finally obtained results showed that predicted cracks characteristics were in good agreements with actual data

crack detection, genetic algorithm, finite element method, experimental modal analysis, blade