A benchmark problem for comparison of vibration-based crack identification methods

The vibration-based crack identification problem insists of finding a measured vibration parameter from a complete crack-detection-database constructed by numerical simulation. It is one of the classical optimization problems. Many intelligence methods, such as neural network (NN), genetic algorithm (GA), determinant transformation (DT), and frequency contour (FC) etc., have been extensively employed as optimization tools to achieve this task. The aim of this paper is to propose a benchmark problem to compare these extensive-used optimization methods in terms of crack identification precision and computational time. The merit and demerits for each method are discussed. The results suggest that FC is a visualized, stable and easily applied method for detecting crack in practice. The conclusions of current studies are useful to investigators in deciding which method should be chosen in their crack inspections.