Detection of Cracks in Adhesively Bonded Double-Strap Joints Using Artificial Neural Network Method

This study aims to survey the effects of crack length on the natural frequencies and mode shapes of adhesively bonded double-strap joints (DSJs) in the models with different adherend thicknesses and different adhesive ductility. Hence, the results of these analyses are used as a method of crack detection inside the adhesive layer. For this purpose, a rich database of results for DSJ models with cracks in length of 0 ≤ l _C ≤ 10 mm were used for the aim of training the artificial neural network (ANN) algorithms. Subsequently, the results obtained from ANN models can be used to estimate the existence of crack and its length. The results show that the natural frequencies of the models with different adherend thicknesses and different adhesive ductility follow a logical trend, by which detection of cracks is possible for a wide range of geometries and material properties, using ANN analysis. By contrast, the results show that mode shapes are not affected by cracks. Therefore, the mode shapes are not useful characteristics for judgments.