Crack identification in laminated composites based on modal responses using metaheuristics, artificial neural networks and response surface method: a comparative study

Structures that are difficult to maintain and access need to have an efficient and robust process for continuous monitoring. Such monitoring through damage detection and identification studies is present in several engineering applications as it allows that corrective measures be applied in order to guarantee the structural safety of a given structure, machine or equipment. In particular, laminated composite materials, often used in aeronautical structures, have a complex failure mechanism where delamination or cracks in these materials are often not visible on the surface. Thus, the use of optimization methods for the characterization of damages in these materials becomes relevant. In this study, both the metaheuristic sunflower optimization, the artificial neural networks and the response surface method were used to solve an inverse crack identification problem. The crack was modeled as a thin elliptical hole in a rectangular laminated plate numerically modeled using the finite element method. As a result of the methods used, different approaches to the problem were obtained that present reliable shape, size and position identification of a crack sized between 3 and 30 mm. The results showed substantial and promising results in the uses of both metaheuristic techniques and artificial neural networks. However, neural networks have a certain competitive advantage over optimization techniques as long as the data that feeds the model present a certain level of quantity and quality. Results obtained were able to identify all damage parameters (location, extension and orientation), with errors less than 1%.