Analyzing Concrete Cracks' Characteristics Using Meta-heuristic Computing

Monitoring cracks is vital for successful inspection and maintenance management of bridge infrastructure. Thus, this research introduces an artificial intelligence-based model that can address the limitations of visual inspection models. The developed model is conceptualized on two basic modules. In the first module, Renyi's entropy function is coupled with shuffled frog-leaping algorithm for the efficient detection of the boundaries and features of cracks. In the second one, artificial neural network is hybridized with shuffled frog leaping algorithm (ANN-SFL) for the purpose of analyzing the level of severities of cracks from length and width perspectives. In this regard, shuffled frog leaping algorithm is exploited to improve the prediction performance of artificial neural network through optimizing the weights of connections between neurons. The developed model is tested through its comparison against widely implemented machine learning models, namely back propagation artificial neural network (BPANN), generalized regression neural network (GRNN) and radial basis function network (RBFN). It was manifested the developed ANN-SFL model explicitly performed better than the remainder of the artificial intelligence models for both length and width. The developed model ANN-SFL ameliorated the prediction performances of BPANN, GRNN and RBFN by 81.98%, 79.11% and 77.97%, respectively. It is expected that the developed ANN-SFL model can contribute efficiently in conserving bridges' condition and establishing preemptive maintenance plans.