Minimizing the cost of structural design of RC corbels: A hybrid approach of optimizing pattern search technology and artificial intelligence

In the recent years, modeling of concrete properties based on optimization and prioritization methods has been performed using machine learning techniques including genetic programming, artificial neural networks (ANN), an adaptive neuro-fuzzy inference system, and support vector machines. However, the application of ANN to the task of optimizing the design of reinforced concrete (RC) corbels has been limited in previous studies. In this endeavor, two empirical mathematical models of ANN are developed to calculate the optimal reinforcement ratio and effective depth of RC corbel. More than 1,500 corbel data have been processed utilizing pattern search optimization and used in this computational artificial neural solution. Five model parameters were adopted to formulate simple ANN mathematical models. Parametric analysis was performed to justify the independent parameters of ANN models. This procedure showed the accuracy of ANN models, despite changes in independent parameters. Thus, these variables are successfully integrated into the neural models. The stability and simplicity of the proposed mathematical models of ANN make them suitable for optimizing the design of RC corbels, as these formulas are applicable to predict the optimal reinforcement ratio and edge impact depth, considering the cost factor.