Using GMDH-type neural network model to predict the response of triangular plates under the hydrodynamic loading

The aim of this study is investigating the center of mass deflection to thickness ratio of triangular plates using the GMDH-type neural networks and comparing it with results of laboratory tests performed on a narrow triangular plate using water-hammer apparatus. Also, the study focuses on the overall deformation, strain and impact transmission. Dimensionless input variables are used to investigate the center of mass deflection of triangular plate with changing variables. A simpler polynomial expression is derived using GMDH-type neural network and dimensionless number. The vector of coefficients of quadratic sub-expressions involved in GMDH-type networks is obtained by Singular Value Decomposition (SVD) method. SVD can improve the proficiency of GMDH-type networks to model the intricate process of deformation of triangular plates. Obtaining results by applying a GMDH model and comparing them with actual data indicates good agreement between model output and experimental data. The advantages of this approach are in the simplification of computation and convenient application to parametric study for impact behavior.