Analysis of the prediction of a bilayered cylindrical shell’s reduced cutoff frequency with data-driven approaches

Abstract The reduced cutoff frequency is one of the most important properties of elastic circumferential waves of cylindrical shells. For homogeneous thin cylindrical shells, this frequency can be calculated based on the eigen modes theory. In contrast, this theory is not practical for inhomogeneous cylindrical shells. This paper presented the methods of reduced cutoff frequency prediction using data-driven approaches. Three widely known data-driven methods, the Artificial Neural Network (ANN), the Adaptive-Network-based Fuzzy Inference System (ANFIS), the Support Vector Machine (SVM), are employed to develop the prediction models. The training and testing datasets for these models are determined from the plane of modal identification. In this study, the considered bilayered cylindrical shells are made of stainless steel/polymer. Three prediction scenarios are assessed. The difference between these scenarios is the physical and geometrical parameters which considered as relevant entries of data-driven approaches. In the first and second scenarios, a comparative analysis indicates that the ANN models performed better than other models with highest prediction accuracy value of 0.9761 and the lowest mean relative error (MRE) of 0.4017%. In the third scenario the SVM model was the best model with highest accuracy of 0.9895 and lowest root mean square error (RMSE) of 0.0227.