In this study, as the Artificial Neural Networks (ANN) is becoming increasingly common in the modeling and forecasting in many disciplines of engineering, Generalized Regression Neural Networks (GRNN) model is used to obtain natural frequencies of elastically supported Timoshenko beams with attached masses and the obtained frequencies are compared with the frequencies obtained from the exact solution of the same beam model. The elastically supported cantilever Timoshenko beam carrying multiple attached masses with rotary inertia is chosen as a mathematical model. The concept of fixity factor is used to define the stiffness of the elastic support. The governing equation of the beam elements is solved by applying the separation of variables method in the transfer matrix method (TMM) algorithm that gives the exact solution. The data used in the training, testing and verification phases of GRNN are obtained from TMM. The comparison graphs are presented for training, testing and verification data in numerical analysis to show the effectiveness of the GRNN, and it is resulted that neural network approach gives close results to TMM in training and testing phases.

Bülent Yõldõrõm

Papers

Comparing Exact and Generalized Regression Neural Network Solutions for Free Vibration of Elastically Supported Timoshenko Beams with Attached Masses