Thermal Stress Analysis in Two-Directional Functionally Graded Plates with Artificial Neural Network Training Algorithms

It is very important to determine the volumetric distribution because Functionally Graded Materials (FGMs) are not mass-produced at present. By determining the volumetric distribution, the allowable stress limits of the material are also determined. FGMs are used in high temperature effects so thermal stress limits become very important. For the calculation of thermal stress distribution and levels, numerical analysis methods are used. In this study, the models based on volumetric distribution for the equivalent stress levels, which are the most important parameter for the determination of the thermo-mechanical behavior of the 2D-FG plates, were formed by using the finite difference method (FDM). These models were obtained by three different training algorithms in artificial neural network (ANN). In this study, the models which can determine the equivalent stress level of different composition gradient exponent values for the existing conditions of the 2D-FGs are presented. These models provide 340 times faster solution than the FDM. Suggested models will be the guide to reach the optimum volumetric distribution in both production and theoretical studies of FGMs which cannot be mass-produced yet. In the study, the training stages, performance values and functionality of different training algorithms are examined and interpreted in detail