Static, dynamic and stability analysis of multi-dimensional functional graded plate with variable thickness using deep neural network
This work addresses structural analysis challenges for engineers dealing with complex plate designs, though it is incremental as it applies existing DNN methods to a specific domain.
The paper tackled the analysis of multi-directional functionally graded plates with variable thickness by predicting central deflection, natural frequency, and critical buckling load, using a deep neural network that demonstrated effectiveness in handling extensive engineering data.
The goal of this paper is to analyze and predict the central deflection, natural frequency, and critical buckling load of the multi-directional functionally graded (FG) plate with variable thickness resting on an elastic Winkler foundation. First, the mathematical models of the static and eigenproblems are formulated in great detail. The FG material properties are assumed to vary smoothly and continuously throughout three directions of the plate according to a Mori-Tanaka micromechanics model distribution of volume fraction of constituents. Then, finite element analysis (FEA) with mixed interpolation of tensorial components of 4-nodes (MITC4) is implemented in order to eliminate theoretically a shear locking phenomenon existing. Next, influences of the variable thickness functions (uniform, non-uniform linear, and non-uniform non-linear), material properties, length-to-thickness ratio, boundary conditions, and elastic parameters on the plate response are investigated and discussed in detail through several numerical examples. Finally, a deep neural network (DNN) technique using batch normalization (BN) is learned to predict the non-dimensional values of multi-directional FG plates. The DNN model also shows that it is a powerful technique capable of handling an extensive database and different vital parameters in engineering applications.