<p>This paper introduces a deep learning-based surrogate model based on predicting the nonlinear vibration behaviors of porous nanocomposites plates reinforced with graphene platelets (GPLs) with central cutouts. Conventional numerical approaches can provide precision but are too slow, particularly when dealing with high-dimensional parametric analysis, or in real time. To overcome this shortcoming, an artificial neural network (ANN) surrogate model is formulated to effectively learn the complicated nonlinear interactions among design variables, such as plate geometry, GPL’s distribution patterns, reinforcement volume fractions, and porosity. Hamilton’s principle is used to obtain the motion equations of the reinforced porous plates, where von Kármán geometric nonlinearity and Mindlin plate theory are incorporated to include both bending and shear effects. The plate with a central cutout is divided into sub-domains to find vibration solutions whereby orthogonal polynomials are used to meet the geometric boundary conditions of each sub-domain. Lastly, the continuity conditions are imposed and the Rayleigh–Ritz and Newton–Raphson methods are used to calculate the linear and nonlinear frequencies. The model is also verified to agree with the finite element simulation in ABAQUS with the errors of less than 3%. It is shown that plates with bigger cutouts have lower linear and nonlinear frequency as well as higher frequency ratio than the plates with smaller cutouts. Enhancing the porosity results in 21% increase in nonlinear frequency ratio and reduces the linear frequency about 28%. Also, the nonlinear frequency sensitivity to GPL’s weight fraction is felt more at higher amplitudes, where its change is 25% at higher amplitude. About the deep learning analysis, the best model is obtained by a ANN model with 8 nodes in first hidden layer and 2 nodes in second hidden nodes.</p>