Architects always use a range of aesthetic elements, including fractionally graded materials (FGMs), at a high acoustic and thermal demand level. Numerous partial differential equations, especially wave equations, have previously been the focus of extensive analytical or numerical approaches. However, the application of Neural Networks guided by physics raises the standard for acoustic results. Using the novel paradigm outlined in that paper, metal-ceramic composites, for instance, demonstrate extremely effective wave behavior to demonstrate changes in stiffness and density, including radiation, scattering, and noise transmission. Several kinds of PINN can help precisely define the error when comparing square error, absolute error, and mean square error compared to finite element simulations. The MATLAB NEURAL simulation for neural network toolboxes were used to view the simulation in this research. The research revealed that the results were extremely accurate, with a maximum inaccuracy of 2.6%. Intending to improve the acoustic management of homogenous materials, this study correspondingly examines the impact of material gradient on reflection and sound insulation properties. This suggested strategy offers a highly motivated basis for resolving wave propagation, opening the door to far better soundproofing outcomes, noise management, and more effective building material design.