The field of nonlinear numerical simulations has garnered significant interest among researchers in engineering, physics, and mathematics, being essential for understanding complex phenomena in static and dynamic systems. In particular, space trusses are widely studied structures due to their practical relevance and the challenges related to nonlinear behavior. This work explores the application of Physics-Informed Neural Networks (PINNs) in the nonlinear dynamic analysis of these structures. The methodology employs a Multilayer Perceptron (MLP) neural network, combined with an exact vector formulation that incorporates parameters such as length, stress, area, strain, nodal positioning, and vector directions. The neural network is trained using a loss function based on force equilibrium and boundary conditions, employing the L-BFGS optimizer and the hyperbolic tangent (Tanh) activation function. The implementation, named FRSTrussNN, integrates the Finite Element Method (FEM) and utilizes the Newmark method for nonlinear dynamic analysis. The proposed solution was validated using classic examples from the literature and compared with commercial software (e.g., ANSYS), demonstrating that PINNs are a promising alternative to traditional methods, offering accuracy and efficiency in solving complex structural problems.