Dynamical problems governed by non-linear initial value problems (IVP), in general, are of great interest of the scientific community. The knowledge of the solution of these IVPs facilitates the understanding of the dynamic characteristics of the problem. However, unfortunately, many of the IVPs of interest does not present a known solution. In this case, an alternative is to calculate approximations for the solution. Numerical and analytical methods are efficient in this assignment and can provide approximations with the desired precision. Numerical methods have been developed over the last years and have been widely applied to dynamical problems in various engineering areas. Computational packages, easy to use, were created and today are part of the most traditional numerical simulation programs. However, numerical approximations have a disadvantage in relation to analytical approaches. They do not allow the understanding of how the solution depends on the problem parameters. Given this, this dissertation focuses on the analysis and implementation of analytical techniques called perturbation methods. The Lindstedt-Poincaré method and multiple time scales method were studied. The methodologies were applied in an IVP involving the non-damped Duffing equation. Symbolic algebra programs were developed with the purpose of calculating hierarchical analytical approximations to the solution of this problem. A parametric analysis was performed, in other words, a study of how the approximations are influenced by initial conditions and parameter values. In addition, the analytical approximations obtained were compared with numerical approximations calculated using the Runge-Kutta method. The multiple scales method was also applied in a IVP that represents the dynamics of a mass-spring-damper oscillator with dry friction. Due to friction, the system response can be characterized in two alternating phases, the stick phase and the slip phase, composing a phenomenon called stick-slip. It was verified that the approximations obtained for system response by the multiple scales method represent the stick-slip dynamics with good accuracy.