This paper presents a fault diagnosis strategy for actuators faults in an octorotor unmanned aerial vehicle (UAV) using sliding modes techniques. First, system states are estimated using a second order sliding-mode observer (SOSMO) based on the modified super-twisting algorithm. Second, faults are estimated by considering the system control as an unknown input. After the convergence of the observer, the equivalent output injection is used to estimate the unknown input and thus to identify losses in the actuators. The effectiveness of this approach is illustrated by numerical simulations on Matlab/Simulink, and also a real experimental application on a coaxial octorotor UAV.