This work is devoted to the design of interval observers for a class of Linear-Parameter-Varying (LPV) systems. Applying High Order Sliding Mode (HOSM) techniques it is possible to decrease the initial level of uncertainty in the system, which leads to improvement of set-membership estimates generated by an interval observer and enlarges the class of the estimated systems. Next, this approach is applied to fault detection by verifying the consistency between the output trajectory and its estimated domain. The efficiency of the proposed approach is demonstrated through computer simulations of an electromechanical system.