An adaptive numerical method called, the adaptive random differential quadrature (ARDQ) method is presented in this paper. In the ARDQ method, the random differential quadrature (RDQ) method is coupled with a posteriori error estimator based on relative error norm in the displacement field. An error recovery technique, based on the least square averaging over the local interpolation domain, is proposed which improves the solution accuracy as the spacing, h → 0. In the adaptive refinement, a novel convex hull approach with the vectors cross product is proposed to ensure that the newly created nodes are always within the computational domain. The ARDQ method numerical accuracy is successfully evaluated by solving several 1D, 2D and irregular domain problems having locally high gradients. It is concluded from the convergence values that the ARDQ method coupled with error recovery technique can be effectively used to solve the locally high gradient initial and boundary value problems.