This article presents global optimal solutions for a minimal distance problem between two non-convex surfaces. By canonical duality theory, this non-convex constrained minimization problem can be re-formulated as a concave maximization problem over a convex space, which can be solved by well-developed convex programming techniques. Application is given to the problem with one convex and one non-convex surface studied in Gao and Yang [Optimization 2008; 57: 705-714], which was challenged by Voisei and Za˘linescu [Optimization 2011; 60: 593-602]. Using the points of view presented in Voisei and Za˘linescu, we demonstrate that consideration of the Gao-Strang total complementary function and the canonical duality theory is indeed quite useful for solving a class of real-world global optimization problems with nonconvex constraints. Additionally, we demonstrate how a perturbed canonical dual method can be used to solve the counterexample presented by Voisei and Za˘linescu, which has multiple global minimum solutions.