Numerical investigation of an iterative procedure, called the control averaging technique, is performed in, the framework of three illustrative singular optimal control problems. Effect of various relevant parameters such as penalty constants and initial guesses on the convergence behaviour of this method is examined. For each ease considered, the control averaging iterates are found to give a close approximation to the optimal solution in relatively small computing time. Furthermore, from the viewpoint of computing time the control averaging technique is found to be about six to eight times more efficient than the gradient-like minHmethod.