We have proposed "Draining Method (DM)" in [5,6]. DM is based on the Discrete Gradient Chaos Model (DGCM) and the objective function transformation which is developed by the analysis of DGCM. Applying DM to typical benchmark problems, we have confirmed its superior global optimization capability. However, as Liang et al pointed out in , typical benchmark problems, such as Rastrigin function, have several considerable problems. Besides, DM has a problem that we need to set Objective Function Value (OFV) of global minima (or desired value) at the start of the search. In this study, we propose to improve draining procedure so that OFV of the global minimum is not needed. Then, we apply the improved DM to more proper benchmark problems which are created by recommended methods in . Through several numerical simulations, we confirm that improved DM is generally effective for proper benchmark problems. This result suggests that improved DM is effective in general situations.