One-Dimension Blind-Walking Optimization Method

• 1、China University of Petroleum (East China)

Abstract：The theory and the algorithm of the one-dimension blind-walking optimal method are proposed. Its mathematic realization is halving or doubling detect-step and verifying the applicability condition. Double-step is propitious to making current point near to extremum. Half-step is available to accelerate convergence. In order to improve the optimization, the applicability condition is the precondition of updating current point. The algorithmic flow chart and the characteristic analysis of the method are given. Two optimal problems with single-apices and non-single-apices object function are solved by the proposed method and the ordinary method. Its astringency and applicability are validated. The simulation result shows that the proposed method is better than the ordinary method. The proposed method has the characteristic of rapid convergence speed, little calculation capacity, wide applicable range, etc. If this method is used as teaching material, it has the characteristic of visualizing teaching method, just as renaming the punish function method as wall-hillock method.

Keywords： Mechanism optimization design One-dimension blind-walking optimization method Optimization method Teaching