Harmonious decomposition algorithm and its application in global optimization of pump configurations
调速泵结构配置协调分解优化算法及实现

A two-stage model, which consists of nonlinear continuous programming and nonlinear integer programming sub-problems, is proposed for solving the multi-level pumps configuration problem. The integer nonlinear programming subproblem is solved by using a traditional outer approximation algorithm; the continuous nonlinear programming mainproblem is optimized using a novel harmonious decomposition algorithm based on the cutting angle method (CAM). Some properties of cutting angle method (CAM) are studied and the ill-posed trap limitation is found and proved theoretically. In view of the deficiencies of CAM, we propose a harmonious decomposition algorithm using judgement criterions to exclude the known trap regions in feasible domain; and then, compute the gradually tightening lower bounds on the main optimization problem by building a sequence of piecewise linear relaxation problems. The sequence of solutions to these relaxation problems converges to the global minimum of the optimization problem. Experiment results from a 3-level pump configuration instance show that the proposed method provides better performances than the other optimization methods listed in paper.