Abstract: An improved heuristic method for Optimum searching of mixed discrete variable problems,i.e.a mixed discrete expanded complex programming method(MDECP)is presented.The definition of‘effective function,in literature is modified by normalization and establishing a buffer region for equality constrained problems.The variable interval discrete technigue in engineering problems are discussed,thus broadening the scope of design.Furthermore,a new principle for check is proposed for a convergence criterion,which uses the maximum mixed sub-gradient defined in this paper rather than the sub-gradient of objective function to seek a local optimum solution.Calculation of a real cases has shown that the number of point-to-point check is greatly decreased and optimum searching is more effective.Finally,an optimum design of chain drive used in oil fields with discrete parameter is made in this paper.

摘要： 目前,本文基于现有混合离散变量优化方法的基本思想,提出了一种改进的混合离散变量问题的试探性寻优方法,即MDECP法.该法首先对文献[2]的“有效函数”定义作了一定的改进,即进行了规一化处理和对等式约束问题建立了“缓冲区”;同时,就工程上变均匀间隔离散问题进行了讨论,从而扩大了设计范围,丰富了优化内容.其次,本文还以提出的一种新的查点原理作为收敛准则.该查点原理不是采用一般资料所介绍的利用目标函数值的次梯度信息进行查点的方法,求解局部最优解,而是利用本文所定义的“最大混合次梯度”信息进行查点寻求局部最优解.经过实例计算表明:该法大大地减少了查点数目、提高了寻优效率.最后,还对具有离散性参数的石油链传动进行了优化设计.

关键词: 混合离散变量, 链传动, 局部最优解, 目标函数, 寻优方法, 次梯度, 有效函数, 基本思想, 梯度信息, 混合型