摘要
遗传算法在问题优化中的应用已有了许多研究,但对于大型多目标规划问题而言,由于其问题特性和计算量大而限制了遗传算法的应用。为探索新的问题求解方法,提出了一种基于遗传算法和梯度算法的问题优化混合算法。用梯度法每次迭代得到的结果来改进遗传算法的群体,而用遗传算法的最优个体与梯度算法的迭代解相比较,选择其中的最优点作为梯度法下一步迭代的初始点。通过保持迭代过程的最优解,加快了搜索速度,并保证收敛于全局最优解。算例表明该方法兼具遗传算法的全局搜索能力和梯度算法的局部搜索的特点,且具有良好的工程适应性。
There is a lot of research in genetic algorithm about structural optimization. But as far as the large multi-goal program concerned it here limited the application of genetic algorithm for the reason of its specialty and large calculation. In order to explore new resolution, the author proposed a combining algorithm for structural optimization, which is based on genetic algorithm and gradient algorithm: Use gradient algorithm to superpose, get a result, improve the herd of genetic algorithm with this result, then compare the superior one of genetic algorithm with the root of gradient algorithm, choose the best point to be the incipient point of the next step of super position. With this method, it can keep the best root of all the course, and also it can speed up searching, and keep the best global root. Numerical examples show that the combining algorithm possesses both the merit of genetic algorithm on strong global searching ability and gradient algorithm.
出处
《微电子学与计算机》
CSCD
北大核心
2004年第7期132-134,142,共4页
Microelectronics & Computer
基金
国家自然科学基金资助(60373062)
湖南省杰出中青年专家科技基金项目(02JJYB012)
关键词
遗传算法
梯度算法
混合算法
结构优化
Genetic algorithm, Gradient algorithm, Structural optimization
