论文部分内容阅读
在用数字计算机设计最佳微波网络方面,新近出现的自动的优化设计方法已获得应用,或即将获得应用。本文的目的就是对此作一评述。重点放在公式及用经典综合法(解析式或数值式)所不能胜任的方法上。网络最佳化的目标用公式表示,包括极值法公式与最小P次法公式。还详细地研究了通过变换或惩罚去处理参数和响应约束的方法,特别讨论了不等式约束问题的公式与用贯序无约束极小化的方法求它们的解。以讲座的方式概述了几个一维和多维的极小化策略。这些策略是费班纳赛(Fibonacci)法、黄金分割搜索法、插值法、样步搜索法、罗森布罗克法、鲍威尔法、单纯形法、以及牛顿—拉夫森法、弗莱彻—鲍威尔法和最小二乘法。还介绍了微波线路设计师感兴趣的一些有关例子来说明计算机辅助设计的最优化技术的应用情况,还有分门别类的参考文献。
The recent emergence of automated optimization methods has been or will soon be applied to the design of the best microwave networks with digital computers. The purpose of this article is to comment on this. Emphasis is placed on formulas and methods that can not be performed using classical synthesis (analytic or numerical). The goal of network optimization is expressed in terms of formulas, including the formula for the extreme value method and the formula for the smallest P order. We also study in detail the method of dealing with parameters and response constraints through transformation or punishment. In particular, we discuss the formulas of inequality constraints and their solutions with unconstrained minimization. In a lecture, several one-dimensional and multi-dimensional minimization strategies are outlined. These strategies are Fibonacci, Golden Section, Interpolation, Step-by-Step, Rosenborg, Powell, Simplex, and Newton-Raphson, Fletcher-Powell And least squares method. Also introduced some examples of microwave line designers interested in illustrating the application of computer-aided design optimization techniques, as well as separate references.