机械设计优化设计程序.docxVIP

优化设计与应用Matlab编程作业1、试用牛顿法求函数的极小点。解：牛顿法一维搜索函数文件minNewton如下：function [x,minf]=minNewton(f,x0,eps)%目标函数：f;%初始点：x0;%精度：eps;%目标函数取最小值时的自变量：x;%目标函数的最小值：minf;format longif nargin==2 eps=1.0e-6;enddf=diff(f); %求目标函数的一阶导数d2f=diff(df); %求目标函数的二阶导数k=0;tol=1;while toleps dfx=subs(df,findsym(df),x0); %目标函数的一阶导数值 d2fx=subs(d2f,findsym(d2f),x0); %目标函数的二阶导数值 x1=x0-dfx/d2fx; k=k+1; tol=abs(dfx); x0=x1;endx=x1;minf=subs(f,findsym(f),x);format short;命令行输入：syms t;f=t^4-4*t^3-6*t^2-16*t+4;[x,fx]=minNewton(f,3)M函数文件的运行结果如下：x =4fx = -1562、利用梯度法求目标函数的极小值，设初始点为，收敛精度。解：梯度法函数文件minFD如下：function [x,minf] = minFD(f,x0,var,eps)%目标函数:f;%初始点:x0;%自变量向量:var;%精度:eps;%目标函数取最小值时的自变量值:x;%目标函数的最小值:minf;format long;if nargin == 3 eps = 1.0e-6;endsyms l;tol = 1;gradf = - jacobian(f,var);while toleps v = subs(gradf,var,x0); tol = norm(v); y = x0 + l*v; yf = subs(f,var,y); [a,b] = minJT(yf,0,0.1); xm = minHJ(yf,a,b); %用黄金分割法进行一维搜索 x1 = x0 + xm*v; x0 = x1;endx = x1;minf = subs(subs(f,x(1)),x(2));format short;命令行输入：syms t s;f=t^2+s^2-t*s-10*t-4*s+60;[x,mf]=minFD(f,[0 0],[t s])M函数文件的运行结果如下：x = 8.00006.0000mf =8.00003、给定初始点，用阻尼牛顿法求函数的极小点。解：阻尼牛顿法函数文件minNT如下function [x,minf] = minMNT(f,x0,var,eps)format long;if nargin == 3 eps = 1.0e-6;endtol = 1;x0 = transpose(x0);syms l;gradf = jacobian(f,var);jacf = jacobian(gradf,var);while toleps v = subs(gradf,var,x0); tol = norm(v); pv = subs(jacf,var,x0); p = -inv(pv)*transpose(v); y = x0 + l*p; yf = subs(f,var,y); [a,b] = minJT(yf,0,0.1); %进退法求单峰区间 xm = minHJ(yf,a,b); %黄金分割法进行一维搜素 x1 = x0 + xm*p; x0 = x1;endx = x1;minf = subs(f,var,x);format short;命令行输入：syms t s z;f=(t-s+z)^2+(-t+s+z)^2+(t+s+z)^2[x,mf]=minMNT(f,[0.5 1 0.5],[t s z])M函数文件的运行结果如下：x = 1.0e-015 *-0.3468 -0.6936 -0.3468mf =2.4053e-0304、用内点法求目标函数受约束的最优解。解：内点惩罚函数法文件minNF如下：function[x,minf]=minNF(f,x0,g,u,v,var,eps)%目标函数：f；%初始点：x0；%约束函数：g；%罚因子：u；%缩小