【高等数学基础进阶】多元函数的极值与最值

74 阅读3分钟

本文已参与「新人创作礼」活动,一起开启掘金创作之路。

无约束极值

定义:若在点(x0,y0)(x_{0},y_{0})的某邻域内恒成立不等式

f(x,y)f(x0,y0)(f(x,y)f(x0,y0)) f(x,y)\leq f(x_{0},y_{0})\quad (f(x,y)\geq f(x_{0},y_{0}))

则称ff在点(x0,y0)(x_{0},y_{0})取得极大值(极小值),点(x0,y0)(x_{0},y_{0})称为ff的极大值点(极小值点),极大值与极小值统称为极值,极大值点与极小值点统称为极值点

 

定理(极值的必要条件):设z=f(x,y)z=f(x,y)在点(x0,y0)(x_{0},y_{0})存在偏导数,且(x0,y0)(x_{0},y_{0})f(x,y)f(x,y)的极值点,则

fx(x0,y0)=0,fy(x0,y0)=0 f_{x}(x_{0},y_{0})=0,f_{y}(x_{0},y_{0})=0

 

也可表述为

f(x,y0)f(x,y_{0})x=x0x=x_{0}处的导数等于00f(x0,y)f(x_{0},y)y=y0y=y_{0}处的导数等于00

 

极值点只可能在驻点和fx,fyf_{x},f_{y}至少有一个不存在的点

 

定理(极值的充分条件):设z=f(x,y)z=f(x,y)在点P(x0,y0)P(x_{0},y_{0})的某邻域内有二阶连续偏导数,又fx(x0,y0)=fy(x0,y0)=0f_{x}(x_{0},y_{0})=f_{y}(x_{0},y_{0})=0,记A=fxx(x0,y0),B=fxy(x0,y0),C=fyy(x0,y0)A=f_{xx}(x_{0},y_{0}),B=f_{xy}(x_{0},y_{0}),C=f_{yy}(x_{0},y_{0}),则

  • ACB2>0AC-B^{2}>0时,有极值,若A>0A>0为极小值,若A<0A<0为极大值

  • ACB2<0AC-B^{2}<0时,无极值

  • ACB2=0AC-B^{2}=0时,不一定(一般用定义判定)

 

条件极值与拉格朗日乘数法

函数f(x,y)f(x,y)在条件ϕ(x,y)=0\phi (x,y)=0条件下的极值

F(x,y,λ)=f(x,y)+λϕ(x,y)F(x,y,\lambda)=f(x,y)+\lambda \phi (x,y)

{Fx=f(x,y)+λϕx(x,y)=0Fy=fy(x,y)+λϕy(x,y)=0Fλ=ϕ(x,y)=0 \left\{\begin{aligned}&F_{x}=f(x,y)+\lambda \phi_{x}(x,y)=0\\ &F_{y}=f_{y}(x,y)+\lambda \phi_{y}(x,y)=0\\ &F_{\lambda}=\phi(x,y)=0\end{aligned}\right.

 

函数f(x,y,z)f(x,y,z)在条件ϕ(x,y,z)=0,ψ(x,y,z)=0\phi(x,y,z)=0,\psi (x,y,z)=0条件下的条件极值

F(x,y,z,λ,μ)=f(x,y,z)+λϕ(x,y,z)+μψ(x,y,z)F(x,y,z,\lambda,\mu)=f(x,y,z)+\lambda \phi(x,y,z)+\mu \psi (x,y,z)

 

设给定目标函数f(x,y)f(x,y),约束条件为ϕ(x,y)=0\phi(x,y)=0

![[附件/ae51f3deb48f8c542b093a4c30292df5e1fe7fff.webp]]

如图所示,曲线LL为约束条件ϕ(x,y)=0\phi(x,y)=0f(x,y)=Cf(x,y)=C为目标函数的等值线族

f(x,y),ϕ(x,y)f(x,y),\phi(x,y)偏导数都连续的条件下,目标函数f(x,y)f(x,y)在约束条件ϕ(x,y)=0\phi(x,y)=0下的可能极值点M(x0,y0)M(x_0,y_0),从几何上看,必是目标函数等值线曲线族中与约束条件相切的那个切点

因为两曲线在切点处必有公切线,所以目标函数等值线在点M(x0,y0)M(x_0,y_0)处法向量{fx(x0,y0),fy(x0,y0)}\{f'_x(x_0,y_0),f'_y(x_0,y_0)\}与约束条件曲线在点M(x0,y0)M(x_0,y_0)处法向量{ϕx(x0,y0),ϕy(x0,y0)}\{\phi'_x(x_0,y_0),\phi'_y(x_0,y_0)\}平行,即

fx(x0,y0)ϕx(x0,y0)=fy(x0,y0)ϕy(x0,y0)\begin{aligned}\frac{f'_x(x_0,y_0)}{\phi'_x(x_0,y_0)}=\frac{f'_y(x_0,y_0)}{\phi'_y(x_0,y_0)}\end{aligned}

也就是说存在实数λ\lambda,使下式成立

{fx(x0,y0),fy(x0,y0)}+λ{ϕx(x0,y0),ϕy(x0,y0)}=0\{f'_x(x_0,y_0),f'_y(x_0,y_0)\}+\lambda\{\phi'_x(x_0,y_0),\phi'_y(x_0,y_0)\}=0

需要注意的是,目标函数等值线与约束条件曲线的切点未必就是目标函数f(x,y)f(x,y)在约束条件ϕ(x,y)=0\phi(x,y)=0下的极值点(如图中的M2M_2点)

链接:拉格朗日乘数法_百度百科 (baidu.com)

最大最小值

求连续函数f(x,y)f(x,y)在有界闭域DD上的最大最小值

  1. f(x,y)f(x,y)DD内部可能的极值点(无约束极值)

  2. f(x,y)f(x,y)DD的边界的最大最小值(条件极值)

  3. 比较

 

应用题

  1. 建立函数关系

  2. f(x,y)f(x,y)DD内部可能的极值点(无约束极值)

  3. f(x,y)f(x,y)DD的边界的最大最小值(条件极值)

  4. 比较

 

常考题型方法与技巧

求极值(无条件)

例1:设函数z=f(x,y)z=f(x,y)的全微分为dz=xdx+ydydz=xdx+ydy,证明点(0,0)(0,0)f(x,y)f(x,y)的极小值点

 

可以用极值的充分条件,这里不展示过程。这里展示偏积分的方法得到原函数

 

函数z=f(x,y)z=f(x,y)的全微分为dz=xdx+ydydz=xdx+ydy,可知

zx=x,zy=y z_{x}=x,z_{y}=y

zx=xz_{x}=x,偏积分得

z=xdx=12x2+ϕ(y) z=\int_{}^{}xdx=\frac{1}{2}x^{2}+\phi(y)

再代入zy=yz_{y}=y,确定ϕ(y)\phi(y)

zy=ϕ(y)ϕ(y)=yϕ(y)dy=ydyϕ(y)=12y2+Cz=12x2+12y2+C \begin{aligned} z_{y}&=\phi'(y)\\ \Rightarrow \phi'(y)&=y\\ \int_{}^{}\phi'(y)dy&=\int_{}^{}ydy\\ \phi(y)&=\frac{1}{2}y^{2}+C\\ z&=\frac{1}{2}x^{2}+ \frac{1}{2}y^{2}+C \end{aligned}

根据定义后面不再展示过程

 

还可以通过凑微分得到原函数

 

dz=xdx+ydydz=d(12x2)+d(12y2)dz=d(12x2+12y2)z=12x2+12y2+C \begin{aligned} dz&=xdx+ydy\\ dz&=d (\frac{1}{2}x^{2})+d (\frac{1}{2}y^{2})\\ dz&=d(\frac{1}{2}x^{2}+ \frac{1}{2}y^{2})\\ z&= \frac{1}{2}x^{2}+ \frac{1}{2}y^{2}+C \end{aligned}

根据定义后面不再展示过程

 

求最大最小值

例2:求函数u=x2+y2+z2u=x^{2}+y^{2}+z^{2}在约束条件下z=x2+y2z=x^{2}+y^{2}x+y+z=4x+y+z=4下的最大值和最小值

 

F(x,y,z,λ,μ)=x2+y2+z2+λ(x2+y2z)+μ(x+y+z4) F(x,y,z,\lambda,\mu)=x^{2}+y^{2}+z^{2}+\lambda(x^{2}+y^{2}-z)+\mu(x+y+z-4)

{Fx=2x+2λx+μ=0Fy=2y+2λy+μ=0Fz=2zλ+μ=0Fλ=x2+y2z=0Fμ=x+y+z4=0 \left\{\begin{aligned} &F_{x}=2x+2\lambda x+\mu=0\\ &F_{y}=2y+2\lambda y+\mu=0\\ &F_{z}=2z-\lambda+\mu=0\\ &F_\lambda=x^{2}+y^{2}-z=0\\ &F_{\mu}=x+y+z-4=0\end{aligned}\right.

解得(x1,y1,z1)=(1,1,2),(x2,y2,z2)=(2,2,8)(x_{1},y_{1},z_{1})=(1,1,2),(x_{2},y_{2},z_{2})=(-2,-2,8)

故所求的最大值为7272,最小值为66

 

例3:已知z=f(x,y)z=f(x,y)的全微分dz=2xdx2ydydz=2xdx-2ydyf(1,1)=2f(1,1)=2。求f(x,y)f(x,y)D={(x,y)x2+y241}\begin{aligned} D=\left\{(x,y)\Big|_{}^{}x^{2}+ \frac{y^{2}}{4} \leq 1\right\}\end{aligned}上的最大最小值

 

先找f(x,y)f(x,y),这里用凑微分(偏积分也行)

dz=2xdx2ydydz=dx2dy2dz=d(x2y2)z=x2y2+C代入f(1,1)=22=11+CC=2z=x2y2+2 \begin{aligned} dz&=2xdx-2ydy\\ dz&=dx^{2}-dy^{2}\\ dz&=d(x^{2}-y^{2})\\ z&=x^{2}-y^{2}+C\\ &代入f(1,1)=2\\ 2&=1-1+C \Rightarrow C=2\\ z&=x^{2}-y^{2}+2 \end{aligned}

fx=2x=0,fy=2y=0\begin{aligned} \frac{\partial f}{\partial x}=2x=0,\frac{\partial f}{\partial y}=-2y=0\end{aligned},得驻点为(0,0)(0,0)

 

接下来可以用拉格朗日乘数法,运算不难,这里不展示步骤。考虑另一个思路,由于已经知道了约束条件,该约束条件可以代入z=f(x,y)z=f(x,y),化条件为无条件

 

由于点在x2+y24=1x^{2}+ \frac{y^{2}}{4}=1上,有

z=x2(44x2)+2x[1,1]z=5x22zmax=zx=±1=3zmin=zx=0=2 \begin{aligned} z&=x^{2}-(4-4x^{2})+2\quad x \in [-1,1]\\ z&=5x^{2}-2\\ z_{\max}&=z \Big|_{x=\pm 1}^{}=3\\ z_{\min}&=z \Big|_{x=0}^{}=-2 \end{aligned}

因此最大值为33,最小值为2-2

 

对于圆和椭圆可以用参数方程化条件为无条件

 

椭圆x2+y24=1\begin{aligned} x^{2}+ \frac{y^{2}}{4}=1\end{aligned}的参数方程为

{x=costy=2sint \left\{\begin{aligned}&x= \cos t\\&y=2 \sin t\end{aligned}\right.

z=f(x,y)=x2y2+2=cos2t4sin2t+2=35sin2tt[0,2π] \begin{aligned} z=f(x,y)&=x^{2}-y^{2}+2\\ &=\cos ^{2}t-4\sin ^{2}t+2\\ &=3-5\sin ^{2}t \quad t \in [0,2\pi] \end{aligned}