s=tf('s')
转换laplace
- find state response
X=zpk(minreal(inv(s*eye(x的维度=N)-A)*(B*U+x0)))
A,B,X0,U known
[num_X1,den_X1]=tfdata(X(1),'v')
[r1,p1]=residue(num_X1,den_X1)
这里r1 和 p1 是对应的,之后x(1)。。。分别求出
- find output response `
Y=zpk(minreal(C*inv(s*eye(2)-A)*(B*U+x0)))
[num_Y,den_Y]=tfdata(Y,'v')
[r,p]=residue(num_Y, den_Y)
find magnitude and phase
>> M=abs(r(1)), 2*M
>> phi=angle(r(1))
y(t) = [(2Me^-3t)cos(2t+0.58)−0.15]ε(t)
y(t)=2\*abs(r(1))exp(real(p(1)))\*cos((abs(imag(p(1)))\*t+angle(r(1)))+r(3))
residue的结果是根据指数从小到大排列的
stability 研究state matrix的特征值情况
1.internally stable
所有特征值的实数部分<=0
而且geometric multiplicity of eigenvalues is 1 的特征值实数部分为0
if the zero input state response xzi(t) is bounded for any initial state x0
2.asymptotically stable
if the zero input response xzi(t) converges to 0, as t ∞, for any initial state x0
所有特征值的实数部分<0
3.unstable
if the zero input state response xzi(t) is unbounded
特征值全部>0 ,而且小于零的 geometric multiplicity大于1
BIBO stable
An LTI system is BIBO stable if and only if all the poles of the transfer function have strictly negative real parts
minimal
Are such that ρ(MC) = ρ(MO) = n, where n = dim(x )
M_C=ctrb(A,B)
r=rank(M_C)
M_O=obsv(A,C)
o=rank(M_O)
if (r==0 && r==dim(x) ) ,那么这个LTI sys(A,B,C,D) 就是minimal
relations between internal stable and BIBO stable
internal stability.Asimptotically stable====>>>BIBO stability.stable
Examples
matlab DCGAIN
H=1/((s+2)*(s+10));
transfer func,K = dcgain(H)
and k=0.05
The statement dcgain can also be used to compute the generalized steady-state gain (generalized dc-gain) K = lim s r H (s )
s=tf('s');
>> H=1/(s*(s+2)^2);
>> K = dcgain(s*H)
sinusoid input
s=tf('s');
>> H=1/((s+2)*(s+10)); >> w0=0.5;
>> [m,f]=bode(H,w0)
The magnitude value is not expressed in dB, but in linear scale, the phase value is expressed in degrees
transfer function with input delay
H=1/(s^2+s+1);
>> H.inputdelay=2;
Padé approximation
s=tf('s');
>> H=1/(s^2+s+1);
>> H.inputdelay=2;
>> SYSX = zpk(PADE(H,1))
bode figure
H=1/(s^2+3*s+2)
figure, bode(H)
polar diagrams
[re,im]=nyquist(H)
figure, plot(squeeze(re), squeeze(im))
nyquist
[re,im]=nyquist(H);
figure, plot(squeeze(re)
nichols
figure, nichols(H)
feedback
串联相乘,并联加减
反馈电路实例
控制电路
结构和命名
符号 | 结构名称 |
---|---|
G_p(s) | plant transfer function |
C (s ) | controller tf |
A | actuator gain |
G_s | Sensor gain |
G_y | Conditioning filter gain |
r : reference signal | desired behavior of the controlled output |
y | (controlled) output |
e=r -y | tracking error |
u | control input (command signal) |
d'_a | input disturbance |
d'_y | output disturbance |
G_d (s) | tf between d’y and y |
d_s | sensor noise |
d_a = d’_a / A | actuator disturbance |
G_d (s) =1 | d’_y = d_y |
d_t =d_s | Gy |
G (s ) = Gp (s ) A | - |
Td=GyGs | - |
简化结构图
loop function
loop function is defined as
L(s)= G(s)C(s)
sensitivity function
is define as
S(s)=1/(1+L(s))=y(s)/d_y(s)
complementary sensitivity function
control sensitivity function
actuator disturbance sensitivity function
the steady state design
e^inf_r,根据t前面的系数确定P的值
y^inf_d_y ,根据系数确定derta_y
例题
解题过程
steady state下的Css=Kc/s
先确定h和g的值,h就是r(t)中t的指数,
1.根据r(t)在e^inf_r表格中确定公式
2.根据K=lim s^2 L(s)=Kg*Kc,Kg=G(s)所有零点(s->0)相乘的绝对值,例如本题是K2=24Kc
3.求出Kc=K2/Kg=(2/0.1)/24=0.833
4.Css=0.833/s
lead and lag network design
解题过程
- 根据er_inf 或者y_da 求出Css
- 根据S_ave,tr,ts求出Tp,Sp,Wc,des
- 画图nichols和Tgrid Sgrid,描出Wc,des点并移出大圈之外得到相位角
-
根据相位角带入到diagram中找到md和wd或者mi和wi
wd=Wc,des/Wnorm
alfa=Wnorm/Mi - 根据4的结果带入到Ci(s)或者Cd(s)中
- L(s)=Ci(s)G(s) L(s)=Cd(s)G(s)
大题总结
求css
根据表求范围 找到范围内的一个合适的值
手画 L 的nyquist曲线判断稳定性 ,如果L实部>0的极点数量=卷绕数
根据稳定性判断正负号
lead和lag##
加零点右上移动,极点左下移动
图一是分贝,图2是相位
结合得到合适的W/Wz的值 求出Wz,带入lead公式里面,调节Md和Wd的值适合simulink
如果开环传递函数在原点上有重极点,则奈奎斯特图在频率有不连续性。在进一步分析中,应当假设相矢量在极点附近沿着一个半径无限大的半圆转了 次。通过附加这一条件,这些位于原点上的极点可以被忽略掉,也就是说,如果系统中没有其他不稳定的极点,开环传递函数应被认为是稳定的。
如果开环传递函数在右半平面内没有极点,则它是稳定的,而若其对应的闭环传递函数对(-1+j0)点没有卷绕,则闭环传递函数也是稳定的,否则会是不稳定的。
如果闭环传递函数对(-1+j0)点有顺时针卷绕,则它是不稳定的;如果对(-1+j0)点有逆时针卷绕,且逆时针卷绕数等于开环传递函数在右半平面内的极点数,则闭环系统是稳定的,否则会是不稳定的。