《控制理论》课程教学资源（参考书籍）定量过程控制理论 Quantitative Process Control Theory_Chapter 03 Essentials of Robust Control

Chapter 3 Essentials of Robust Control 4口，+@，4定4=定0C Zhang.W.D..CRC Press.2011 Version 1.0 1/69

Chapter 3 Essentials of Robust Control Zhang, W.D., CRC Press, 2011 Version 1.0 1/69

Essentials of Robust Control 13.1 Norms and System Gains 23.2 Internal Stability and Performance 33.3 Controller Parameterization 43.4 Robust Stability and Robust Performance 53.5 Robustness of Systems with Time Delays 4口，+@，4定4定9QC Zhang.W.D..CRC Press.2011 Version 1.0 2/69

Essentials of Robust Control 1 3.1 Norms and System Gains 2 3.2 Internal Stability and Performance 3 3.3 Controller Parameterization 4 3.4 Robust Stability and Robust Performance 5 3.5 Robustness of Systems with Time Delays Zhang, W.D., CRC Press, 2011 Version 1.0 2/69

Section 3.1 Norms and System Gains 3.1 Norms and System Gains Why Do We Need Norms The performance of a control system is usually specified in terms of the "size"of certain signals. Solution:The "size"of a signal can be defined by introducing norms."The signal is small"means its norm is small Consider a signal r(t).A norm is a nonnegative real number. denoted by r(t).that satisfying the following properties: r(t)=0 if and only if r(t)=0.vt. 3or(t)=ar(t)o is any real number. n(e)+2(c)≤n()+2(e 4口，+@，4定4=定0C Zhang.W.D..CRC Press.2011 Version 1.0 3/69

Section 3.1 Norms and System Gains 3.1 Norms and System Gains Why Do We Need Norms The performance of a control system is usually specified in terms of the ”size” of certain signals. Solution:The ”size” of a signal can be defined by introducing norms. “The signal is small” means its norm is small Consider a signal r(t). A norm is a nonnegative real number, denoted by kr(t)k, that satisfying the following properties: 1 kr(t)k = 0 if and only if r(t) = 0 , ∀t. 2 kαr(t)k = |α|kr(t)k, α is any real number. 3 kr1(t) + r2(t)k ≤ kr1(t)k + kr2(t)k. Zhang, W.D., CRC Press, 2011 Version 1.0 3/69

Section 3.1 Norms and System Gains 3.1 Norms and System Gains Why Do We Need Norms The performance of a control system is usually specified in terms of the "size"of certain signals. Solution:The "size"of a signal can be defined by introducing norms."The signal is small"means its norm is small Consider a signal r(t).A norm is a nonnegative real number, denoted by (t),that satisfying the following properties: llr(t)ll =0 if and only if r(t)=0.Vt. llar(t)=alr(t),a is any real number. ③ln(t)+2(t)川≤ln(t)l+I2(t)l- 4口，4心4定4生，定QC Zhang.W.D..CRC Press.2011 Version 1.0 3/69

Section 3.1 Norms and System Gains 3.1 Norms and System Gains Why Do We Need Norms The performance of a control system is usually specified in terms of the ”size” of certain signals. Solution:The ”size” of a signal can be defined by introducing norms. “The signal is small” means its norm is small Consider a signal r(t). A norm is a nonnegative real number, denoted by kr(t)k, that satisfying the following properties: 1 kr(t)k = 0 if and only if r(t) = 0 , ∀t. 2 kαr(t)k = |α|kr(t)k, α is any real number. 3 kr1(t) + r2(t)k ≤ kr1(t)k + kr2(t)k. Zhang, W.D., CRC Press, 2011 Version 1.0 3/69

Section 3.1 Norms and System Gains Frequently Used Signal Norms 1-norm.The 1-norm of r(t)is the integral of its absolute value: lr()l =lr(e)l de 2-norm.The 2-norm of r(t)is 1/2 Ir()=P()de Suppose that r(t)is the current through a 10 resistor.The instantaneous power equals r(t)2 and the energy equals r(t) x-norm.The oo-norm of r(t)is the least upper bound of its absolute value: r(t)oo=sup r(t) 4口，+@，4定4=定0C Zhang.W.D..CRC Press.2011 Version 1.0 4/69

Section 3.1 Norms and System Gains Frequently Used Signal Norms 1-norm. The 1-norm of r(t) is the integral of its absolute value: kr(t)k1 := Z ∞ −∞ |r(t)| dt 2-norm. The 2-norm of r(t) is kr(t)k2 := Z ∞ −∞ r 2 (t) dt1/2 Suppose that r(t) is the current through a 1Ω resistor. The instantaneous power equals r(t) 2 and the energy equals kr(t)k 2 2 ∞-norm. The ∞-norm of r(t) is the least upper bound of its absolute value: kr(t)k∞ := sup t |r(t)| Zhang, W.D., CRC Press, 2011 Version 1.0 4/69

Section 3.1 Norms and System Gains Frequently Used Signal Norms 1-norm.The 1-norm of r(t)is the integral of its absolute value: r()=Ir(e)l de 2-norm.The 2-norm of r(t)is 11/2 Ir(2()d Suppose that r(t)is the current through a 10 resistor.The instantaneous power equals r(t)2 and the energy equals (t)2 x-norm.The oo-norm of r(t)is the least upper bound of its absolute value: r(t)oo=sup r(t) 4口，+@，4定4=定0C Zhang.W.D..CRC Press.2011 Version 1.0 4/69

Section 3.1 Norms and System Gains Frequently Used Signal Norms 1-norm. The 1-norm of r(t) is the integral of its absolute value: kr(t)k1 := Z ∞ −∞ |r(t)| dt 2-norm. The 2-norm of r(t) is kr(t)k2 := Z ∞ −∞ r 2 (t) dt1/2 Suppose that r(t) is the current through a 1Ω resistor. The instantaneous power equals r(t) 2 and the energy equals kr(t)k 2 2 ∞-norm. The ∞-norm of r(t) is the least upper bound of its absolute value: kr(t)k∞ := sup t |r(t)| Zhang, W.D., CRC Press, 2011 Version 1.0 4/69

Section 3.1 Norms and System Gains Frequently Used Signal Norms 1-norm.The 1-norm of r(t)is the integral of its absolute value: r()=Ir(e)l de 2-norm.The 2-norm of r(t)is 11/2 Ir(()d Suppose that r(t)is the current through a 10 resistor.The instantaneous power equals r(t)2 and the energy equals lr(t)2 oo-norm.The oo-norm of r(t)is the least upper bound of its absolute value: lr(t)川o:=supr(t)l 4口，+@，4定4=定0C Zhang.W.D..CRC Press.2011 Version 1.0 4/69

Section 3.1 Norms and System Gains Frequently Used Signal Norms 1-norm. The 1-norm of r(t) is the integral of its absolute value: kr(t)k1 := Z ∞ −∞ |r(t)| dt 2-norm. The 2-norm of r(t) is kr(t)k2 := Z ∞ −∞ r 2 (t) dt1/2 Suppose that r(t) is the current through a 1Ω resistor. The instantaneous power equals r(t) 2 and the energy equals kr(t)k 2 2 ∞-norm. The ∞-norm of r(t) is the least upper bound of its absolute value: kr(t)k∞ := sup t |r(t)| Zhang, W.D., CRC Press, 2011 Version 1.0 4/69

Section 3.1 Norms and System Gains Frequently Used System Norms Consider a linear time-invariant and causal system T(t),of which the input is r(t)and the output is y(t): y(t)=T(t)*r(t) T(t-T)r(T)dr Let T(s)denote the transfer function of T(t).Norms can also be defined for the system T(s): 2-norm. 11/2 Irse=层ITVP d oo-norm. I‖T(s)Ilo:=sup|TUw)川 4口，+@，4定4=定0C Zhang.W.D..CRC Press.2011 Version 1.0 5/69

Section 3.1 Norms and System Gains Frequently Used System Norms Consider a linear time-invariant and causal system T(t), of which the input is r(t) and the output is y(t): y(t) = T(t) ∗ r(t) = Z ∞ −∞ T(t − τ )r(τ ) dτ Let T(s) denote the transfer function of T(t). Norms can also be defined for the system T(s): 2-norm. kT(s)k2 :=  1 2π Z ∞ −∞ |T(jω)| 2 dω 1/2 ∞-norm. kT(s)k∞ := sup ω |T(jω)| Zhang, W.D., CRC Press, 2011 Version 1.0 5/69

Section 3.1 Norms and System Gains Theorem The 2-norm of T(s)is finite if and only if T(s)is strictly proper and has no poles on the imaginary axis.The oo-norm of T(s)is finite if and only if T(s)is proper and has no poles on the imaginary axis. Proof. Assume that T(s)is strictly proper and has no poles on the imaginary axis.Then the Bode magnitude plot rolls off at high frequencies.It is not hard to see that the plot of c/(rs+1)is higher than that of T(s)for sufficiently large positive c and sufficiently small positive r,but the 2-norm of c/(rs +1)equals c/V2T.Hence T(s)has finite 2-norm. The rest of the proof follows similar lines 4口，+@，4定4=定0C Zhang.W.D..CRC Press.2011 Version 1.0 6/69

Section 3.1 Norms and System Gains Theorem The 2-norm of T(s) is finite if and only if T(s) is strictly proper and has no poles on the imaginary axis. The ∞-norm of T(s) is finite if and only if T(s) is proper and has no poles on the imaginary axis. Proof. Assume that T(s) is strictly proper and has no poles on the imaginary axis. Then the Bode magnitude plot rolls off at high frequencies. It is not hard to see that the plot of c/(τ s + 1) is higher than that of T(s) for sufficiently large positive c and sufficiently small positive τ , but the 2-norm of c/(τ s + 1) equals c/ √ 2τ . Hence T(s) has finite 2-norm. The rest of the proof follows similar lines. Zhang, W.D., CRC Press, 2011 Version 1.0 6/69

Section 3.1 Norms and System Gains Theorem The 2-norm of T(s)is finite if and only if T(s)is strictly proper and has no poles on the imaginary axis.The oo-norm of T(s)is finite if and only if T(s)is proper and has no poles on the imaginary axis. Proof. Assume that T(s)is strictly proper and has no poles on the imaginary axis.Then the Bode magnitude plot rolls off at high frequencies.It is not hard to see that the plot of c/(Ts+1)is higher than that of T(s)for sufficiently large positive c and sufficiently small positive T,but the 2-norm of c/(Ts+1)equals c/v27.Hence T(s)has finite 2-norm. The rest of the proof follows similar lines. 4口，44定4生，定QC Zhang.W.D..CRC Press.2011 Version 1.0 6/69

Section 3.1 Norms and System Gains Theorem The 2-norm of T(s) is finite if and only if T(s) is strictly proper and has no poles on the imaginary axis. The ∞-norm of T(s) is finite if and only if T(s) is proper and has no poles on the imaginary axis. Proof. Assume that T(s) is strictly proper and has no poles on the imaginary axis. Then the Bode magnitude plot rolls off at high frequencies. It is not hard to see that the plot of c/(τ s + 1) is higher than that of T(s) for sufficiently large positive c and sufficiently small positive τ , but the 2-norm of c/(τ s + 1) equals c/ √ 2τ . Hence T(s) has finite 2-norm. The rest of the proof follows similar lines. Zhang, W.D., CRC Press, 2011 Version 1.0 6/69