Control of Electromechanical Systems | Compiled Lecture Notes (no proper citations, for
personal use only)
Control of Electromechanical
Systems
Prof. Claudio Roberto Gaz
gaz@[Link]
Solayman Salindato Maso
EMMC STEPS
maso.1851070@[Link]
“La Sapienza” Universita di Roma
September 2018
1st Semester, AY 2018-2019
Course Information
Course : Control of Electromechanical Systems
Professor : Claudia Roberto Gaz
email : gaz@[Link]
Textbook : Digital Control of Electrical Drives by Slobodan N. Vukosavić
Class Schedule
Monday : 2:00 – 4:00 AM
Wednesday : 8:00 – 10:00 AM
Thursday : 11:00 – 1:00 PM
Friday : 8:00 – 10:00 AM
Date: 16/10/2018
Exercise: Choosing 𝜔𝑠 = 10
Find a proper sampling time T for the following 2𝜋 2𝜋 𝜋
𝑇𝑠 = = = = 𝟎. 𝟐𝝅
transfer function. 𝜔𝑠 10 5
1 1 2𝜋
𝐺(𝑠) = 𝑇= ; 𝜔 = 2𝜋𝑓 ; 𝑇 =
𝑠(𝑠 + 1) 𝑓 𝜔
Solution:
𝜔 + 𝑒 𝜇 𝜇 = 1.5
Solve for |𝐺(𝑠)| , 𝑠 = 𝑗𝜔 𝜇 G(s) y
−
1
|𝐺(𝑠)| = | |
𝑗𝜔(𝑗𝜔 + 1)
For the denominator part, 𝜔 + 𝑒 𝑇 𝑒∗ 𝜇∗ 𝜇 𝑦
𝜇 ZOH 𝐺(𝑠)
|𝑗𝜔(𝑗𝜔 + 1)| = |−𝜔2 + 𝑗𝜔| = √(−𝜔 2 )2 + 𝜔 2 −
= √𝜔 4 + 𝜔 2
𝑒 − error
1 1
= 𝑇 = 0.2𝜋
√𝜔𝐵𝑊 4 + 𝜔𝐵𝑊 2 √2
√𝜔𝐵𝑊 4 + 𝜔𝐵𝑊 2 = √2 → 2 = 𝜔𝐵𝑊 4 + 𝜔𝐵𝑊 2 1
𝑊𝑠 (𝑠) = 𝐻 (𝑠)𝑅 ∗ (𝑒 𝑠𝑇 )𝐺(𝑠)
𝑇 0
Let 𝜔𝐵𝑊 2 = 𝑥 1 1 − 𝑒 −𝑠𝑇 1.5 1 − 𝑒 −0.2𝜋𝑠 1.5
= ∙ = ∙
𝑇 𝑠 𝑠(𝑠 + 1) 0.2𝜋𝑠 𝑠(𝑠 + 1)
𝑥2 + 𝑥 − 2 = 0
(𝑥 + 2)(𝑥 − 1) = 0 where:
1 − 𝑒 −𝑠𝑇
𝑥 = 1 ; = −2 𝐻0 (𝑠) =
𝑠
We take the positive solution because there is no
negative frequency.
𝜔𝐵𝑊 2 = 1 , 𝜔𝐵𝑊 = 1
2𝑓𝑚𝑎𝑥 < 𝑓𝑠 → Nyquist Shannon Theorem
Shannon’s give sets this as the minimum
We extend this theorem; therefore, for bandwidth
2𝜔𝑚𝑎𝑥 < 𝜔𝑠
But in practice (industry):
10𝑓𝑚𝑎𝑥 < 𝑓𝑠
5𝜔𝐵𝑊 ≤ 𝜔𝑠 ≤ 10𝜔𝐵𝑊
The Z Transform
𝑢∗ (𝑡) = 𝑢0 𝛿(𝑡) + 𝑢1 𝛿(𝑡 − 𝑇) 𝑢(𝑡)
+ 𝑢2 𝛿(𝑡 − 2𝑇)+ . . . +𝑢𝑘 𝛿(𝑡 − 𝑘𝑇)
∞
= ∑ 𝑢𝑘 𝛿(𝑡 − 𝑘𝑇)
𝑘=0 𝑢0
𝑢(𝑡) t
𝑢2 0 T 2T 3T
𝑢1
𝑢3 Fig: unit impulse
𝑢0
Step function
t
0 T 2T 3T 𝑢(𝑡)
Fig: train of pulses
The Laplace transform of 𝜇∗ (𝑡) is 𝑈 ∗ (𝑠) 1
𝑈 ∗ (𝑠) = 𝑢0 + 𝑢1 𝑒 −𝑠𝑇 + 𝑢2 𝑒 −2𝑠𝑇 + . . . +𝑢𝑘 𝑒 −𝑘𝑠𝑇
∞
t
= ∑ 𝑢𝑘 (𝑒 −𝑠𝑇 )𝑘 0 T 2T 3T
𝑘=0
Fig: Step input
Let 𝑧 be defined by 𝑧 = 𝑒 𝑠𝑇 , then we have
{𝑢𝑘 }∞
𝑘=0 = {1,1,1, . . .1} → 𝑈(𝑧) = 1 + 𝑧
−1
+ 𝑧 −2 + ⋯
∞
𝑈(𝑧) = ∑ 𝑢𝑘 𝑧 −𝑘 1 𝑧
𝑘=0 𝑈(𝑧) = −1
=
1−𝑧 𝑧−1
Consider
𝑛 Exponential function
1 − 𝑎𝑛+1
∑ 𝑎𝑘 = ,𝑎 ≠ 1
1−𝑎
𝑘=0
𝑢(𝑡)
1
We can see that this converges to
∞
1
∑ 𝑎𝑘 = , |𝑎| < 1
1−𝑎
𝑘=0
t
The Z Transform of common functions 0 T 2T 3T
Unit impulse Fig: Exponential function
1 , 𝑘=0 𝑈(z) = 1 + 𝑎𝑧 −1 + 𝑎2 𝑧 −2 + . . . (𝑎⁄𝑧)𝑘
𝜇(𝑘) = 𝛿(𝑘) = { 𝑈(z) = 1 1 𝑧
0 , 𝑘≠0 𝑈(z) = 𝑎 =𝑧−𝑎
1−𝑧
Ramp: 𝐹(𝑧) 1
𝑧 𝐴=𝑧∙ | = = 50
𝑍{𝑡} = 𝑧 𝑧=0 0.02
(𝑧 − 1)2
𝐹(𝑧) 1 − 0.1
Sine: 𝐵 = (𝑧 + 0.1) | = = −90
𝑧 sin 𝜔 𝑧 𝑧=−0.1 (−0.1)(−0.1 + 0.2)
𝑍{sin 𝜔𝑡} =
𝑧 2 − 2𝑧 cos 𝜔 + 1
𝐹(𝑧) 1 − 0.2
𝐶 = (𝑧 + 0.2) | = = 40
Cosine: 𝑧 𝑧=−0.2 (−0.2)(−0.2 + 0.1)
𝑧 2 + 𝑧 cos 𝜔
𝑍{cos 𝜔𝑡} = 𝐹(𝑧) 50 90 40
𝑧 2 − 2𝑧 cos 𝜔 + 1 = − +
𝑧 𝑧 𝑧 + 0.1 𝑧 + 0.2
50𝑧 90𝑧 40𝑧
𝑡
𝑎𝑧 sin 𝜔 𝐹(𝑧) = − +
𝑍{𝑎 sin 𝜔𝑡} = 2 𝑧 𝑧 + 0.1 𝑧 + 0.2
𝑧 − 2𝑧 cos 𝜔 + 𝑎2
50𝛿(𝑘) − 90(−0.1)𝑘 + 40(−0.2)𝑘 , 𝑘≥0
𝑧(𝑧 − 𝑎 sin 𝜔) 𝑓(𝑘) = {
0 , 𝑘<0
𝑍{𝑎𝑡 cos 𝜔𝑡} = 2
𝑧 − 2𝑧 cos 𝜔 + 𝑎2
𝑘 𝑘
Properties of Z-Transform 𝑓(𝑘) = { −90(−0.1) + 40(−0.2) , 𝑘≥1
0 𝑘<1
Linearity: 𝑍{𝛼𝑓1 (𝑘) + 𝛽𝑓2 (𝑘)} = 𝛼𝐹1 (𝑧) + 𝛽𝐹2 (𝑧)
Bilinear Transformation
Time delay: 𝑍{𝑓(𝑘 − 𝑛)} = 𝑧 −𝑛 𝐹(𝑧)
Time advance: 𝑍{𝑓(𝑘 + 1)} = 𝑧𝐹(𝑧) − 𝑧𝑓(0) 𝑒 = ∆𝜔 𝑢 = 𝑇𝑠𝑐
𝑅∗
Multiplication by exponent: 𝑍{𝑎−𝑘 𝑓(𝑘)} = 𝐹(𝑎𝑧)
where 𝑅 ∗ is the state of the regulator
Complex differentiation:
𝑑 𝑚 𝑅(𝑠) = 𝐶(𝑠𝐼 − 𝐴)−1 𝐵 + 𝐷 (𝑎)
𝑍{𝑘 𝑚 𝑓(𝑘)} = (−𝑧 ) 𝐹(𝑧)
𝑑𝑧
Time convolution: 𝑍{𝑓1 (𝑘) ∗ 𝛽(𝑘)} = 𝐹1 (𝑧) ∙ 𝐹2 (𝑧) 𝑥̇ (𝑡) = 𝐴𝑥(𝑡) + 𝐵𝑒(𝑡) (𝑏)
{
𝑢(𝑡) = 𝐶𝑥(𝑡) + 𝐷𝑒(𝑡)
Initial value theorem: 𝑓(0) = lim 𝐹(𝑧) (𝑐)
𝑧→∞
𝐼𝑛𝑡𝑒𝑔𝑟𝑎𝑡𝑒 𝑡ℎ𝑒 𝑡𝑤𝑜 𝑚𝑒𝑚𝑏𝑒𝑟𝑠
Final value theorem: lim 𝑓(𝑘) = lim(𝑧 − 1) 𝐹(𝑧)
𝑘→∞ 𝑧→1 𝑥 ∗ 𝑘 = 𝑥(𝑘𝑇)
Exercise: 𝑥 ∗ (𝑘 + 1) − 𝑥 ∗ (𝑘) (𝑑)
(𝑘+1)𝑇 (𝑘+1)𝑇
Obtain the inverse transform of = 𝐴∫ 𝑥(𝑡)𝑑𝑡 + 𝐵 ∫ 𝑒(𝑡)𝑑𝑡
𝑘𝑇 𝑘𝑇
𝑧+1
𝐹(𝑧) =
𝑧2 + 0.3𝑧 + 0.02
Solution:
𝐹(𝑧) 𝑧+1 𝐴 𝐵 𝐶
= = + +
𝑧 𝑧(𝑧 2 + 0.3𝑧 + 0.02) 𝑧 𝑧 + 0.1 𝑧 + 0.2
Consider a general integral
(𝑘+1)𝑇
∫ 𝑓(𝑡)𝑑𝑡 ≅ [(1 − 𝛼)𝑓(𝑘𝑇) + 𝑎𝑓(𝑘 + 1)𝑇]𝑇 , 0 ≤ 𝛼 ≤ 1 (𝑒)
𝑘𝑇
𝛼 = 0 → Forward Integration (or Eurler ′ s Rule)
𝛼 = 1 → Backward Integration
1
𝛼 = → Tustin′ s Method (or Trapezoidal Rule)
2
Substitute e in d
𝑥 ∗ (𝑘 + 1) − 𝑥 ∗ (𝑘) = 𝐴[(1 − 𝛼)𝑥 ∗ (𝑘) + 𝑥 ∗ (𝑘 + 1)]𝑇 + 𝐵[(1 − 𝛼)𝑒 ∗ (𝑘) + 𝑒 ∗ (𝑘 + 1)]𝑇 (𝑓)
𝑒 ∗ (𝑘) = 𝑒(𝑘𝑇)
From (c), we have
𝑢∗ (𝑘) = 𝐶𝑥 ∗ (𝑘) + 𝐷 𝑒 ∗ (𝑘) (𝑔)
𝑢∗ (𝑘) = 𝑢(𝑘𝑇)
The two relations (f) and (g) are an approximate representation of the original system.
Apply the Z-transform to (f)
𝑋 ∗ (𝑧)𝑧 − 𝑋 ∗ (𝑧) = 𝐴[𝑋 ∗ (𝑧) − 𝛼𝑋 ∗ (𝑧) + 𝛼𝑋 ∗ (𝑧)𝑧]𝑇 + 𝐵[𝐸 ∗ (𝑧) − 𝛼𝐸 ∗ (𝑧) + 𝛼𝐸 ∗ (𝑧)𝑧]𝑇
𝑥(𝑘 + 1) = 𝑋 ∗ (𝑧)𝑧
𝑥(𝑘) = 𝑋 ∗ (𝑧)
𝑋 ∗ (𝑧)𝑧 − 𝑋 ∗ (𝑧) = 𝐴[𝑋 ∗ (𝑧)(1 − 𝛼 + 𝛼𝑧)]𝑇 + 𝐵[𝐸 ∗ (𝑧)(1 − 𝛼 + 𝛼𝑧)]𝑇
𝑋 ∗ (𝑧)𝑧 − 𝑋 ∗ (𝑧) − 𝑋 ∗ (𝑧)[𝐴𝑇(1 − 𝛼 + 𝛼𝑧)] = 𝐸 ∗ (𝑧)[𝐵𝑇(1 − 𝛼 + 𝛼𝑧)]
𝑋 ∗ (𝑧)[𝑧 − 1 − 𝐴𝑇(1 − 𝛼 + 𝛼𝑧)]
= 𝐵𝐸 ∗ (𝑧)
𝑇(1 − 𝛼 + 𝛼𝑧)
1 𝑧−1
𝑋 ∗ (𝑧) [ ∙ 𝐼 − 𝐴] = 𝐵𝐸 ∗ (𝑧) (ℎ)
𝑇 1 − 𝛼 + 𝛼𝑧
−1
∗ (𝑧)
1 𝑧−1
𝑋 =[ ∙ 𝐼 − 𝐴] 𝐵𝐸 ∗ (𝑧)
𝑇 1 − 𝛼 + 𝛼𝑧
Hence, using (g) and recalling (𝑎)
𝑢∗ (𝑘) = 𝐶𝑥 ∗ (𝑘) + 𝐷 𝑒 ∗ (𝑘) → (𝑔)
𝑅(𝑠) = 𝐶(𝑠𝐼 − 𝐴)−1 𝐵 + 𝐷 → (𝑎)
We have the transfer function
𝑈 ∗ (𝑧)
𝑅 ∗ (𝑧) =
𝐸 ∗ (𝑧)
−1
1 𝑧−1 1 𝑧−1
𝑅 ∗ (𝑧) = 𝐶 [ ∙ 𝐼 − 𝐴] 𝐵 + 𝐷 = 𝑅 ∗ ( )
𝑇 1 − 𝛼 + 𝛼𝑧 𝑇 𝛼𝑧 + 1 − 𝛼
𝟏 𝒛−𝟏
𝒔= → sampling time
𝑻 𝜶𝒛 + 𝟏 − 𝜶
