Date post: | 16-Apr-2015 |
Category: |
Documents |
Upload: | asif-al-rasheed |
View: | 39 times |
Download: | 2 times |
Page 1 of 28
Report
For
Project 5
Kalman Estimation of the State of a DC Motor
ASIF AL – RASHEED
EE 5521
Page 2 of 28
The state of a large dc motor, the shaft angle and the angular velocity of the shaft, is estimated from noisy
measurements of the shaft angle. The input to this motor is the applied voltage u(t) plus noise on this
voltage signal w(t). The output y(t) is the motor shaft angle θ(t) plus additive measurement noise v(t).
The transfer function for the motor is:
( ) 2
( ) ( 0.02)
s
U s s s.
Estimates of both states are required to monitor the status of the motor.
Now,
2( 0.02 ) ( ) 2 ( )
( ) 0.02 ( ) 2 ( )
s s s U s
t t u t
Let,
1 1 2
2 2 20.02 ( ) 2 ( ) 0.02 2 ( )
x x x
x x t u t x u t
The state model
1 1
2 2
0 1 0( )
0 0.02 2
x xu t
x x
1
2
( ) ( ) ( ) ( ) ( )
( ) 10 10 ( )
y t t v t x t v t
xy t v t
x
Initial condition for all of the simulations,
0 deg(0)
deg(0) 0
sec
.
Page 3 of 28
Case a:
22
4
deg1
secw
; 2 220 degv
;
0 degˆ(0)degˆ 0(0)sec
;
22
2 2
2
deg0 deg 0
sec(0)
deg deg0 0
sec sec
e .
Simulation results:
Fig 1: Plot of Kalman filter gain.
Page 4 of 28
Fig 2: Plot of estimation error standard deviation.
Fig 3: State and Estimation for shaft angle
Page 5 of 28
Fig 4: Estimation error for shaft angle
Fig 5: State, Estimation and Error for shaft angle
Page 6 of 28
Fig 6: State and Estimation for angular velocity
Fig 7: Estimation error for angular velocity
Page 7 of 28
Fig 8: State, Estimation and Error for angular velocity
Page 8 of 28
Case b: The same as case a, except
22
4
deg10
secw
Simulation results:
Fig 9: Plot of Kalman filter gain.
Page 9 of 28
Fig 10: Plot of estimation error standard deviation.
Fig 11: State and Estimation for shaft angle
Page 10 of 28
Fig 12: Estimation error for shaft angle
Fig 13: State, Estimation and Error for shaft angle
Page 11 of 28
Fig 14: State and Estimation for angular velocity
Fig 15: Estimation error for angular velocity
Page 12 of 28
Fig 16: State, Estimation and Error for angular velocity
Page 13 of 28
Case c: The same as case a, except 2 2200 degv
.
Simulation results:
Fig 17: Plot of Kalman filter gain
Page 14 of 28
Fig 18: Plot of estimation error standard deviation
Fig 19: State and Estimation for shaft angle
Page 15 of 28
Fig 20: Estimation error for shaft angle
Fig 21: State, Estimation and Error for shaft angle
Page 16 of 28
Fig 22: State and Estimation for angular velocity
Fig 23: Estimation error for angular velocity
Page 17 of 28
Fig 24: State, Estimation and Error for angular velocity
Page 18 of 28
Case d: The same as case a, except
5 degˆ(0)degˆ 1(0)sec
;
22
2 2
2
deg100 deg 0
sec(0)
deg deg0 10
sec sec
eS .
Simulation results:
Fig 25: Plot of Kalman filter gain
Page 19 of 28
Fig 26: Plot of estimation error standard deviation
Fig 27: State and Estimation for shaft angle
Page 20 of 28
Fig 28: Estimation error for shaft angle
Fig 29: State, Estimation and Error for shaft angle
Page 21 of 28
Fig 30: State and Estimation for angular velocity
Fig 31: Estimation error for angular velocity
Page 22 of 28
Fig 32: State, Estimation and Error for angular velocity
Page 23 of 28
Case e: The same as case a, except use the steady-state Kalman gain.
Simulation Results:
Fig 33: Plot of Kalman filter gain
Page 24 of 28
Fig 34: Plot of estimation error standard deviation
Fig 35: State and Estimation for shaft angle
Page 25 of 28
Fig 36: Estimation error for shaft angle
Fig 37: State, Estimation and Error for shaft angle
Page 26 of 28
Fig 38: State and Estimation for angular velocity
Fig 39: Estimation error for angular velocity
Page 27 of 28
Fig 40: State, Estimation and Error for shaft angle
Page 28 of 28
Comparison of the five cases:
Cases Kalman Filter
Gains
Estimation
error standard
deviations
Mean of
Estimation Error
Rate of
convergence
(s)
Standard Deviation
of Estimation
G1 G2 σe1 σe2 Shaft
angle
Angular
velocity
G1 G2 Shaft
angle
Angular
velocity
a 0.4806 0.1515 3.1 1.718 0.4861 1.6044 15 13 311.3252 4.3741
b 0.6992 0.3738 4.25 3.74 -0.114 0.2775 9 9 776.0154 9.1066
c 0.3007 0.05312 7.755 2.219 -0.436 0.3904 26 24 286.7357 3.0982
d 0.4806 0.1515 3.1 1.718 -0.288 0.0773 17 9 163.0103 3.6302
e 0.6004 0.2707 1.735 1.457 0.0625 0.1256 1 1 116.5068 4.8030
Comments:
A common trend that we observe among all cases is that the shaft angle estimate more accurately tracks
than the angular velocity estimate. From the results we see that using the steady-state Kalman gain yields
the best filter evident by the fastest rate of convergence, the lowest estimation error standard deviations,
and the lower values of mean and standard deviation of estimation and estimation errors. Also cases a and
d give almost identical results in terms of gains, convergence rates and estimation error standard
deviations but approach the steady state from opposite directions. Increasing the plant noise in case b
increases the output resulting in increase in gain. But increasing the output noise in case c reduces the
output, decreasing the gain. So we can conclude that the gain is proportional to the plant noise and
inversely proportional to the output noise.