 Open Access
 Total Downloads : 129
 Authors : Bijily Rose Varghese, Manju G
 Paper ID : IJERTV4IS100555
 Volume & Issue : Volume 04, Issue 10 (October 2015)
 DOI : http://dx.doi.org/10.17577/IJERTV4IS100555
 Published (First Online): 30102015
 ISSN (Online) : 22780181
 Publisher Name : IJERT
 License: This work is licensed under a Creative Commons Attribution 4.0 International License
Longitudinal Parameter Estimation of Aerospace Vehicle with Different Noise Levels
Bijily Rose Varghese
Department of Electrical and Electronics Engineering, MBCET, Kerala University,
Trivandrum, India
Ms. Manju G
Assistant Professor,
Department of Electrical and Electronics Engineering, MBCET, KeralaUniversity,
Trivandrum, India.
Abstract This paper aims to estimate the longitudinal aerodynamic parameters of the aerospace vehicle from the available measurements with different noise levels. In this current work Filter Error Method (FEM) which accounts for both process and measurement noise, is used to estimate the aerodynamic parameters from vehicle data. FEM algorithm is developed in MATLAB environment. With the developed algorithm studies are carried out at different instance of the vehicle and the estimated parameters are compared with wind tunnel predictions.
Keywords Aerodynamic parameter; Filter Error Method; Gauss Newton Method; Kalman Filter.

INTRODUCTION
Parameter estimation, is the subfield of system identification, which determines the best estimates of the parameters occurring in the model. Aerospace vehicle parameter estimation is the best example of system identification. An aircraft motion is described by equations of motion derived from the Newtonian mechanics, considering the vehicle as a rigid body. The aerodynamic forces and moments cannot be
Fig. 1. Aerodynamic forces and moments
Fig.1 shows the different forces and moments acting on the vehicle along the three axis. Aerodynamic forces acting on the vehicle with respect to longitudinal plane are given from (1)(2). [1]
measured directly. However, aerodynamic modeling followed by parameter estimation helps to determine the aerodynamic parameters. Conventional methods of parameter estimation are (i)Equation Error Method[2] (ii) Output Error Methods (OEM)[3,4] and (iii) Filter Error Methods (FEM). Out of these methods Filter Error Method (FEM), is the most general stochastic approach for aircraft parameter estimation, which
Fx CAqS
FZ CN qS
Aerodynamic moment is given by (3).
M CmqSc
(1)
(2)
(3)
accounts for both process and measurement noise. In this current work Filter Error Method (FEM), is used to estimate the aerodynamic parameters. FEM algorithm is developed in MATLAB environment. State estimation is carried out with Kalman Filter and the parameters are updated using Gauss Newton Method. With the developed algorithm studies are carried out at different instance of the vehicle and the
Where q is the dynamic pressure, S is the reference area, c
is
the longitudinal reference length, CA ,CN ,Cm are the total axial force coefficient, total normal force coefficients and total pitching moment coefficients. The state equations are given from (4)(7)
estimated parameters are compared with wind tunnel predictions.

LONGITUDINAL DYNAMICS Aerospace vehicle experiences aerodynamic forces and
V uu vv ww
u2 v2 w2
uw wu
u2 w2
(4)
(5)
moments during their motion in atmosphere.
1 cos
(q cos r sin )
(6)
C qSc I (r2 p2 ) rp(I I ) I ( p qr) I q m zx zz xx xy yz
Iyy
(r pq)
(7)

EFFECT OF NOISE
The real aerospace vehicle measurements are affected by both
Where u, v, w are the translational velocity, p, q, r are the
measurement noise and process noise. To study the effect of
the noise on the estimated coefficients, noise is added in q,
body rates, V is the vehicle velocity, is the angle of attack,
ax , az by specifying the signal to noise ratio. The nosie
Ixx , I yy , Izz
are the moment of inertia and Ixy , I yz , Izx
are
added is shown in table 1.
the product of inertia, ,, are the euler angles. The aerodynamic coefficients are modelled in linear form as:
Table 1: Noise added in the variable
CA CA0
CA
C
A e1e1
CA e2e2
CA r1r1
CA r 2r 2
(8)
C C C C
C
C C
(9)
N N 0 N
N e1 e1
N e2 e2 N r1 r1
N r 2 r 2
C C C C
C C
C
m m0 m
(10)
m e1 e1 me2 e2
m r1 r1 m r 2 r 2
Variable
Noise added (signal to noise ratio)
Case1
Case 2
Case3
Case 4
q
–
40
50
80
ax
–
40
50
80
az
–
40
50
80
Due to the presence of process noise, Equation Error Method
where CA0,CN 0,Cm0,CA ,CN ,Cm are the basic coefficients and CA e1,CA e2 ,CA r1,CA r 2 ,
CN e1,CN e2 ,CN r1,CN r 2 , Cm e1,Cm e2 ,Cm r1,Cm r 2 are
the incremental coefficients due to control surface deflections. Thus the unknown parameters to be estimated are given as:
T [C C C C C C C C
,Output Error Method estimation techniques yield poor estimation results. Therefore, Filter Error Method [5] is used in this work which account for process and measurement noise. Since the system is stochastic in nature a steady state Kalman filter is used for obtaining the true state variables from the noisy measurements.

RESULTS AND DISCUSSION Estimation is carried out for a period of 10s with 500
A0 A A e1 A e2 A r1 A r 2 N 0 N
CN e1CN e2CN r1CN r 2Cm0Cm C Cm e2Cm r1Cm r 2 ]
Observation equations are given by (12)(18)
Vm V
m
m e1
(11)
(12)
(13)
samples. Region considered for the estimation process is for the vehicle time 310320s. The initial values of the estimates are taken from the wind tunnel data. The variation of angle of attack, dynamic pressure, Mach number, relative velocity and control surface deflections during this region are shown in Fig.3 and 4. The estimated longitudinal aerodynamic coefficients are shown in Fig 5,6 and 7.
C qSc I
m
qm q
(r2 p2 ) rp(I

I ) I
( p qr) I
(r pq)
(14)
(15)
3.2
Mach Number
3.1
3
2.9
10000
Dynamic Pressure(Pa)
9000
8000
7000
qm
m zx zz xx xy yz
Iyy
a CAqS xm m
(16)
(17)
2.8
Relative Velocity(m/s)
1000
0 5 10
Time(s)
6000
22
Alpha(deg)
21
0 5 10
Time(s)
azm
CN qS
m
(18)
950
20
900
Where, m is the mass of the vehicle, V , , , q are the 19
m m m m
measured velocity, angle of attack, euler angle, and body
850
0 5 10
18
0 5 10
rate, qm is the measured angular accelerations and
axm , azm
Time(s)
Time(s)
are the measured accelerations.
Fig. 3. Mach number, relative velocity, dynamic pressure and alpha during the estimation region.
Left Elevon Deflection(deg)
2
0
4
0 5 10
4
Right Elevon Deflection(deg)
2
0
2
0 5 10
4
x 10
2
Cm Total
1
0
1
2
Wind Tunnel No Noise
With Noise(case1) With Noise(case2) With Noise(case3)
Time(s)
Time(s)
0 1 2 3 4 5 6 7 8 9 10
No Noise
With Noise(case1) With Noise(case2) With Noise(case3)
Time(s)
Left Rudder Deflection(deg)
Right Rudder Deflection(deg)
5 8
4
x 10
2
Difference
6 1
0
4 0
5
0 5 10
Time(s)
2
0 5 10
Time(s)
1
2
0 1 2 3 4 5 6 7 8 9 10
Time(s)
Fig.4.Control surface deflections during the estimation region
Wind Tunnel
0.076
No Noise
With Noise(case1)
With Noise(case2) With Noise(case3)
CA total
0.075
0.074
0.073
0.072
0 1 2 3 4 5 6 7 8 9 10
No Noise
Time(s)
With Noise(case1)
With Noise(case2) With Noise(case3)
1
%Error
0.5
0
0.5
0 1 2 3 4 5 6 7 8 9 10
Time(s)
Fig. 5 The estimated total aerodynamic force coefficient CA
Wind Tunnel No Noise With Noise(case1) With Noise(case2) With Noise(case3) 

0 
1 
2 
3 
4 
5 Time(s) 
6 
No Noise With Noise(case1) With Noise(case2) With Noise(case3) 

0.7
0.68
CN total
0.66
0.64
7 8 9 10
0.62
0.2
%Error
0.1
0
0.1
Fig.7.The estimated total aerodynamic moment coefficient Cm
CONCLUSION
Filter error algorithm is formulated and developed in MATLAB environment. With the developed algorithm longitudinal aerodynamic parameters are estimated with different noise levels .The estimated longitudinal aerodynamic coefficients are compared with wind tunnel prediction. Maximum of 0.55%, 0.13%, error, are observed in total aerodynamic force coefficients, 0.000012 difference are observed in total aerodynamic moment coefficient.
REFERENCES

R.E.Malne and L.W.Lfiff, Identification of dynamic systems applications to aircraft, part1output error approach, AGARDL flight test techniques series, vol. 3, pp. 733, July 1986.

Rakesh Kumar, Lateral parameter estimation using regression, International Journal of Engineering Inventions ISSN, vol. 1,pp: 7681, October2012.

Luiz C.S.GÃ³es, Elder M.Hemerly and Benedito C.O Maciel, Parameter estimation and flight path reconstruction using outputerror method,24th international congress of the aeronautical sciences,pp:110,2004.

R V Jategaonkar and F Thieleckec, Aircraft parameter estimation, A tool for development of aerodynamic databases, Saadhanaa, Vol. 25, Part 2, pp.119135,April 2000.

R.V Jategoankar, Flight Vehicle System Identification : A time domain methodology , volume 216, 2006.
0.2
0 1 2 3 4 5 6 7 8 9 10
Time(s)
Fig. 6 The estimated total aerodynamic force coefficient CN