ThM21.

Fault Tolerant Flight Control Using Sliding

Modes and Subspace Identification-based
Predictive Control

Bilal A. Siddiqui (DSU)

Sami El-Ferik (KFUPM)
M. Abdelkader (KAUST)

June 23, 2016 6th Symposium on System Structure and Control (SSSC2016)
 2

Outline
• Introduction
• Problem Statement
• Fault Tolerant Control Algorithm
• System Modeling
• Simulation Results

Fault Tolerant Flight Control Using Sliding Modes and Subspace SSSC’16
Identification-based Predictive Control
Introduction Literature Review Problem Statement SMC-MPC Algorithm System Modeling Simulations Conclusion 3

Flight Control Robust to

Faults/Uncertainties
• Aircraft can suffer fatal loss due to
▫ Structural damage
▫ Sensor malfunctioning
▫ Severe Weather Conditions
▫ Untuned Controller due to change in aircraft
dynamics
• Two approaches for Fault Tolerant FCS
▫ Robust control (SMC)
▫ Reconfigurable control (Identification
based MPC)
Fault Tolerant Flight Control Using Sliding Modes and Subspace SSSC’16
Identification-based Predictive Control
Introduction Literature Review Problem Statement SMC-MPC Algorithm System Modeling Simulations Conclusion 4

Fault Tolerant FCS Literature

• Multiple model-based adaptive estimation and
control [Maybeck 1991]. Model reference adaptive
control based on RLS parameter identification
[Shore 2005].
• Model predictive control (MPC) [Kale 2004].
• Multi-model fault detection and optimal control
allocation [Urnes 1990].
• Sliding Mode Control (SMC) based control
allocation [Edwards 2010]
Fault Tolerant Flight Control Using Sliding Modes and Subspace SSSC’16
Identification-based Predictive Control
 Problem
Introduction Literature Review
Statement
SMC-MPC Algorithm System Modeling Simulations Conclusion 5

Problem Statement
• Nonlinear Dynamics of Aircraft

Modelling uncertainty,
fault, disturbance etc Measurement Noise
• Physical Constraints on Under-actuated System

• For applying multi-variable SMC, consider a square subset of the output space
,such that the remaining outputs are stable
•
• The above requirement is not conservative if y­2(t) can be stabilized as it is common in
aerospace cascaded autopilot design. A slower outer loop for controlling y2(t) which
produces virtual commands in terms of y1(t) can serve as the desired trajectory for a
faster inner loop controlling y1(t). In such a case, the loops have to obey some time scale
separation.
Fault Tolerant Flight Control Using Sliding Modes and Subspace SSSC’16
Identification-based Predictive Control
Introduction Literature Review Problem Statement SMC-MPC Algorithm System Modeling Simulations Conclusion 6

Proposed Fault Tolerant Algorithm

Model Nominal
Identification Model

Model Sliding
Predictive Mode
Control Control Actuators Aircraft Dynamics
(Outer (Inner
Loop) Loop)

Denoising
Sensors
Filters

Fault Tolerant Flight Control Using Sliding Modes and Subspace SSSC’16
Identification-based Predictive Control
Introduction Literature Review Problem Statement SMC-MPC Algorithm System Modeling Simulations Conclusion 7

System Modeling
• Aircraft model used is the nonlinear model of an F-16, based on extensive
wind-tunnel tests, represented in polynomial form using global nonlinear
parametric modelling based on orthogonal functions.
• Control limits
• Inertial Measurement Unit for linear accelerations, 3σ = 0.06 g, Gyro
measurements for Euler’s angles, 3σ = 0.35°, Air Data Probe providing
measurements of angles of attack and sideslip, 3σ = 0.15° and forward
speed, 3σ=0.1m/s.
• For angular rates, we assumed military grade sensors providing 3σ = 1°/hr
[25]. The sensor noise was simulated as band-limited white noise with
correlation time Tc=10.5ms (much smaller than the system bandwidth).
• The aircraft is flying level initially at a pitch angle of 10°, at a speed of 160
kts and an altitude of 6km.

Fault Tolerant Flight Control Using Sliding Modes and Subspace SSSC’16
Identification-based Predictive Control
Introduction Literature Review Problem Statement SMC-MPC Algorithm System Modeling Simulations Conclusion 8

System Modeling

Fault Tolerant Flight Control Using Sliding Modes and Subspace SSSC’16
Identification-based Predictive Control
 System
Introduction Literature Review Problem Statement SMC-MPC Algorithm
Modeling
Simulations Conclusion 9

Nominal Model
• We will assume that the Param.
Table 2 Best Estimates of Parametric Values
% Error in Estimate Value (% of nominal values)

aerodynamic coefficients are xy

ay
x1 x2
-20.1 0.5
x3 x4
11.7 -5.3
x5 x6 x7
-8.9 10.7 -4.4 ----
x8

by
known with an accuracy of
9.4 2.5 1.5 -8.6 6.7 ---- ---- ----
cy -8.1 4.2 0.7 ---- ---- ---- ---- ----
dy -2.7 3.5 1.3 -6.2 ---- ---- ---- ----

20% only. ey
fy
0.7
-1.4
12.7
-7.0
6.3
2.6
-0.2
-5.9
----
4.1
---- ----
-0.1 ----
----
----
gy -3.5 -8.6 3.9 -1.5 17.8 ---- ---- ----
• This uncertainty may be hy
iy
-2.9
-2.2
-19.9 -8.2 8.7
-0.7 0.4 -2.9
9.9
----
0.5
----
-7.2 -5.8
---- ----

because of structural jy
ky
2.3
-4.0
16.2 -6.7 5.8
-4.2 4.4 0.2
-4.3 ---- ----
11.4 -2.4 -7.8 ----
----

-
damage, as it is ‘big’ enough ly

my
8.0 -10.7

-11.2 8.1
11.6
-
-14.6 3.7

-11.9 6.1
-7.7 -2.7 ----

6.8 6.2 9.1

10.5
to cater for quite off nominal ny
oy
-3.3
10.7
-0.4 -2.1 4.9
2.2 -9.0 -7.8
-7.7 0.1
-2.5 0.9
----
3.4
----
----

conditions py
qy
18.5
9.2
-17.5 0.6
-4.6
5.6
-3.7 ----
8.5
----
----
----
----
----
----
----
ry -7.5 3.5 0.4 1.3 -7.3 3.7 6.2 -4.2
r9,10 7.8 -3.8 ---- ---- ---- ---- ---- ----
sy -4.9 -1.1 9.4 0.4 -6.7 0.1 ---- ----
Fault Tolerant Flight Control Using Sliding Modes and Subspace SSSC’16
Identification-based Predictive Control
Introduction Literature Review Problem Statement SMC-MPC Algorithm System Modeling Simulations Conclusion 10

Inner Loop SMC

• The inner loop represents the controller for
tracking virtual commands in angular rates
y1=[p,q,r] produced by the outer loop.
• We define the sliding surfaces as
t t t
s (p,t)=p +0.1 pdτ , s (q,t) = q +0.3 qdτ , s (r,t) =r +0.25  r dτ
1 2 3
0 0 0
Deadzone for
chattering 1 s  1 
u = u eq +K  hgˆ  dsat(s)
-1 
Command  0 s 

Filter dsat(s)=  s   s  1 
 1 s  1  
Kp=Kr=0.4 and Kq=4 
pd (s) 4q d (s) 0.81rd (s) 
p(s)= , q(s)= , r (s)=
s 2 +2s+1 s 2 +3s+4 s 2 +1.8s+0.81

Fault Tolerant Flight Control Using Sliding Modes and Subspace SSSC’16
Identification-based Predictive Control
Introduction Literature Review Problem Statement SMC-MPC Algorithm System Modeling Simulations Conclusion 11

Equivalent Control u eq  λ , %
y1  = -  hgˆ 
-1
 h fˆ - y&1,d +λ%
y 1

a,eq 
 QSb 2  r(ˆj4   ˆj3   ˆj2   ˆj1  ˆj0 ) 
4 3 2 
  a %p  I xz pq / I x    

2

2VI x   p(iˆ3   ˆi2   ˆi1  ˆi0 )
3 2  

 
  I xz pq / I x  Iz qr  (I y  Iz )(pq  I xz qr / I x ) 
2

 2VI xz   I y  Iz 
  QSb r (qˆ 2   qˆ 1  qˆ 0 )   qr
2 2

 (I z  I xz / I x ) 
2
 Ix 
 QSb p(pˆ 4   pˆ 3   pˆ 2   pˆ 1  pˆ 0 ) 
2 4 3 2
 
 I x (rˆ7  3  rˆ6  2  rˆ1  rˆ0 )  I xz (kˆ 6 3  kˆ 3  2  kˆ 1  kˆ 0 ) 
QSbI xz  
(Iz  I xz / I x ) ˆ 3 ˆ 2 ˆ
2
I x 2 (Iz  I 2xz / I x )   (k 6   k 3   k1  kˆ 0 )


 I xz 

Fault Tolerant Flight Control Using Sliding Modes and Subspace SSSC’16
Identification-based Predictive Control
Introduction Literature Review Problem Statement SMC-MPC Algorithm System Modeling Simulations Conclusion 12

Close-Loop System Identification

• For
  MPC in outer loop, we must have a prediction model
for the inner loop closed loop plant.
• While the relationship between Euler angles <φ,θ,ψ>ε y1
and body angular rates <p,q,r>y1 is a well known nonlinear
kinematic relation , the relation with flow angles <α,β>y2
depends on the vehicle’s aerodynamics.
• Aircraft is persistently excited with PRBS inputs (pd,qd,rd).
• Using the N4SID (Numerical Algorithms for Subspace
State-Space System Identification) method, two discrete
state-space models were identified (θ0=α0=10°), one for
longitudinal mode, and another for lateral
Fault Tolerant Flight Control Using Sliding Modes and Subspace SSSC’16
Identification-based Predictive Control
Introduction Literature Review Problem Statement SMC-MPC Algorithm System Modeling Simulations Conclusion 13

Close-Loop System Identification-2

Fault Tolerant Flight Control Using Sliding Modes and Subspace SSSC’16
Identification-based Predictive Control
Introduction Literature Review Problem Statement SMC-MPC Algorithm System Modeling Simulations Conclusion 14

Outer Loop MPC Controller

• Even though the models identified are valid in the vicinity of the
initial conditions, due to the inherent robustness of MPC, the
models were seen to be adequate for the flight envelope, even for
aggressive maneuvers.
• The constraints placed on manipulated variables are |y1,d|
≤60°/s. Output variables were constrained at -10° ≤(α-α­0)≤35°
and |β| ≤ 5°.
• Weights on the inputs were Rc=1 for each input, while the
weights on outputs were 10 on each of φ and β, 50 for θ, 20 for α
• Prediction horizon was 1 sec, and the control horizon was 0.25
sec for both long/lat controllers.

Fault Tolerant Flight Control Using Sliding Modes and Subspace SSSC’16
Identification-based Predictive Control
Introduction Literature Review Problem Statement SMC-MPC Algorithm System Modeling Simulations Conclusion 15

Simulation Results
• Several simulations were performed to show robustness
and fault tolerance in the event of
▫ parametric uncertainty
▫ measurement noise
▫ severe wind turbulence
▫ strong gusts
▫ actuator/sensor faults
• Very aggressive combat-like maneuvers were
considered.

Fault Tolerant Flight Control Using Sliding Modes and Subspace SSSC’16
Identification-based Predictive Control
Introduction Literature Review Problem Statement SMC-MPC Algorithm System Modeling Simulations Conclusion 16

Simulation 1 – Air Combat Maneuver with

Parametric Uncertainty
• Minute long air-
combat-maneuvre
(ACM) involving 40°
banking reversals
followed by a pitch-
up to 45°, typical of
dog-fights, showing
robustness to
parametric
uncertainty and
measurement noise.

Fault Tolerant Flight Control Using Sliding Modes and Subspace SSSC’16
Identification-based Predictive Control
Introduction Literature Review Problem Statement SMC-MPC Algorithm System Modeling Simulations Conclusion 17

Simulation 2 – Severe Turbulence

Bank to bank reversals in severe wind turbulence of 3σ = 35
knots as specified in (MIL-F-8785C ) standards

Fault Tolerant Flight Control Using Sliding Modes and Subspace SSSC’16
Identification-based Predictive Control
Introduction Literature Review Problem Statement SMC-MPC Algorithm System Modeling Simulations Conclusion 18

Simulation 3 – Severe Cross-Wind and

Gusts
• FAR.25 specifications
for cross-wind and
gust tolerance are 25
knots
• We consider severe
gusts of 30 knots in
horizontal and vertical
direction and a severe
cross wind of 50 knots
during the 45° pitch-
up maneuver.
Fault Tolerant Flight Control Using Sliding Modes and Subspace SSSC’16
Identification-based Predictive Control
Introduction Literature Review Problem Statement SMC-MPC Algorithm System Modeling Simulations Conclusion 19

Simulation 4 – Sensor Fault

• Pitot-system for measuring
airspeed often malfunctions, and
is responsible for some major air
disasters.

• Effect of pitot blockage is

simulated by fixing sensor
readings α=10° and β=0°,
altitude reading to be fixed at
6km and airspeed indicator to
read a constant airspeed of 70
knots which is much below the
stall speed (110 knots), and 40%
of the actual speed (160 knots),
which are typical results of pitot
blockage

Fault Tolerant Flight Control Using Sliding Modes and Subspace SSSC’16
Identification-based Predictive Control
Introduction Literature Review Problem Statement SMC-MPC Algorithm System Modeling Simulations Conclusion 20

Simulation 5 – Control Surface Loss

• The level of fault tolerance is
inversely proportional to the usage
of the control surface for that
maneuver in the healthy aircraft’s
case.
• A 50% loss of control surface area
of all three surfaces, i.e. elevator,
rudder and aileron is considered.
• ACM task was achieved with the
same performance as the
undamaged case.
• Actuators were saturated for longer
periods, which suggest that
instability may occur if more
aggressive maneuvers, particularly
in severe gusts and turbulence are
attempted.

Fault Tolerant Flight Control Using Sliding Modes and Subspace SSSC’16
Identification-based Predictive Control
Introduction Literature Review Problem Statement SMC-MPC Algorithm System Modeling Simulations Conclusion 21

Thankyou
• Thankyou.
• You are welcome to question.
• [email protected]

Fault Tolerant Flight Control Using Sliding Modes and Subspace SSSC’16
Identification-based Predictive Control