This action might not be possible to undo. Are you sure you want to continue?
OF MARINE VEHICLES
Michael S. Triantafyllou
Franz S. Hover
Department of Ocean Engineering
Massachusetts Institute of Technology
Cambridge, Massachusetts USA
These notes were developed in the instruction of the MIT graduate subject
13.49: Maneuvering and Control of Surface and Underwater Vehicles. We plan
many enhancements; your comments are welcome!
Maneuvering and Control of Marine Vehicles
Latest Revision: January 11, 2002
c (Michael S. Triantafyllou and Franz S. Hover
Contents
1 MATH FACTS 1
1.1 Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Deﬁnition . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.2 Vector Magnitude . . . . . . . . . . . . . . . . . . . . . 2
1.1.3 Vector Dot Product . . . . . . . . . . . . . . . . . . . . 2
1.1.4 Vector Cross Product . . . . . . . . . . . . . . . . . . . 2
1.2 Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.1 Deﬁnition . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.2 Multiplying a Vector by a Matrix . . . . . . . . . . . . 3
1.2.3 Multiplying a Matrix by a Matrix . . . . . . . . . . . . 4
1.2.4 Common Matrices . . . . . . . . . . . . . . . . . . . . 4
1.2.5 Transpose . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.6 Determinant . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.7 Inverse . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2.8 Trace . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2.9 Eigenvalues and Eigenvectors . . . . . . . . . . . . . . 7
1.2.10 Modal Decomposition . . . . . . . . . . . . . . . . . . 9
1.2.11 Singular Value . . . . . . . . . . . . . . . . . . . . . . . 9
1.3 Laplace Transform . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3.1 Deﬁnition . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3.2 Convergence . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3.3 Convolution Theorem . . . . . . . . . . . . . . . . . . . 11
1.3.4 Solution of Diﬀerential Equations by Laplace Transform 13
2 KINEMATICS OF MOVING FRAMES 14
2.1 Rotation of Reference Frames . . . . . . . . . . . . . . . . . . 14
2.2 Diﬀerential Rotations . . . . . . . . . . . . . . . . . . . . . . . 16
2.3 Rate of Change of Euler Angles . . . . . . . . . . . . . . . . . 18
2.4 Dead Reckoning . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3 VESSEL INERTIAL DYNAMICS 20
3.1 Momentum of a Particle . . . . . . . . . . . . . . . . . . . . . 20
3.2 Linear Momentum in a Moving Frame . . . . . . . . . . . . . 21
3.3 Example: Mass on a String . . . . . . . . . . . . . . . . . . . 22
3.3.1 Moving Frame Aﬃxed to Mass . . . . . . . . . . . . . 23
3.3.2 Rotating Frame Attached to Pivot Point . . . . . . . . 23
i
3.3.3 Stationary Frame . . . . . . . . . . . . . . . . . . . . . 24
3.4 Angular Momentum . . . . . . . . . . . . . . . . . . . . . . . 24
3.5 Example: Spinning Book . . . . . . . . . . . . . . . . . . . . . 26
3.5.1 xaxis . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.5.2 yaxis . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.5.3 zaxis . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.6 Parallel Axis Theorem . . . . . . . . . . . . . . . . . . . . . . 28
3.7 Basis for Simulation . . . . . . . . . . . . . . . . . . . . . . . 28
4 HYDRODYNAMICS: INTRODUCTION 30
4.1 Taylor Series and Hydrodynamic Coeﬃcients . . . . . . . . . . 30
4.2 Surface Vessel Linear Model . . . . . . . . . . . . . . . . . . . 31
4.3 Stability of the Sway/Yaw System . . . . . . . . . . . . . . . . 32
4.4 Basic Rudder Action in the Sway/Yaw Model . . . . . . . . . 34
4.4.1 Adding Yaw Damping through Feedback . . . . . . . . 36
4.4.2 Heading Control in the Sway/Yaw Model . . . . . . . . 36
4.5 Response of the Vessel to Step Rudder Input . . . . . . . . . . 37
4.5.1 Phase 1: Accelerations Dominate . . . . . . . . . . . . 37
4.5.2 Phase 3: Steady State . . . . . . . . . . . . . . . . . . 37
4.6 Summary of the Linear Maneuvering Model . . . . . . . . . . 38
4.7 Stability in the Vertical Plane . . . . . . . . . . . . . . . . . . 38
5 SIMILITUDE 40
5.1 Use of Nondimensional Groups . . . . . . . . . . . . . . . . . 40
5.2 Common Groups in Marine Engineering . . . . . . . . . . . . 42
5.3 Similitude in Maneuvering . . . . . . . . . . . . . . . . . . . . 44
5.4 Roll Equation Similitude . . . . . . . . . . . . . . . . . . . . . 46
6 CAPTIVE MEASUREMENTS 48
6.1 Towtank . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
6.2 Rotating Arm Device . . . . . . . . . . . . . . . . . . . . . . . 48
6.3 PlanarMotion Mechanism . . . . . . . . . . . . . . . . . . . . 49
7 STANDARD MANEUVERING TESTS 52
7.1 Dieudonn´e Spiral . . . . . . . . . . . . . . . . . . . . . . . . . 52
7.2 ZigZag Maneuver . . . . . . . . . . . . . . . . . . . . . . . . . 52
7.3 Circle Maneuver . . . . . . . . . . . . . . . . . . . . . . . . . . 53
7.3.1 Drift Angle . . . . . . . . . . . . . . . . . . . . . . . . 53
ii
7.3.2 Speed Loss . . . . . . . . . . . . . . . . . . . . . . . . 53
7.3.3 Heel Angle . . . . . . . . . . . . . . . . . . . . . . . . . 53
7.3.4 Heeling in Submarines with Sails . . . . . . . . . . . . 54
8 STREAMLINED BODIES 56
8.1 Nominal Drag Force . . . . . . . . . . . . . . . . . . . . . . . 56
8.2 Munk Moment . . . . . . . . . . . . . . . . . . . . . . . . . . 56
8.3 Separation Moment . . . . . . . . . . . . . . . . . . . . . . . . 57
8.4 Net Eﬀects: Aerodynamic Center . . . . . . . . . . . . . . . . 57
8.5 Role of Fins in Moving the Aerodynamic Center . . . . . . . . 58
8.6 Aggregate Eﬀects of Body and Fins . . . . . . . . . . . . . . . 60
8.7 Coeﬃcients Z
w
, M
w
, Z
q
, and M
q
for a Slender Body . . . . . . 60
9 SLENDERBODY THEORY 62
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
9.2 Kinematics Following the Fluid . . . . . . . . . . . . . . . . . 62
9.3 Derivative Following the Fluid . . . . . . . . . . . . . . . . . . 63
9.4 Diﬀerential Force on the Body . . . . . . . . . . . . . . . . . . 64
9.5 Total Force on a Vessel . . . . . . . . . . . . . . . . . . . . . . 64
9.6 Total Moment on a Vessel . . . . . . . . . . . . . . . . . . . . 65
9.7 Relation to Wing Lift . . . . . . . . . . . . . . . . . . . . . . . 67
9.8 Convention: Hydrodynamic Mass Matrix A . . . . . . . . . . 67
10 PRACTICAL LIFT CALCULATIONS 68
10.1 Characteristics of LiftProducing Mechanisms . . . . . . . . . 68
10.2 Jorgensen’s Formulas . . . . . . . . . . . . . . . . . . . . . . . 68
10.3 Hoerner’s Data: Notation . . . . . . . . . . . . . . . . . . . . 70
10.4 SlenderBody Theory vs. Experiment . . . . . . . . . . . . . . 71
10.5 SlenderBody Approximation for Fin Lift . . . . . . . . . . . . 73
11 FINS AND LIFTING SURFACES 74
11.1 Origin of Lift . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
11.2 ThreeDimensional Eﬀects: Finite Length . . . . . . . . . . . . 74
11.3 Ring Fins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
12 PROPELLERS AND PROPULSION 77
12.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
12.2 Steady Propulsion of Vessels . . . . . . . . . . . . . . . . . . . 77
12.2.1 Basic Characteristics . . . . . . . . . . . . . . . . . . . 77
iii
12.2.2 Solution for Steady Conditions . . . . . . . . . . . . . 81
12.2.3 Engine/Motor Models . . . . . . . . . . . . . . . . . . 81
12.3 Unsteady Propulsion Models . . . . . . . . . . . . . . . . . . . 83
12.3.1 OneState Model: Yoerger et al. . . . . . . . . . . . . . 83
12.3.2 TwoState Model: Healey et al. . . . . . . . . . . . . . 84
13 TOWING OF VEHICLES 85
13.1 Statics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
13.1.1 Force Balance . . . . . . . . . . . . . . . . . . . . . . . 85
13.1.2 Critical Angle . . . . . . . . . . . . . . . . . . . . . . . 87
13.2 Linearized Dynamics . . . . . . . . . . . . . . . . . . . . . . . 89
13.2.1 Derivation . . . . . . . . . . . . . . . . . . . . . . . . . 89
13.2.2 Damped Axial Motion . . . . . . . . . . . . . . . . . . 91
13.3 Cable Strumming . . . . . . . . . . . . . . . . . . . . . . . . . 94
13.4 Vehicle Design . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
14 TRANSFER FUNCTIONS & STABILITY 96
14.1 Partial Fractions . . . . . . . . . . . . . . . . . . . . . . . . . 96
14.2 Partial Fractions: Unique Poles . . . . . . . . . . . . . . . . . 96
14.3 Example: Partial Fractions with Unique Real Poles . . . . . . 97
14.4 Partial Fractions: ComplexConjugate Poles . . . . . . . . . . 98
14.5 Example: Partial Fractions with Complex Poles . . . . . . . . 98
14.6 Stability in Linear Systems . . . . . . . . . . . . . . . . . . . . 98
14.7 Stability ⇐⇒ Poles in LHP . . . . . . . . . . . . . . . . . . . 99
14.8 General Stability . . . . . . . . . . . . . . . . . . . . . . . . . 99
15 CONTROL FUNDAMENTALS 100
15.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
15.1.1 Plants, Inputs, and Outputs . . . . . . . . . . . . . . . 100
15.1.2 The Need for Modeling . . . . . . . . . . . . . . . . . . 100
15.1.3 Nonlinear Control . . . . . . . . . . . . . . . . . . . . . 101
15.2 Representing Linear Systems . . . . . . . . . . . . . . . . . . . 101
15.2.1 Standard StateSpace Form . . . . . . . . . . . . . . . 101
15.2.2 Converting a StateSpace Model into a Transfer Function 102
15.2.3 Converting a Transfer Function into a StateSpace Model 102
15.3 PID Controllers . . . . . . . . . . . . . . . . . . . . . . . . . . 103
15.4 Example: PID Control . . . . . . . . . . . . . . . . . . . . . . 103
15.4.1 Proportional Only . . . . . . . . . . . . . . . . . . . . 104
iv
15.4.2 ProportionalDerivative Only . . . . . . . . . . . . . . 104
15.4.3 ProportionalIntegralDerivative . . . . . . . . . . . . . 105
15.5 Heuristic Tuning . . . . . . . . . . . . . . . . . . . . . . . . . 105
15.6 Block Diagrams of Systems . . . . . . . . . . . . . . . . . . . . 106
15.6.1 Fundamental Feedback Loop . . . . . . . . . . . . . . . 106
15.6.2 Block Diagrams: General Case . . . . . . . . . . . . . . 106
15.6.3 Primary Transfer Functions . . . . . . . . . . . . . . . 107
16 MODAL ANALYSIS 109
16.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
16.2 Matrix Exponential . . . . . . . . . . . . . . . . . . . . . . . . 109
16.2.1 Deﬁnition . . . . . . . . . . . . . . . . . . . . . . . . . 109
16.2.2 Modal Canonical Form . . . . . . . . . . . . . . . . . . 109
16.2.3 Modal Decomposition of Response . . . . . . . . . . . 110
16.3 Forced Response and Controllability . . . . . . . . . . . . . . 111
16.4 Plant Output and Observability . . . . . . . . . . . . . . . . . 112
17 CONTROL SYSTEMS – LOOPSHAPING 113
17.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
17.2 Roots of Stability – Nyquist Criterion . . . . . . . . . . . . . . 113
17.2.1 Mapping Theorem . . . . . . . . . . . . . . . . . . . . 114
17.2.2 Nyquist Criterion . . . . . . . . . . . . . . . . . . . . . 114
17.2.3 Robustness on the Nyquist Plot . . . . . . . . . . . . . 115
17.3 Design for Nominal Performance . . . . . . . . . . . . . . . . . 116
17.4 Design for Robustness . . . . . . . . . . . . . . . . . . . . . . 116
17.5 Robust Performance . . . . . . . . . . . . . . . . . . . . . . . 118
17.6 Implications of Bode’s Integral . . . . . . . . . . . . . . . . . . 118
17.7 The Recipe for Loopshaping . . . . . . . . . . . . . . . . . . . 119
18 LINEAR QUADRATIC REGULATOR 121
18.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
18.2 FullState Feedback . . . . . . . . . . . . . . . . . . . . . . . . 121
18.3 Dynamic Programming . . . . . . . . . . . . . . . . . . . . . . 121
18.4 Dynamic Programming and FullState Feedback . . . . . . . . 123
18.5 Properties and Use of the LQR . . . . . . . . . . . . . . . . . 125
v
19 KALMAN FILTER 128
19.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
19.2 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . 128
19.3 Step 1: An Equation for
˙
Σ . . . . . . . . . . . . . . . . . . . . 129
19.4 Step 2: H as a Function of Σ . . . . . . . . . . . . . . . . . . 131
19.5 Properties of the Solution . . . . . . . . . . . . . . . . . . . . 132
19.6 Combination of LQR and KF . . . . . . . . . . . . . . . . . . 133
19.7 Proofs of the Intermediate Results . . . . . . . . . . . . . . . . 134
19.7.1 Proof that E(e
T
We) = trace(ΣW) . . . . . . . . . . . 134
19.7.2 Proof that
∂
∂H
trace(−ΛHCΣ) = −Λ
T
ΣC
T
. . . . . . . 135
19.7.3 Proof that
∂
∂H
trace(−ΛΣC
T
H
T
) = −ΛΣC
T
. . . . . . 135
19.7.4 Proof of the Separation Principle . . . . . . . . . . . . 136
20 LOOP TRANSFER RECOVERY 137
20.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
20.2 A Special Property of the LQR Solution . . . . . . . . . . . . 138
20.3 The Loop Transfer Recovery Result . . . . . . . . . . . . . . . 139
20.4 Usage of the Loop Transfer Recovery . . . . . . . . . . . . . . 141
20.5 Three Lemmas . . . . . . . . . . . . . . . . . . . . . . . . . . 141
21 SYSTEM IDENTIFICATION 143
21.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
21.2 Visual Output from a Simple Input . . . . . . . . . . . . . . . 143
21.3 Transfer Function Estimation – Sinusoidal Input . . . . . . . . 145
21.4 Transfer Function Estimation – Broadband Input . . . . . . . 146
21.4.1 Fourier Transform of Sampled Data . . . . . . . . . . . 147
21.4.2 Estimating the Transfer Function . . . . . . . . . . . . 148
21.5 TimeDomain Simulation . . . . . . . . . . . . . . . . . . . . . 150
22 CARTESIAN NAVIGATION 153
22.1 Acoustic Navigation . . . . . . . . . . . . . . . . . . . . . . . 153
22.2 Global Positioning System (GPS) . . . . . . . . . . . . . . . . 155
23 REFERENCES 159
24 PROBLEMS 163
vi
1
1 MATH FACTS
1.1 Vectors
1.1.1 Deﬁnition
We use the overhead arrow to denote a column vector, i.e., a number with a
direction. For example, in threespace, we write
a =
2
1
7
.
The elements of a vector have a graphical interpretation, which is particularly
easy to see in two or three dimensions.
2
x
y
1
7
z
a
1. Vector addition is pointwise.
a +
b = c
2
1
7
+
3
3
2
=
5
4
9
.
Graphically, addition is stringing the vectors together head to tail.
2 1 MATH FACTS
2. Scalar multiplication is pointwise.
−2
2
1
7
=
−4
−2
−14
.
1.1.2 Vector Magnitude
The total length of a vector of dimension m, its Euclidean norm, is given by
[[x[[ =
m
¸
i=1
x
2
i
;
this scalar is commonly used to normalize a vector to length one.
1.1.3 Vector Dot Product
The dot product of two vectors is the sum of the products of the elements:
x y = x
T
y =
m
¸
i=1
x
i
y
i
.
The dot product also satisﬁes
x y = [[x[[[[y[[ cos θ,
where θ is the angle between the vectors.
1.1.4 Vector Cross Product
The cross product of two threedimensional vectors is another vector, xy = z,
whose
1. direction is normal to the plane formed by the two vectors,
2. direction is given by the righthand rule, rotating from x to y,
1.2 Matrices 3
3. magnitude is the area of the parallelogram formed by the two vectors –
the cross product of two parallel vectors is zero – and
4. (signed) magnitude is equal to [[x[[[[y[[ sin θ, where θ is the angle between
the two vectors, measured from x to y.
The schoolbook formula is
x y =
x
2
y
3
−x
3
y
2
x
3
y
1
−x
1
y
3
x
1
y
2
−x
2
y
1
.
1.2 Matrices
1.2.1 Deﬁnition
A matrix, or array, is equivalent to a set of row vectors, arranged side by side,
say
A = [a
b] =
2 3
1 3
7 2
¸
¸
¸ .
This matrix has three rows (m = 3) and two columns (n = 2); a vector is
a special case of a matrix with one column. Matrices, like vectors, permit
pointwise addition and scalar multiplication. We usually use an uppercase
symbol to denote a matrix.
1.2.2 Multiplying a Vector by a Matrix
If A
ij
denotes the element of matrix A in the i’th row and the j’th column,
then the multiplication c = Av is constructed as:
c
i
= A
i1
v
1
+ A
i2
v
2
+ + A
in
v
n
=
n
¸
j=1
A
ij
v
j
,
where n is the number of columns in A. c will have as many columns as
A has rows (m). Note that this multiplication is welldeﬁned only if v has
4 1 MATH FACTS
as many rows as A has columns; they have consistent inner dimension n.
The product vA would be wellposed only if A had one row, and the proper
number of columns. There is another important interpretation of this vector
multiplication: Let the subscript : indicate all rows, so that each A
:j
is the
j’th column vector. Then
c = Av = A
:1
v
1
+ A
:2
v
2
+ + A
:n
v
n
.
We are multiplying column vectors of A by the scalar elements of v.
1.2.3 Multiplying a Matrix by a Matrix
The multiplication C = AB is equivalent to a sidebyside arrangement of
column vectors C
:j
= AB
:j
, so that
C = AB = [AB
:1
AB
:2
AB
:k
],
where k is the number of columns in matrix B. The same inner dimension
condition applies as noted above: the number of columns in A must equal the
number of rows in B. Matrix multiplication is:
1. Associative. (AB)C = A(BC).
2. Distributive. A(B + C) = AB + AC, (B + C)A = BA + CA.
3. NOT Commutative. AB = BA, except in special cases.
1.2.4 Common Matrices
Identity . The identity matrix is usually denoted I, and comprises a square
matrix with ones on the diagonal, and zeros elsewhere, e.g.,
I
3×3
=
1 0 0
0 1 0
0 0 1
¸
¸
¸ .
The identity always satisﬁes AI
n×n
= I
m×m
A = A.
1.2 Matrices 5
Diagonal Matrices . A diagonal matrix is square, and has all zeros oﬀ the
diagonal. For instance, the following is a diagonal matrix:
A =
4 0 0
0 −2 0
0 0 3
¸
¸
¸ .
The product of a diagonal matrix with another diagonal matrix is diagonal,
and in this case the operation is commutative.
1.2.5 Transpose
The transpose of a vector or matrix, indicated by a T superscript results
from simply swapping the rowcolumn indices of each entry; it is equivalent to
“ﬂipping” the vector or matrix around the diagonal line. For example,
a =
1
2
3
−→a
T
= ¦1 2 3¦
A =
1 2
4 5
8 9
¸
¸
¸ −→A
T
=
¸
1 4 8
2 5 9
¸
.
A very useful property of the transpose is
(AB)
T
= B
T
A
T
.
1.2.6 Determinant
The determinant of a square matrix A is a scalar equal to the volume of the
parallelepiped enclosed by the constituent vectors. The twodimensional case
is particularly easy to remember, and illustrates the principle of volume:
det(A) = A
11
A
22
−A
21
A
12
det
¸
1 −1
1 1
¸
= 1 + 1 = 2.
6 1 MATH FACTS
1 1
A
:2
A
:1
Area = det(A) = 2
y
x
In higher dimensions, the determinant is more complicated to compute. The
general formula allows one to pick a row k, perhaps the one containing the
most zeros, and apply
det(A) =
j=n
¸
j=1
A
kj
(−1)
k+j
∆
kj
,
where ∆
kj
is the determinant of the submatrix formed by neglecting the k’th
row and the j’th column. The formula is symmetric, in the sense that one
could also target the k’th column:
det(A) =
j=n
¸
j=1
A
jk
(−1)
k+j
∆
jk
.
If the determinant of a matrix is zero, then the matrix is said to be singular –
there is no volume, and this results from the fact that the constituent vectors
do not span the matrix dimension. For instance, in two dimensions, a singular
matrix has the vectors colinear; in three dimensions, a singular matrix has
all its vectors lying in a (twodimensional) plane. Note also that det(A) =
det(A
T
). If det(A) = 0, then the matrix is said to be nonsingular.
1.2.7 Inverse
The inverse of a square matrix A, denoted A
−1
, satisﬁes AA
−1
= A
−1
A = I.
Its computation requires the determinant above, and the following deﬁnition
of the n n adjoint matrix:
adj(A) =
(−1)
1+1
∆
11
(−1)
1+n
∆
1n
(−1)
n+1
∆
n1
(−1)
n+n
∆
nn
.
¸
¸
¸
T
.
1.2 Matrices 7
Once this computation is made, the inverse follows from
A
−1
=
adj(A)
det(A)
.
If A is singular, i.e., det(A) = 0, then the inverse does not exist. The inverse
ﬁnds common application in solving systems of linear equations such as
Ax =
b −→x = A
−1
b.
1.2.8 Trace
The trace of a matrix is simply the sum of the diagonals:
tr(A) =
n
¸
i=1
A
ii
.
1.2.9 Eigenvalues and Eigenvectors
A typical eigenvalue problem is stated as
Ax = λx,
where A is an n n matrix, x is a column vector with n elements, and λ is
a scalar. We ask for what nonzero vectors x (right eigenvectors), and scalars
λ (eigenvalues) will the equation be satisﬁed. Since the above is equivalent to
(A−λI)x =
0, it is clear that det(A−λI) = 0. This observation leads to the
solutions for λ; here is an example for the twodimensional case:
A =
¸
4 −5
2 −3
¸
−→
A −λI =
¸
4 −λ −5
2 −3 −λ
¸
−→
det(A −λI) = (4 −λ)(−3 −λ) + 10
= λ
2
−λ −2
= (λ + 1)(λ −2).
8 1 MATH FACTS
Thus, A has two eigenvalues, λ
1
= −1 and λ
2
= 2. Each is associated with a
right eigenvector x. In this example,
(A −λ
1
I)x
1
=
0 −→
¸
5 −5
2 −2
¸
x
1
=
0 −→
x
1
=
√
2/2,
√
2/2
¸
T
.
(A −λ
2
I)x
2
=
0 −→
¸
2 −5
2 −5
¸
x
2
=
0 −→
x
2
=
5
√
29/29, 2
√
29/29
¸
T
.
Eigenvectors have arbitrary magnitude and sign; they are often normalized to
have unity magnitude, and positive ﬁrst element (as above). A set of n eigen
vectors is always linearly independent. The condition that rank(A − λ
i
I) =
rank(A)−1 indicates that there is only one eigenvector for the eigenvalue λ
i
. If
the lefthand side is less than this, then there are multiple unique eigenvectors
that go with λ
i
.
The above discussion relates only the right eigenvectors, generated from the
equation Ax = λx. Left eigenvectors, also useful for many problems, pertain
to the transpose of A: A
T
y = λy. A and A
T
share the same eigenvalues λ,
since they share the same determinant. Example:
(A
T
−λ
1
I)y
1
=
0 −→
¸
5 2
−5 −2
¸
y
1
=
0 −→
y
1
=
2
√
29/29, −5
√
29/29
¸
T
.
(A
T
−λ
2
I)y
2
=
0 −→
¸
2 2
−5 −5
¸
y
2
=
0 −→
y
2
=
√
2/2, −
√
2/2
¸
T
.
1.2 Matrices 9
1.2.10 Modal Decomposition
The right and left eigenvectors of a particular eigenvalue have unity dot prod
uct, that is x
T
i
y
i
= 1, with the normalization noted above. The dot product
of a left eigenvector with the right eigenvector of a diﬀerent eigenvalue is zero.
Thus, if
X = [x
1
x
n
] , and
Y = [y
1
y
n
] ,
then we have
Y
T
X = I, or
Y
T
= X
−1
.
Next, construct a diagonal matrix of eigenvalues:
Λ =
λ
1
0
0 λ
n
¸
¸
¸ ;
it follows that
AV = V Λ −→
A = V ΛW
T
=
n
¸
i=1
λ
i
v
i
w
T
i
.
Hence A can be written as a sum of modal components.
1
1.2.11 Singular Value
Let G(s) be an mn, possibly complex matrix. The singular value decompo
sition (SVD) computes three matrices satisfying
1
By carrying out successive multiplications, it can be shown that A
k
has its eigenvalues
at λ
k
i
, and keeps the same eigenvectors as A.
10 1 MATH FACTS
G = UΣV
∗
,
where U is mm, Σ is mn, and V is n n. The star notation indicates a
complexconjugate transpose. The matrix Σ is diagonal, with the form
Σ =
σ
1
0 0 0
0 0 0
0 0 σ
p
0
0 0 0 0
¸
¸
¸
¸
¸
,
where p = min(m, n). Each nonzero entry on the diagonal is a real, positive
singular value, ordered such that σ
1
> σ
2
> σ
p
. The notation is common
that σ
1
= σ, the maximum singular value, and σ
p
= σ, the minimum singular
value. The auxiliary matrices U and V are unitary, i.e., they satisfy X
∗
=
X
−1
. Like eigenvalues, the singular values of G are related to projections. σ
i
represents the Euclidean size of the matrix G along the i’th singular vector:
σ = max
x=1
[[Gx[[
σ = min
x=1
[[Gx[[.
Other properties of the singular value include:
• σ(AB) ≤ σ(A)σ(B).
• σ(A) =
λ
max
(A
∗
A).
• σ(A) =
λ
min
(A
∗
A).
• σ(A) = 1/σ(A
−1
).
• σ(A) = 1/σ(A
−1
).
1.3 Laplace Transform 11
1.3 Laplace Transform
1.3.1 Deﬁnition
The Laplace transform converts timedomain signals into a frequencydomain
equivalent. The signal y(t) has transform Y (s) deﬁned as follows:
Y (s) = L(y(t)) =
∞
0
y(τ)e
−sτ
dτ,
where s is an unspeciﬁed complex number; Y (s) is considered to be complex as
a result. Note that the Laplace transform is linear, and so it is is distributive:
L(x(t) + y(t)) = L(x(t)) + L(y(t)) will hold throughout. The following table
gives a list of some useful transform pairs and other properties, for reference.
The last two properties are of special importance: for control system design,
the diﬀerentiation of a signal is equivalent to multiplication of its Laplace
transform by s; integration of a signal is equivalent to division by s. The other
terms that arise will cancel if y(0) = 0, or if y(0) is ﬁnite, so they are usually
ignored.
1.3.2 Convergence
We note ﬁrst that the value of s aﬀects the convergence of the integral. For
instance, if y(t) = e
t
, then the integral converges only for Re(s) > 1, since
the integrand is e
1−s
in this case. Convergence issues are not a problem for
evaluation of the Laplace transform, however, because of analytic continuation.
This result from complex analysis holds that if two complex functions are
equal on some arc (or line) in the complex plane, then they are equivalent
everywhere. This fact allows us to always pick a value of s for which the
integral above converges, and then by extension infer the existence of the
general transform.
1.3.3 Convolution Theorem
One of the main points of the Laplace transform is the ease of dealing with
dynamic systems. As with the Fourier transform, the convolution of two sig
nals in the time domain corresponds with the multiplication of signals in the
frequency domain. Consider a system whose impulse response is g(t), being
12 1 MATH FACTS
y(t) ←→ Y (s)
(Impulse) δ(t) ←→ 1
(Unit Step) 1(t) ←→
1
s
(Unit Ramp) t ←→
1
s
2
e
−αt
←→
1
s + α
sin ωt ←→
ω
s
2
+ ω
2
cos ωt ←→
s
s
2
+ ω
2
e
−αt
sin ωt ←→
ω
(s + α)
2
+ ω
2
e
−αt
cos ωt ←→
s + α
(s + α)
2
+ ω
2
1
b −a
(e
−at
−e
−bt
) ←→
1
(s + a)(s + b)
1
ab
¸
1 +
1
a −b
(be
−at
−ae
−bt
)
←→
1
s(s + a)(s + b)
ω
n
√
1 −ζ
2
e
−ζωnt
sin ω
n
1 −ζ
2
t ←→
ω
2
n
s
2
+ 2ζω
n
s + ω
2
n
1 −
1
√
1 −ζ
2
e
−ζωnt
sin
ω
n
1 −ζ
2
t + φ
←→
ω
2
n
s(s
2
+ 2ζω
n
s + ω
2
n
)
φ = tan
−1
√
1 −ζ
2
ζ
(Pure Delay) y(t −τ)1(t −τ) ←→ Y (s)e
−sτ
(Time Derivative)
dy(t)
dt
←→ sY (s) −y(0)
(Time Integral)
t
0
y(τ)dτ ←→
Y (s)
s
+
0+
0−
y(t)dt
s
1.3 Laplace Transform 13
driven by an input signal x(t); the output is y(t) = g(t)∗x(t). The Convolution
Theorem is
y(t) =
t
0
g(t −τ)x(τ)dτ ⇐⇒Y (s) = G(s)X(s).
Here’s the proof given by Siebert:
Y (s) =
∞
0
y(t)e
−st
dt
=
∞
0
¸
t
0
g(t −τ) x(τ) dτ
e
−st
dt
=
∞
0
¸
∞
0
g(t −τ) 1(t −τ) x(τ) dτ
e
−st
dt
=
∞
0
x(τ)
¸
∞
0
g(t −τ) 1(t −τ) e
−st
dt
dτ
=
∞
0
x(τ) G(s)e
−sτ
dτ
= G(s)X(s)
When g(t) is the impulse response of a dynamic system, then y(t) represents
the output of this system when it is driven by the external signal x(t).
1.3.4 Solution of Diﬀerential Equations by Laplace Transform
The Convolution Theorem allows one to solve (linear timeinvariant) diﬀeren
tial equations in the following way:
1. Transform the system impulse response g(t) into G(s), and the input
signal x(t) into X(s), using the transform pairs.
2. Perform the multiplication in the Laplace domain to ﬁnd Y (s).
3. Ignoring the eﬀects of pure time delays, break Y (s) into partial fractions
with no powers of s greater than 2 in the denominator.
4. Generate the timedomain response from the simple transform pairs.
Apply time delay as necessary.
Speciﬁc examples of this procedure are given in a later section on transfer
functions.
14 2 KINEMATICS OF MOVING FRAMES
2 KINEMATICS OF MOVING FRAMES
2.1 Rotation of Reference Frames
We say that a vector expressed in the inertial frame has coordinates x, and
in a bodyreference frame x
b
. For the moment, we assume that the origins of
these frames are coincident, but that the body frame has a diﬀerent angular
orientation. The angular orientation has several wellknown descriptions, in
cluding the Euler angles and the Euler parameters (quaternions). The former
method involves successive rotations about the principle axes, and has a solid
link with the intuitive notions of roll, pitch, and yaw. Quaternions present a
more elegant and robust method, but with more abstraction. We will develop
the equations of motion using Euler angles.
Tape three pencils together to form a righthanded threedimensional coordi
nate system. Successively rotating the system about three of its own principle
axes, it is easy to see that any possible orientation can be achieved. For ex
ample, consider the sequence of [yaw, pitch, roll]: starting from an orientation
identical to some inertial frame, rotate the movable system about its yaw axis,
then about the new pitch axis, then about the newer still roll axis. Needless
to say, there are many valid Euler angle rotation sets possible to reach a given
orientation; some of them might use the same axis twice.
x
x’
y’=y’’
y
x’’=x’’’
z’’
z’’’
y’’’
z=z’
Figure 1: Successive application of three Euler angles transforms the original
coordinate frame into an arbitrary orientation.
A ﬁrst question is: what is the coordinate of a point ﬁxed in inertial space,
referenced to a rotated body frame? The transformation takes the form of
a 33 matrix, which we now derive through successive rotations of the three
2.1 Rotation of Reference Frames 15
Euler angles. Before the ﬁrst rotation, the bodyreferenced coordinate matches
that of the inertial frame: x
0
b
= x. Now rotate the movable frame yaw axis
(z) through an angle φ. We have
x
1
b
=
cos φ sin φ 0
−sin φ cos φ 0
0 0 1
¸
¸
¸x
0
b
= R(φ)x
0
b
. (1)
Rotation about the zaxis does not change the zcoordinate of the point; the
other axes are modiﬁed according to basic trigonometry. Now apply the second
rotation, pitch about the new yaxis by the angle θ:
x
2
b
=
cos θ 0 −sin θ
0 1 0
sin θ 0 cos θ
¸
¸
¸x
1
b
= R(θ)x
1
b
. (2)
Finally, rotate the body system an angle ψ about its newest xaxis:
x
3
b
=
1 0 0
0 cos ψ sin ψ0
0 −sin ψ cos ψ
¸
¸
¸x
2
b
= R(ψ)x
2
b
. (3)
This represents the location of the original point, in the fullytransformed
bodyreference frame, i.e., x
3
b
. We will use the notation x
b
instead of x
3
b
from
here on. The three independent rotations can be cascaded through matrix
multiplication (order matters!):
x
b
= R(ψ)R(θ)R(φ)x (4)
=
cθcφ cθsφ −sθ
−cψsψ + sψsθcφ cψcφ + sψsθsφ sψcθ
sψsφ + cψsθcφ −sψcφ + cψsθsψ cψcθ
¸
¸
¸x
= R(φ, θ, ψ)x.
All of the transformation matrices, including R(φ, θ, ψ), are orthonormal: their
inverse is equivalent to their transpose. Additionally, we should note that the
rotation matrix R is universal to all representations of orientation, including
quaternions. The roles of the trigonometric functions, as written, are speciﬁc
to Euler angles, and to the order in which we performed the rotations.
In the case that the movable (body) reference frame has a diﬀerent origin than
the inertial frame, we have
16 2 KINEMATICS OF MOVING FRAMES
x = x
0
+ R
T
x
b
, (5)
where x
0
is the location of the moving origin, expressed in inertial coordinates.
2.2 Diﬀerential Rotations
Now consider small rotations from one frame to another; using the small angle
assumption to ignore higherorder terms gives
R ·
1 δφ −δθ
−δφ 1 δψ
δθ −δψ 1
¸
¸
¸ (6)
=
0 δφ −δθ
−δφ 0 δψ
δθ −δψ 0
¸
¸
¸ + I
3×3
.
R comprises the identity plus a part equal to the (negative) crossproduct
operator [−δ
E], where δ
E = [δψ, δθ, δφ], the vector of Euler angles ordered
with the axes [x, y, z]. Small rotations are completely decoupled; the order of
the small rotations does not matter. Since R
−1
= R
T
, we have also R
−1
=
I
3×3
+ δ
E;
x
b
= x −δ
E x (7)
x = x
b
+ δ
E x
b
. (8)
We now ﬁx the point of interest on the body, instead of in inertial space,
calling its location in the body frame r (radius). The diﬀerential rotations
occur over a time step δt, so that we can write the location of the point before
and after the rotation, with respect to the ﬁrst frame as follows:
x(t) = r (9)
x(t + δt) = R
T
r = r + δ
E r.
Dividing by the diﬀerential time step gives
2.2 Diﬀerential Rotations 17
δx
δt
=
δ
E
δt
r (10)
= ω r,
where the rotation rate vector ω · d
E/dt because the Euler angles for this
inﬁnitesimal rotation are small and decoupled. This same crossproduct rela
tionship can be derived in the second frame as well:
x
b
(t) = Rr = r −δ
E r (11)
x
b
(t + δt) = r.
such that
δx
b
δt
=
δ
E
δt
r (12)
= ω r,
On a rotating body whose origin point is ﬁxed, the time rate of change of a
constant radius vector is the crossproduct of the rotation rate vector ω and
the radius vector itself. The resultant derivative is in the moving body frame.
In the case that the radius vector changes with respect to the body frame, we
need an additional term:
dx
b
dt
= ω r +
∂r
∂t
. (13)
Finally, allowing the origin to move as well gives
dx
b
dt
= ω r +
∂r
∂t
+
dx
o
dt
. (14)
This result is often written in terms of bodyreferenced velocity v:
v = ω r +
∂r
∂t
+v
o
, (15)
where v
o
is the bodyreferenced velocity of the origin. The total velocity of the
particle is equal to the velocity of the reference frame origin, plus a component
18 2 KINEMATICS OF MOVING FRAMES
due to rotation of this frame. The velocity equation can be generalized to any
bodyreferenced vector
f:
d
f
dt
=
∂f
∂t
+ ω
f. (16)
2.3 Rate of Change of Euler Angles
Only for the case of inﬁnitesimal Euler angles is it true that the time rate
of change of the Euler angles equals the bodyreferenced rotation rate. For
example, with the sequence [yaw,pitch,roll], the Euler yaw angle (applied ﬁrst)
is deﬁnitely not about the ﬁnal body yaw axis; the pitch and roll rotations
moved the axis. An important part of any simulation is the evolution of the
Euler angles. Since the physics determine rotation rate ω, we seek a mapping
ω →d
E/dt.
The idea is to consider small changes in each Euler angle, and determine the
eﬀects on the rotation vector. The ﬁrst Euler angle undergoes two additional
rotations, the second angle one rotation, and the ﬁnal Euler angle no additional
rotations:
ω = R(ψ)R(θ)
0
0
dφ
dt
+ R(ψ)
0
dθ
dt
0
+
dψ
dt
0
0
(17)
=
1 0 −sin θ
0 cos ψ sin ψ cos θ
0 −sin ψ cos ψ cos θ
¸
¸
¸
dψ
dt
dθ
dt
dφ
dt
.
Taking the inverse gives
d
E
dt
=
1 sin ψ tan θ cos ψ tan θ
0 cos ψ −sin ψ
0 sin ψ/ cos θ cos ψ/ cos θ
¸
¸
¸ ω (18)
= Γ(
E) ω.
Singularities exist in Γ at θ = ¦π/2, 3π/2¦, because of the division by cos θ, and
hence this otherwise useful equation for propagating the angular orientation of
a body fails when the vehicle rotates about the intermediate yaxis by ninety
2.4 Dead Reckoning 19
degrees. In applications where this is a real possibility, for example in orbiting
satellites and robotic arms, quaternions provide a seamless mapping. For most
ocean vessels, the singularity is acceptable, as long as it is not on the yaw axis!
2.4 Dead Reckoning
The measurement of heading and longitudinal speed gives rise to one of the
oldest methods of navigation: dead reckoning. Quite simply, if the estimated
longitudinal speed over ground is U, and the estimated heading is φ, ignoring
the lateral velocity leads to the evolution of Cartesian coordinates:
˙ x = U cos φ
˙ y = U sin φ.
Needless to say, currents and vehicle sideslip will cause this to be in error.
Nonetheless, some of the most remarkable feats of navigation in history have
depended on dead reckoning.
20 3 VESSEL INERTIAL DYNAMICS
3 VESSEL INERTIAL DYNAMICS
We consider the rigid body dynamics with a coordinate system aﬃxed on the
body. A common frame for ships, submarines, and other marine vehicles has
the bodyreferenced xaxis forward, yaxis to port (left), and zaxis up. This
will be the sense of our bodyreferenced coordinate system here.
3.1 Momentum of a Particle
Since the body moves with respect to an inertial frame, dynamics expressed in
the bodyreferenced frame need extra attention. First, linear momentum for
a particle obeys the equality
F =
d
dt
(mv) (19)
A rigid body consists of a large number of these small particles, which can be
indexed. The summations we use below can be generalized to integrals quite
easily. We have
F
i
+
R
i
=
d
dt
(m
i
v
i
) , (20)
where
F
i
is the external force acting on the particle and
R
i
is the net force
exerted by all the other surrounding particles (internal forces). Since the
collection of particles is not driven apart by the internal forces, we must have
equal and opposite internal forces such that
N
¸
i=1
R
i
= 0. (21)
Then summing up all the particle momentum equations gives
N
¸
i=1
F
i
=
N
¸
i=1
d
dt
(m
i
v
i
) . (22)
Note that the particle velocities are not independent, because the particles are
rigidly attached.
Now consider a body reference frame, with origin 0, in which the particle i
resides at bodyreferenced radius vector r; the body translates and rotates,
and we now consider how the momentum equation depends on this motion.
3.2 Linear Momentum in a Moving Frame 21
z, w,
y, v,
x, u,
θ
φ
ψ
Figure 2: Convention for the bodyreferenced coordinate system on a vessel:
x is forward, y is sway to the left, and z is heave upwards. Looking forward
from the vessel bridge, roll about the x axis is positive counterclockwise, pitch
about the yaxis is positive bowdown, and yaw about the zaxis is positive
turning left.
3.2 Linear Momentum in a Moving Frame
The expression for total velocity may be inserted into the summed linear mo
mentum equation to give
N
¸
i=1
F
i
=
N
¸
i=1
d
dt
(m
i
(v
o
+ ω r
i
)) (23)
= m
∂v
o
∂t
+
d
dt
¸
ω
N
¸
i=1
m
i
r
i
¸
,
where m =
¸
N
i=1
m
i
, and v
i
= v
o
+ ωr
i
. Further deﬁning the center of gravity
vector r
G
such that
mr
G
=
N
¸
i=1
m
i
r
i
, (24)
we have
N
¸
i=1
F
i
= m
∂v
o
∂t
+ m
d
dt
( ω r
G
). (25)
Using the expansion for total derivative again, the complete vector equation
in body coordinates is
22 3 VESSEL INERTIAL DYNAMICS
F =
¸
i=1
N = m
∂v
o
∂t
+ ω v
o
+
dω
dt
r
G
+ ω ( ω r
G
)
. (26)
Now we list some conventions that will be used from here on:
v
o
= ¦u, v, w¦ (bodyreferenced velocity)
r
G
= ¦x
G
, y
G
, z
g
¦ (bodyreferenced location of center of mass)
ω = ¦p, q, r¦ (rotation vector, in body coordinates)
F = ¦X, Y, Z¦ (external force, body coordinates).
The last term in the previous equation simpliﬁes using the vector triple product
identity
ω ( ω r
G
) = ( ω r
G
) ω −( ω ω)r
G
,
and the resulting three linear momentum equations are
X = m
¸
∂u
∂t
+ qw −rv +
dq
dt
z
G
−
dr
dt
y
G
+ (qy
G
+ rz
G
)p −(q
2
+ r
2
)x
G
¸
(27)
Y = m
¸
∂v
∂t
+ ru −pw +
dr
dt
x
G
−
dp
dt
z
G
+ (rz
G
+ px
G
)q −(r
2
+ p
2
)y
G
¸
Z = m
¸
∂w
∂t
+ pv −qu +
dp
dt
y
G
−
dq
dt
x
G
+ (px
G
+ qy
G
)r −(p
2
+ q
2
)z
G
¸
.
Note that about half of the terms here are due to the mass center being in a
diﬀerent location than the reference frame origin, i.e., r
G
=
0.
3.3 Example: Mass on a String
Consider a mass on a string, being swung around around in a circle at speed
U, with radius r. The centrifugal force can be computed in at least three
diﬀerent ways. The vector equation at the start is
F = m
∂v
o
∂t
+ ω v
o
+
dω
dt
r
G
+ ω ( ω r
G
)
.
3.3 Example: Mass on a String 23
3.3.1 Moving Frame Aﬃxed to Mass
Aﬃxing a reference frame on the mass, with the local x oriented forward and
y inward towards the circle center, gives
v
o
= ¦U, 0, 0¦
T
ω = ¦0, 0, U/r¦
T
r
G
= ¦0, 0, 0¦
T
∂v
o
∂t
= ¦0, 0, 0¦
T
∂ ω
∂t
= ¦0, 0, 0¦
T
,
such that
F = mω v
o
= m¦0, U
2
/r, 0¦
T
.
The force of the string pulls in on the mass to create the circular motion.
3.3.2 Rotating Frame Attached to Pivot Point
Aﬃxing the moving reference frame to the pivot point of the string, with the
same orientation as above but allowing it to rotate with the string, we have
v
o
= ¦0, 0, 0¦
T
ω = ¦0, 0, U/r¦
T
r
G
= ¦0, r, 0¦
T
∂v
o
∂t
= ¦0, 0, 0¦
T
∂ ω
∂t
= ¦0, 0, 0¦
T
,
giving the same result:
F = mω ( ω r
G
) = m¦0, U
2
/r, 0¦
T
.
24 3 VESSEL INERTIAL DYNAMICS
3.3.3 Stationary Frame
A frame ﬁxed in inertial space, and momentarily coincident with the frame
on the mass (3.3.1), can also be used for the calculation. In this case, as the
string travels through a small arc δψ, vector subtraction gives
δv = ¦0, U sin δψ, 0¦
T
· ¦0, Uδψ, 0¦
T
.
Since
˙
ψ = U/r, it follows easily that in the ﬁxed frame dv/dt = ¦0, U
2
/r, 0¦
T
,
as before.
3.4 Angular Momentum
For angular momentum, the summed particle equation is
N
¸
i=1
(
M
i
+r
i
F
i
) =
N
¸
i=1
r
i
d
dt
(m
i
v
i
), (28)
where
M
i
is an external moment on the particle i. Similar to the case for linear
momentum, summed internal moments cancel. We have
N
¸
i=1
(
M
i
+r
i
F
i
) =
N
¸
i=1
m
i
r
i
¸
∂v
o
∂t
+ ω v
o
¸
+
N
¸
i=1
m
i
r
i
∂ ω
∂t
r
i
+
N
¸
i=1
m
i
r
i
( ω ( ω r
i
)).
The summation in the ﬁrst term of the righthand side is recognized simply as
mr
G
, and the ﬁrst term becomes
mr
G
¸
∂v
o
∂t
+ ω v
o
¸
. (29)
The second term expands as (using the triple product)
N
¸
i=1
m
i
r
i
∂ ω
∂t
r
i
=
N
¸
i=1
m
i
(r
i
r
i
)
∂ ω
∂t
−
∂ ω
∂t
r
i
r
i
(30)
=
¸
N
i=1
m
i
((y
2
i
+ z
2
i
) ˙ p −(y
i
˙ q + z
i
˙ r)x
i
)
¸
N
i=1
m
i
((x
2
i
+ z
2
i
) ˙ q −(x
i
˙ p + z
i
˙ r)y
i
)
¸
N
i=1
m
i
((x
2
i
+ y
2
i
) ˙ r −(x
i
˙ p + y
i
˙ q)z
i
)
.
3.4 Angular Momentum 25
Employing the deﬁnitions of moments of inertia,
I =
I
xx
I
xy
I
xz
I
yx
I
yy
I
yz
I
zx
I
zy
I
zz
¸
¸
¸ (inertia matrix)
I
xx
=
N
¸
i=1
m
i
(y
2
i
+ z
2
i
)
I
yy
=
N
¸
i=1
m
i
(x
2
i
+ z
2
i
)
I
zz
=
N
¸
i=1
m
i
(x
2
i
+ y
2
i
)
I
xy
= I
yx
= −
N
¸
i=1
m
i
x
i
y
i
(crossinertia)
I
xz
= I
zx
= −
N
¸
i=1
m
i
x
i
z
i
I
yz
= I
zy
= −
N
¸
i=1
m
i
y
i
z
i
,
the second term of the angular momentum righthand side collapses neatly
into I∂ ω/∂t. The third term can be worked out along the same lines, but
oﬀers no similar condensation:
N
¸
i=1
m
i
r
i
(( ω r
i
) ω −( ω ω)r
i
) =
N
¸
i=1
m
i
r
i
ω( ω r
i
) (31)
=
¸
N
i=1
m
i
(y
i
r −z
i
q)(x
i
p + y
i
q + z
i
r)
¸
N
i=1
m
i
(z
i
p −x
i
r)(x
i
p + y
i
q + z
i
r)
¸
N
i=1
m
i
(x
i
q −y
i
p)(x
i
p + y
i
q + z
i
r)
=
I
yz
(q
2
−r
2
) + I
xz
pq −I
xy
pr
I
xz
(r
2
−p
2
) + I
xy
rq −I
yz
pq
I
xy
(p
2
−q
2
) + I
yz
pr −I
xz
qr
+
(I
zz
−I
yy
)rq
(I
xx
−I
zz
)rp
(I
yy
−I
xx
)qp
.
26 3 VESSEL INERTIAL DYNAMICS
Letting
M = ¦K, M, N¦ be the total moment acting on the body, i.e., the left
side of Equation 28, the complete moment equations are
K = I
xx
˙ p + I
xy
˙ q + I
xz
˙ r + (32)
(I
zz
−I
yy
)rq + I
yz
(q
2
−r
2
) + I
xz
pq −I
xy
pr +
m[y
G
( ˙ w + pv −qu) −z
G
( ˙ v + ru −pw)]
M = I
yx
˙ p + I
yy
˙ q + I
yz
˙ r +
(I
xx
−I
zz
)pr + I
xz
(r
2
−p
2
) + I
xy
qr −I
yz
qp +
m[z
G
( ˙ u + qw −rv) −x
G
( ˙ w + pv −qu)]
N = I
zx
˙ p + I
zy
˙ q + I
zz
˙ r +
(I
yy
−I
xx
)pq + I
xy
(p
2
−q
2
) + I
yz
pr −I
xz
qr +
m[x
G
( ˙ v + ru −pw) −y
G
( ˙ u + qw −rv)] .
3.5 Example: Spinning Book
Consider a homogeneous rectangular block with I
xx
< I
yy
< I
zz
and all oﬀ
diagonal moments of inertia are zero. The linearized angular momentum equa
tions, with no external forces or moments, are
I
xx
dp
dt
+ (I
zz
−I
yy
)rq = 0
I
yy
dq
dt
+ (I
xx
−I
zz
)pr = 0
I
zz
dr
dt
+ (I
yy
−I
xx
)qp = 0.
We consider in turn the stability of rotations about each of the main axes,
with constant angular rate Ω. The interesting result is that rotations about
the x and z axes are unstable, while rotation about the y axis is not.
3.5.1 xaxis
In the case of the xaxis, p = Ω + δp, q = δq, and r = δr, where the δ preﬁx
indicates a small value compared to Ω. The ﬁrst equation above is uncoupled
3.5 Example: Spinning Book 27
from the others, and indicates no change in δp, since the small term δqδr can
be ignored. Diﬀerentiate the second equation to obtain
I
yy
∂
2
δq
∂t
2
+ (I
xx
−I
zz
)Ω
∂δr
∂t
= 0
Substitution of this result into the third equation yields
I
yy
I
zz
∂
2
δq
∂t
2
+ (I
xx
−I
zz
)(I
xx
−I
yy
)Ω
2
δq = 0.
A simpler expression is δ¨ q + αδq = 0, which has response δq(t) = δq(0)e
√
−αt
,
when δ ˙ q(0) = 0. For spin about the xaxis, both coeﬃcients of the diﬀerential
equation are positive, and hence α > 0. The imaginary exponent indicates
that the solution is of the form δq(t) = δq(0)cos
√
αt, that is, it oscillates but
does not grow. Since the perturbation δr is coupled, it too oscillates.
3.5.2 yaxis
Now suppose q = Ω + δq: diﬀerentiate the ﬁrst equation and substitute into
the third equation to obtain
I
zz
I
xx
∂
2
δp
∂t
2
+ (I
yy
−I
xx
)(I
yy
−I
zz
)Ω
2
δp = 0.
Here the second coeﬃcient has negative sign, and therefore α < 0. The ex
ponent is real now, and the solution grows without bound, following δp(t) =
δp(0)e
√
−αt
.
3.5.3 zaxis
Finally, let r = Ω+δr: diﬀerentiate the ﬁrst equation and substitute into the
second equation to obtain
I
yy
I
xx
∂
2
δp
∂t
2
+ (I
xx
−I
zz
)(I
yy
−I
zz
)Ω
2
δp = 0.
The coeﬃcients are positive, so bounded oscillations occur.
28 3 VESSEL INERTIAL DYNAMICS
3.6 Parallel Axis Theorem
Often, the mass center of an body is at a diﬀerent location than a more con
venient measurement point, the geometric center of a vessel for example. The
parallel axis theorem allows one to translate the mass moments of inertia refer
enced to the mass center into another frame with parallel orientation, and vice
versa. Sometimes a translation of coordinates to the mass center will make the
crossinertial terms I
xy
, I
yz
, I
xz
small enough that they can be ignored; in this
case r
G
=
0 also, so that the equations of motion are signiﬁcantly reduced, as
in the spinning book example.
The formulas are:
I
xx
=
¯
I
xx
+ m(δy
2
+ δz
2
) (33)
I
yy
=
¯
I
yy
+ m(δx
2
+ δz
2
)
I
zz
=
¯
I
zz
+ m(δx
2
+ δy
2
)
I
yz
=
¯
I
yz
−mδyδz
I
xz
=
¯
I
xz
−mδxδz
I
xy
=
¯
I
xy
−mδxδy,
where
¯
I represents an MMOI in the axes of the mass center, and δx, for ex
ample, is the translation of the xaxis to the new frame. Note that translation
of MMOI using the parallel axis theorem must be either to or from a frame
resting exactly at the center of gravity.
3.7 Basis for Simulation
Except for external forces and moments
F and
M, we now have the neces
sary terms for writing a full nonlinear simulation of a rigid body, in body
coordinates. There are twelve states, comprising the following components:
• v
o
, the vector of bodyreferenced velocities.
• ω, body rotation rate vector.
• x, location of the body origin, in inertial space.
•
E, Euler angle vector.
3.7 Basis for Simulation 29
The derivatives of bodyreferenced velocity and rotation rate come from Equa
tions 27 and 32, with some coupling which generally requires a 6 6 matrix
inverse. The Cartesian position propagates according to
˙
x = R
T
(
E)v
o
, (34)
while the Euler angles follow:
˙
E = Γ(
E) ω. (35)
30 4 HYDRODYNAMICS: INTRODUCTION
4 HYDRODYNAMICS: INTRODUCTION
The forces and moments on a vessel are complicated functions of many fac
tors, including water density, viscosity, surface tension, pressure, vapor pres
sure, and motions of the body. The most important factors for large ocean
vehicles are density and motion, and we can make simpliﬁcations to param
eterize the most prominent relationships. This section pertains to the use of
hydrodynamic coeﬃcients for predicting hydrodynamic response.
4.1 Taylor Series and Hydrodynamic Coeﬃcients
Recall the Taylor expansion of a function:
f(x) = f(x
o
) +
∂f(x
o
)
∂x
(x −x
o
) +
1
2!
∂
2
f(x
o
)
∂x
2
(x −x
o
)
2
+ . (36)
We introduce the notation
f
x
=
∂f(x
o
)
∂x
f
x
x =
1
2!
∂
2
f(x
o
)
∂x
2
,
and so on, so that a twovariable Taylor expansion would have the form
f(x, y) = f(x
o
, y
o
) + (37)
f
x
(x −x
o
) + f
y
(y −y
o
) +
f
xx
(x −x
o
)
2
+ f
yy
(y −y
o
)
2
+ f
xy
(x −x
o
)(y −y
o
) +
f
xxx
(x −x
o
)
3
+ . (38)
Note that all of the factorials are included in the coeﬃcients. This notation
covers some instances where the formal Taylor series is meaningless, but the
notation is still clear. As one example, ﬂuid drag is often written as
F =
1
2
ρC
d
Au[u[ = F
uu
u[u[.
4.2 Surface Vessel Linear Model 31
4.2 Surface Vessel Linear Model
We now discuss some of the hydrodynamic parameters which govern a ship
maneuvering in the horizontal plane. The body xaxis is forward and the y
axis is to port, so positive r has the boat turning left. We will consider motions
only in the horizontal plane, which means θ = ψ = p = q = w = 0. Since the
vessel is symmetric about the x −z plane, y
G
= 0; z
G
is inconsequential. We
then have at the outset
X = m
∂u
∂t
−rv −x
G
r
2
(39)
Y = m
∂v
∂t
+ ru + x
G
∂r
∂t
N = I
zz
∂r
∂t
+ mx
G
∂v
∂t
+ ru
.
Letting u = U + u, where U >> u, and eliminating higherorder terms, this
set is
X = m
∂u
∂t
(40)
Y = m
∂v
∂t
+ rU + x
G
∂r
∂t
N = I
zz
∂r
∂t
+ mx
G
∂v
∂t
+ rU
.
A number of coeﬃcients can be discounted. First, in a homogeneous sea, with
no current, wave, or wind eﬀects, ¦X
x
, X
y
, X
φ
, Y
x
, Y
y
, Y
φ
, N
x
, N
y
, N
φ
¦ are all
zero. We assume that no hydrodynamic forces depend on the position of the
vessel.
2
Second, consider X
v
: since this longitudinal force would have the same
sign regardless of the sign of v (because of sidetoside hull symmetry), it must
have zero slope with v at the origin. Thus X
v
= 0. The same argument shows
that ¦X
r
, X
˙ v
, X
˙ r
, Y
u
, Y
˙ u
, N
u
, N
˙ u
¦ = 0. Finally, since ﬂuid particle acceleration
relates linearly with pressure or force, we do not consider nonlinear acceleration
2
Note that the linearized heave/pitch dynamics of a submarine do depend on the pitch
angle; this topic will be discussed later.
32 4 HYDRODYNAMICS: INTRODUCTION
terms, or higher time derivatives. It should be noted that some nonlinear terms
related to those we have eliminated above are not zero. For instance, Y
uu
= 0
because of hull symmetry, but in general X
vv
= 0 only if the vessel is bowstern
symmetric.
We have so far, considering only the linear hydrodynamic terms,
(m−X
˙ u
) ˙ u = X
u
u + X
(41)
(m−Y
˙ v
) ˙ v + (mx
G
−Y
˙ r
) ˙ r = Y
v
v + (Y
r
−mU)r + Y
(42)
(mx
G
−N
˙ v
) ˙ v + (I
zz
−N
˙ r
) ˙ r = N
v
v −(N
r
−mx
G
U)r + N
. (43)
The right side here carries also the imposed forces from a thruster(s) and
rudder(s) ¦X
, Y
, N
¦. Note that the surge equation is decoupled from the
sway and yaw, but that sway and yaw themselves are coupled, and there
fore are of immediate interest. With the state vector s = ¦v, r¦ and exter
nal force/moment vector
F = ¦Y
, N
¦, a statespace representation of the
sway/yaw system is
¸
m−Y
˙ v
mx
G
−Y
˙ r
mx
G
−N
˙ v
I
zz
−N
˙ r
¸
ds
dt
=
¸
Y
v
Y
r
−mU
N
v
N
r
−mx
G
U
¸
s +
F, or (44)
M
˙
s = Ps +
F
˙
s = M
−1
Ps + M
−1
F
˙
s = As + B
F. (45)
The matrix M is a mass or inertia matrix, which is always invertible. The
last form of the equation is a standard one wherein A represents the internal
dynamics of the system, and B is a gain matrix for the control and disturbance
inputs.
4.3 Stability of the Sway/Yaw System
Consider the homogeneous system
˙
s = As:
˙ s
1
= A
11
s
1
+ A
12
s
2
˙ s
2
= A
21
s
1
+ A
22
s
2
.
4.3 Stability of the Sway/Yaw System 33
We can rewrite the second equation as
s
2
=
d()
dt
−A
22
−1
A
21
s
1
(46)
and substitute into the ﬁrst equation to give
¨ s
1
+ (−A
11
−A
22
) ˙ s
1
+ (A
11
A
22
−A
12
A
21
)s
1
= 0. (47)
Note that these operations are allowed because the derivative operator is linear;
in the language of the Laplace transform, we would simply use s. A necessary
and suﬃcient condition for stability of this ODE system is that each coeﬃcient
must be greater than zero:
−A
11
−A
22
> 0 (48)
A
11
A
22
−−A
12
A
21
> 0
The components of A for the sway/yaw problem are
A
11
=
(I
zz
−N
˙ r
)Y
v
+ (Y
˙ r
−mx
G
)N
v
(m−Y
˙ v
)(I
zz
−N
˙ r
) −(mx
G
−Y
˙ r
)(mx
G
−N
˙ v
)
(49)
A
12
=
−(I
zz
−N
˙ r
)(mU −Y
r
) −(Y
˙ r
−mx
G
)(mx
G
U −N
r
)
(m−Y
˙ v
)(I
zz
−N
˙ r
) −(mx
G
−Y
˙ r
)(mx
G
−N
˙ v
)
A
21
=
(N
˙ v
−mx
G
)Y
v
+ (m−Y
˙ v
)N
v
(m−Y
˙ v
)(I
zz
−N
˙ r
) −(mx
G
−Y
˙ r
)(mx
G
−N
˙ v
)
A
22
=
−(N
˙ v
−mx
G
)(mU −Y
r
) −(m−Y
˙ v
)(mx
G
U −N
r
)
(m−Y
˙ v
)(I
zz
−N
˙ r
) −(mx
G
−Y
˙ r
)(mx
G
−N
˙ v
)
.
The denominators are identical, and can be simpliﬁed. First, let x
G
· 0;
valid for many vessels with the origin is at the geometric center. If the origin
is at the center of mass, x
G
= 0. Next, if the vessel is reasonably balanced
with regard to forward and aft areas with respect to the origin, the terms
¦N
˙ v
, Y
˙ r
, N
v
, Y
r
¦ take very small values in comparison with the others. To wit,
the added mass term −Y
˙ v
is of the order of the vessel’s material mass m,
and similarly N
˙ r
· −I
zz
. Both Y
˙ v
and N
˙ r
take large negative values. Linear
drag and rotational drag are signiﬁcant also; these are the terms Y
v
and N
r
,
34 4 HYDRODYNAMICS: INTRODUCTION
both large and negative. The denominator for A’s components reduces to
(m−Y
˙ v
)(I
zz
−N
˙ r
), and
A
11
=
Yv
m−Y
˙ v
< 0
A
22
=
Nr
Izz−N
˙ r
< 0.
Hence the ﬁrst condition for stability is met: −A
11
−A
22
> 0. For the second
condition, since the denominators of the A
ij
are identical, we have only to look
at the numerators. For stability, we require
(I
zz
−N
˙ r
)Y
v
(m−Y
˙ v
)N
r
(50)
−[N
˙ v
Y
v
+ (m−Y
˙ v
)N
v
] [−(I
zz
−N
˙ r
)(mU −Y
r
) + Y
˙ r
N
r
] > 0.
The ﬁrst term is the product of two large negative and two large positive
numbers. The second part of the second term contains mU, which has a large
positive value, generally making stability critical on the (usually negative) N
v
.
When only the largest terms are considered for a vessel, a simpler form is
common:
C = Y
v
N
r
+ N
v
(mU −Y
r
) > 0. (51)
C is called the vessels stability parameter. The terms of C compete, and
yaw/sway stability depends closely on the magnitude and sign of N
v
. Adding
more surface area aft drives N
v
more positive, increasing stability as ex
pected. Stability can also be improved by moving the center of gravity forward.
Nonzero x
G
shows up as follows:
C = Y
v
(N
r
−mx
G
U) + N
v
(mU −Y
r
) > 0. (52)
Since N
r
and Y
v
are both negative, positive x
G
increases the (positive) inﬂuence
of C’s ﬁrst term.
4.4 Basic Rudder Action in the Sway/Yaw Model
Rudders are devices which develop large lift forces due to an angle of attack
with respect to the oncoming ﬂuid. As in our discussion of lift on the body,
the form is as follows: L =
1
2
ρAU
2
C
l
(α), where α is the angle of attack. The
4.4 Basic Rudder Action in the Sway/Yaw Model 35
δ
y
x
r
Figure 3: Convention for positive rudder angle in the vessel reference system.
lift coeﬃcient C
l
is normally linear with α near α = 0, but the rudder stalls
when the angle of attack reaches a critical value, and thereafter develops much
less lift. We will assume that α is small enough that the linear relationship
applies:
C
l
(α) =
∂C
l
∂α
α=0
α. (53)
Since the rudder develops force (and a small moment) far away from the body
origin, say a distance l aft, the moment equation is quite simple. We have
Y
α
=
1
2
ρA
∂C
l
∂α
α=0
U
2
(54)
N
α
= −
1
2
ρA
∂C
l
∂α
α=0
lU
2
. (55)
Note the diﬀerence between the rudder angle expressed in the body frame, δ,
and the total angle of attack α. Angle of attack is inﬂuenced by δ, as well
as v/U and lr. Thus, in tank testing with v = 0, δ = α and N
δ
= N
α
,
etc., but in real conditions, other hydrodynamic derivatives are augmented to
capture the necessary eﬀects, for example N
v
and N
r
. Generally speaking, the
hydrodynamic characteristics of the vessel depend strongly on the rudder, even
when δ = 0. In this case the rudder still opposes yaw and sway perturbations
and acts to stabilize the vessel.
A positive rudder deﬂection (deﬁned to have the same sense as the yaw angle)
causes a negative yaw perturbation, and a very small positive sway perturba
tion.
36 4 HYDRODYNAMICS: INTRODUCTION
4.4.1 Adding Yaw Damping through Feedback
The stability coeﬃcient C resulting from the addition of a control law δ = k
r
r,
where k
r
> 0 is a feedback gain, is
C = Y
v
(N
r
−mx
G
U + k
r
N
δ
) + N
v
(mU −Y
r
−k
r
Y
δ
). (56)
Y
δ
is small positive, but N
δ
is large and negative. Hence C becomes more
positive, since Y
v
is negative.
Control system limitations and the stalling of rudders make obvious the fact
that even a very large control gain k
r
cannot completely solve stability prob
lems of a poorlydesigned vessel with an inadequate rudder. On the other
hand, a vessel which is overly stable (C >> 0 with no rudder action) is unma
neuverable. A properlybalanced vessel just achieves stability with zero rudder
action, so that a reasonable amount of control will provide good maneuvering
capabilities.
4.4.2 Heading Control in the Sway/Yaw Model
Considering just the yaw equation of motion, i.e., v = 0, with a rudder, we
have
(I
zz
−N
˙ r
)
¨
φ + (mx
G
U −N
r
)
˙
φ = N
δ
δ. (57)
Employing the control law δ = k
φ
φ, the system equation becomes a homoge
neous, secondorder ODE:
(I
zz
−N
˙ r
)
¨
φ + (mx
G
U −N
r
)
˙
φ −N
δ
k
φ
φ = 0. (58)
Since all coeﬃcients are positive (recall N
δ
< 0), the equation gives a stable θ
response, settling under secondorder dynamics to θ(∞) = 0. The control law
δ = k
φ
(φ − φ
desired
) + k
r
r is the basis for heading autopilots, which are used
to track φ
desired
. This use of an error signal to drive an actuator is in fact the
essence of feedback control. In this case, we require sensors to obtain r and φ,
a controller to calculate δ, and an actuator to implement the corrective action.
4.5 Response of the Vessel to Step Rudder Input 37
4.5 Response of the Vessel to Step Rudder Input
4.5.1 Phase 1: Accelerations Dominate
When the rudder ﬁrst moves, acceleration terms dominate, since the velocities
are zero. The equation looks like this:
¸
m−Y
˙ v
mx
G
−Y
˙ r
mx
G
−N
˙ v
I
zz
−N
˙ r
¸
˙ v
˙ r
¸
=
Y
δ
N
δ
¸
δ. (59)
Since Y
˙ r
and mx
G
are comparatively small in the ﬁrst row, we have
˙ v(0) =
Y
δ
δ
m−Y
˙ v
, (60)
and the vessel moves to the left, the positive vdirection. The initial yaw is in
the negative rdirection, since N
δ
< 0:
˙ r(0) =
N
δ
δ
I
zz
−N
˙ r
. (61)
The ﬁrst phase is followed by a period (Phase 2), in which many terms are
competing and contributing to the transient response.
4.5.2 Phase 3: Steady State
When the transients have decayed, the vessel is in a steady turning condition,
and the accelerations are zero. The system equations reduce to
v
r
¸
=
δ
C
(mx
G
U −N
r
)Y
δ
+ (Y
r
−mU)N
δ
N
v
Y
δ
−Y
v
N
δ
¸
. (62)
Note that the denominator is the stability parameter. The steady turning rate
is thus approximated by
r = −
Y
v
N
δ
C
δ. (63)
With C > 0, the steadystate yaw rate is negative. If the vessel is unstable
(C < 0), it turns in the opposite direction than expected. This turning rate
equation can also be used to estimate turning radius R:
R =
U
r
=
UC
−Y
v
N
δ
δ
. (64)
38 4 HYDRODYNAMICS: INTRODUCTION
The radius goes up directly with C, indicating that too stable a ship has poor
turning performance. We see also that increasing the rudder area increases
N
δ
, decreasing R as desired. Increasing the deﬂection δ to reduce R works
only to the point of stalling.
4.6 Summary of the Linear Maneuvering Model
We conclude our discussion of the yaw/sway model by noting that
1. The linearized sway/yaw dynamics of a surface vessel are strongly cou
pled, and they are independent of the longitudinal dynamics.
2. The design parameter C should be slightly greater than zero for easy
turning, and “ handsoﬀ” stability. The case C < 0 should only be
considered under active feedback control.
3. The analysis is valid only up to small angles of attack and turning rates.
Very tight maneuvering requires the nonlinear inertial components and
hydrodynamic terms. Among other eﬀects, the nonlinear equations cou
ple surge to the other motions, and the actual vessel loses forward speed
during maneuvering.
4.7 Stability in the Vertical Plane
Stability in the horizontal plane changes very little as a function of speed,
because drag and lift eﬀects generally scale with U
2
. This fact is not true
in the vertical plane, for which the dimensional weight/buoyancy forces and
moments are invariant with speed. For example, consider the case of heave
and pitch, with x
G
= 0 and no actuation:
m
∂w
∂t
−Uq
= Z
˙ w
˙ w + Z
w
w + Z
˙ q
˙ q + Z
q
q + (B −W) (65)
I
yy
dq
dt
= M
˙ w
˙ w + M
w
w + M
˙ q
˙ q + M
q
q −Bl
b
sin θ. (66)
The last term in each equation is a hydrostatic eﬀect induced by opposing net
buoyancy B and weight W. l
b
denotes the vertical separation of the center
of gravity and the center of buoyancy, creating the socalled righting moment
which nearly all underwater vehicles possess. Because buoyancy eﬀects do not
4.7 Stability in the Vertical Plane 39
change with speed, the dynamic properties and hence stability of the vehicle
may change with speed.
40 5 SIMILITUDE
5 SIMILITUDE
5.1 Use of Nondimensional Groups
For a consistent description of physical processes, we require that all terms
in an equation must have the same units. On the basis of physical laws,
some quantities are dependent on other, independent quantities. We form
nondimensional groups out of the dimensional ones in this section, and apply
the technique to maneuvering.
The Buckingham πtheorem provides a basis for all nondimensionalization.
Let a quantity Q
n
be given as a function of a set of n −1 other quantities:
Q
n
= f
Q
(Q
1
, Q
2
, , Q
n−1
). (67)
There are n variables here, but suppose also that there are only k independent
ones; k is equivalent to the number of physical unit types encountered. The
theorem asserts that there are n − k dimensionless groups π
i
that can be
formed, and the functional equivalence is reduced to
π
n−k
= f
π
(π
1
, π
2
, , π
n−k−1
). (68)
Example. Suppose we have a block of mass m resting on a frictionless hor
izontal surface. At time zero, a steady force of magnitude F is applied. We
want to know X(T), the distance that the block has moved as of time T. The
dimensional function is X(T) = f
Q
(m, F, T), so n = 4. The (MKS) units are
[X()] = m
[m] = kg
[F] = kgm/s
2
[T] = s,
and therefore k = 3. There is just one nondimensional group in this re
lationship; π
1
assumes only a constant (but unknown) value. Simple term
cancellation gives π
1
= X(T)m/FT
2
, not far at all from the known result that
X(T) = FT
2
/2m!
Example. Consider the ﬂow rate Q of water from an open bucket of height
h, through a drain nozzle of diameter d. We have
5.1 Use of Nondimensional Groups 41
Q = f
Q
(h, d, ρ, µ, g),
where the water density is ρ, and its absolute viscosity µ; g is the acceleration
due to gravity. No other parameters aﬀect the ﬂow rate. We have n = 6, and
the (MKS) units of these quantities are:
[Q] = m
3
/s
[h] = m
[d] = m
[ρ] = kg/m
3
[µ] = kg/ms
[g] = m/s
2
There are only three units that appear: [length, time, mass], and thus k = 3.
Hence, only three nondimensional groups exist, and one is a unique func
tion of the other two. To arrive at a set of valid groups, we must create
three nondimensional quantities, making sure that each of the original (di
mensional) quantities is represented. Intuition and additional manipulations
come in handy, as we now show.
Three plausible ﬁrst groups are: π
1
= ρQ/dµ, π
2
= dρ
√
gh/µ, and π
3
= h/d.
Note that all six quantities appear at least once. Since h and d have the same
units, they could easily change places in the ﬁrst two groups. However, π
1
is
recognized as a Reynolds number pertaining to the oriﬁce ﬂow. π
2
is more
awkward, but products and fractions of groups are themselves valid groups,
and we may construct π
4
= π
1
/π
2
= Q/d
2
√
gh to nondimensionalize Q with
a pressure velocity, and then π
5
= π
1
/π
4
= ρd
√
gh/µ to establish an oriﬁce
Reynolds number independent of Q. We ﬁnally have the useful result
π
4
= f
π
(π
5
, π
2
) −→
Q
d
2
√
gh
= f
π
ρd
√
gh
µ
,
h
d
.
The uncertainty about where to use h and d, and the questionable importance
of h/d as a group are remnants of the theorem. Intuition is that h/d is im
42 5 SIMILITUDE
material, and the other two terms have a nice physical meaning, e.g., π
5
is a
Reynolds number.
The power of the πtheorem is primarily in reducing the number of parameters
which must be considered independently to characterize a process. In the ﬂow
example, the theorem reduced the number of independent parameters from
ﬁve to two, with no constraints about the actual physics taking place.
5.2 Common Groups in Marine Engineering
One frequently encounters the following groups in ﬂuid mechanics and marine
engineering:
1. Froude number:
Fr =
U
√
gL
, (69)
where U is the speed of the vessel, g is the acceleration due to gravity,
and L is the waterline length of the vessel. The Froude number appears
in problems involving pressure boundary conditions, such as in waves on
the ocean surface. Roughly speaking, it relates the vessel speed U to
water wave speeds of wavelength L; the phase speed of a surface wave is
V =
λg/2π, where λ is the wavelength.
2. Cavitation number:
δ =
P
∞
−P
v
1
2
ρU
2
, (70)
where P
∞
represents the ambient total pressure, P
v
the vapor pressure
of the ﬂuid, and U the propeller inlet velocity. A low cavitation number
means that the Bernoulli pressure loss across the lifting surface will cause
the ﬂuid to vaporize, causing bubbles, degradation of performance, and
possible deterioration of the material.
3. Reynolds number:
Re =
Ul
µ/ρ
, (71)
5.2 Common Groups in Marine Engineering 43
where U is velocity, µ is absolute viscosity, and ρ is density. Since Re
appears in many applications, l represents one of many length scales.
Reynolds number is a ratio of ﬂuid inertial pressures to viscous pres
sures: When Re is high, viscous eﬀects are negligible. Re can be used to
characterize pipe ﬂow, bluﬀ body wakes, and ﬂow across a plate, among
others.
4. Weber number:
W =
ρU
2
l
σ
, (72)
where σ is the surface tension of a ﬂuid. Given that [σ] = N/m (MKS),
ρU
2
normalizes pressure, and l normalizes length. The Weber number
indicates the importance of surface tension.
To appreciate the origins of these terms from a ﬂuid particle’s point of view,
consider a box having side lengths [dx, dy, dz]. Various forces on the box scale
as
(inertia) F
i
= ρ
∂v
∂t
dxdydz + ρv
∂v
∂x
dxdydz · ρU
2
l
2
(gravity) F
g
= ρgdxdydz · ρgl
3
(pressure) F
p
= Pdxdy · Pl
2
(shear) F
s
= µ
∂v
∂z
dxdy · µUl
(surface tension) F
σ
= σdx · σl.
Thus the groups listed above can be written as
Fr =
F
i
F
g
·
U
2
gl
δ =
F
p
F
i
·
P
ρU
2
Re =
F
i
F
s
·
ρUl
µ
W =
F
i
F
σ
·
ρU
2
l
σ
44 5 SIMILITUDE
When testing models, it is imperative to maintain as many of the nondimen
sional groups as possible of the fullscale system. This holds for the geometry
of the body and the kinematics of the ﬂow, the surface roughness, and the
all of the relevant groups governing ﬂuid dynamics. Consider the example of
nozzle ﬂow from a bucket. Suppose that we conduct a model test in which
Re is abnormally large, i.e., the viscous eﬀects are negligible. Under inviscid
conditions, the ﬂow rate is Q = πd
2
√
2gh/4. This rate cannot be achieved for
lowerRe conditions because of ﬂuid drag in the oriﬁce, however.
In a vessel, we write the functional relationship for drag as a starting point:
C
r
=
D
1
2
ρAU
2
= f
Q
(ρ, µ, g, σ, U, l)
= f
π
(Re, Fr, W).
First, since l is large, W is very large, and hence surface tension plays no role.
Next, we look at Re = Ul/ν and Fr = U/
√
gl, both of which are important
for surface vessels. Suppose that l
ship
= λl
model
, so that usually λ >> 1;
additionally, we set g
model
= g
ship
, i.e., the model and the true vessel operate
in the same gravity ﬁeld.
Froude number similitude requires U
model
= U
ship
/
√
λ. Then Reynolds number
scaling implies directly ν
model
= ν
ship
/λ
3/2
. Unfortunately, few ﬂuids with this
property are workable in a large testing tank. As a result, accurate scaling of
Re for large vessels to model scale is quite diﬃcult.
For surface vessels, and submarines near the surface, it is a routine procedure
to employ turbulence stimulators to achieve ﬂow that would normally occur
with shipscale Re. Above a critical value Re · 500, 000, C
f
is not sensitive
to Re. With this achieved, one then tries to match Fr closely.
5.3 Similitude in Maneuvering
The linear equations of motion for the horizontal yaw/sway problem are:
(m−Y
˙ v
) ˙ v −Y
v
v + (mU −Y
r
)r + (mx
G
−Y
˙ r
) ˙ r = Y
(I
zz
−N
˙ r
) ˙ r + (mx
G
U −Nr)r + (mx
G
−N
˙ v
) ˙ v −N
v
v = N.
These equations can be nondimensionalized in a standard way, by using the
quantities [U, L, ρ]: these three values provide the necessary units of length,
5.3 Similitude in Maneuvering 45
time, and mass, and furthermore are readily accessible to the user. First, we
create nondimensional states, denoted with a prime symbol:
˙ v
=
L
U
2
˙ v (73)
v
=
1
U
v
˙ r
=
L
2
U
2
˙ r
r
=
L
U
r.
We follow a similar procedure for the constant terms as follows, including a
factor of 1/2 with ρ, for consistency with our previous expressions:
m
=
m
1
2
ρL
3
(74)
I
zz
=
I
zz
1
2
ρL
5
x
G
=
x
G
L
U
=
U
U
= 1
Y
˙ v
=
Y
˙ v
1
2
ρL
3
Y
v
=
Y
v
1
2
ρUL
2
Y
˙ r
=
Y
˙ r
1
2
ρL
4
Y
r
=
Y
r
1
2
ρUL
3
Y
=
Y
1
2
ρU
2
L
2
N
˙ v
=
N
˙ v
1
2
ρL
4
N
v
=
N
v
1
2
ρUL
3
46 5 SIMILITUDE
N
˙ r
=
N
˙ r
1
2
ρL
5
N
r
=
N
r
1
2
ρUL
4
N
=
N
1
2
ρU
2
L
3
.
Note that every force has been normalized with
1
2
ρU
2
L
2
, and every moment
with
1
2
ρU
2
L
3
; time has been also nondimensionalized with L/U. Thus we
arrive at a completely equivalent set of nondimensional system equations,
(m
−Y
˙ v
) ˙ v
−Y
v
v
+ (m
U
−Y
r
)r
+ (m
x
G
−Y
˙ r
) ˙ r
= Y
(75)
(I
zz
−N
˙ r
) ˙ r
+ (m
x
G
U
−N
r
)r
+ (m
x
G
−N
˙ v
) ˙ v
−N
v
v
= N
.
Since ﬂuid forces and moments generally scale with U
2
, the nondimensionalized
description holds for a range of velocities.
5.4 Roll Equation Similitude
Certain nondimensional coeﬃcients may arise which depend explicitly on U,
and therefore require special attention. Let us carry out a similar normaliza
tion of the simpliﬁed roll equation
(I
xx
−K
˙ p
) ˙ p −K
p
p −K
ψ
ψ = K. (76)
For a surface vessel, the roll moment K
ψ
is based on metacentric stability, and
has the form K
ψ
= −ρg∇(GM), where ∇ is the displaced ﬂuid volume of the
vessel, and GM is the metacentric height. The nondimensional terms are
I
xx
=
I
xx
1
2
ρL
5
(77)
K
˙ p
=
K
˙ p
1
2
ρL
5
K
p
=
K
p
1
2
ρUL
4
K
ψ
=
K
ψ
1
2
ρU
2
L
3
5.4 Roll Equation Similitude 47
˙ p
=
L
2
U
2
˙ p
p
=
L
U
p,
leading to the equivalent system
(I
xx
−K
˙ p
) ˙ p
−K
p
p
−
2gL
U
2
∇
(GM
)φ = K
. (78)
Note that the roll angle φ was not nondimensionalized. The Froude number
has a very strong inﬂuence on roll stability, since it appears explicitly in the
nondimensional righting moment term, and also has a strong inﬂuence on K
p
.
In the case of a submarine, the righting moment has the form K
ψ
= −Bh,
where B is the buoyant force, and h is the righting arm. The nondimensional
coeﬃcient becomes
K
ψ
= −
Bh
1
2
ρU
2
L
3
.
K
ψ
again depends strongly on U, since B and h are ﬁxed; this K
ψ
needs to be
maintained in model tests.
48 6 CAPTIVE MEASUREMENTS
6 CAPTIVE MEASUREMENTS
Before making the decision to measure hydrodynamic derivatives, a prelimi
nary search of the literature may turn up useful estimates. For example, test
results for many hullforms have already been published, and the basic lift
ing surface models are not diﬃcult. The available computational approaches
should be considered as well; these are very good for predicting added mass
in particular. Finally, modern sensors and computer control systems make
possible the estimation of certain coeﬃcients based on openwater tests of a
model or a fullscale design.
In model tests, the Froude number Fr =
U
√
gL
, which scales the inﬂuence
of surface waves, must be maintained between model and fullscale surface
vessels. Reynolds number Re =
UL
ν
, which scales the eﬀect of viscosity, need
not be matched as long as the scale model attains turbulent ﬂow (supercritical
Re). One can use turbulence stimulators near the bow if necessary. Since the
control surface(s) and propeller(s) aﬀect the coeﬃcients, they should both be
implemented in model testing.
6.1 Towtank
In a towtank, tow the vehicle at diﬀerent angles of attack, measuring sway
force and yaw moment. The slope of the curve at zero angle determines Y
v
and N
v
respectively; higherorder terms can be generated if the points deviate
from a straight line. Rudder derivatives can be computed also by towing with
various rudder angles.
6.2 Rotating Arm Device
On a rotating arm device, the vessel is ﬁxed on an arm of length R, rotating at
constant rate r: the vessel forward speed is U = rR. The idea is to measure the
crossbody force and yaw moment as a function of r, giving the coeﬃcients Y
r
and N
r
. Note that the lateral force also contains the component (m−Y
˙ v
)r
2
R.
The coeﬃcients Y
v
and N
v
can also be obtained by running with a ﬁxed angle
of attack. Finally, the measurement is made over one rotation only, so that
the vessel does not reenter its own wake.
6.3 PlanarMotion Mechanism 49
6.3 PlanarMotion Mechanism
With a planar motion mechanism, the vessel is towed at constant forward
speed U, but is held by two posts, one forward and one aft, which can each
impose independent sway motions, therefore producing variable yaw. The
model moves in pure sway if y
a
(t) = y
b
(t), in a pure yaw motion about the
midlength point if y
a
(t) = −y
b
(t), or in a combination sway and yaw motion.
The connection points are a distance l forward and aft from the vessel origin.
Usually a sinusoidal motion is imposed:
y
a
(t) = a cos ωt (79)
y
b
(t) = b cos(ωt + ψ),
and the transverse forces on the posts are measured and approximated as
Y
a
(t) = F
a
cos(ωt + θ
a
) (80)
Y
b
(t) = F
b
cos(ωt + θ
b
).
If linearity holds, then
(m−Y
˙ v
) ˙ v −Y
v
v + (mU −Y
r
)r + (mx
G
−Y
˙ r
) ˙ r = Y
a
+ Y
b
(81)
(I
zz
−N
˙ r
) ˙ r + (mx
G
U −N
r
)r + (mx
G
−N
˙ v
) ˙ v −N
v
v = (Y
b
−Y
a
)l.
We have v = ( ˙ y
a
+ ˙ y
a
)/2 and r = ( ˙ y
b
− ˙ y
a
)/2l. When a = b, these become
v = −
aω
2
(sin ωt(1 + cos ψ) + cos ωt sin ψ) (82)
˙ v = −
aω
2
2
(cos ωt(1 + cos ψ) −sinωt sin ψ)
r = −
aω
2l
(sin ωt(cos ψ −1) + cos ωt sin ψ)
˙ r = −
aω
2
2l
(cos ωt(cos ψ −1) −sin ωt sin ψ) .
Equating the sine terms and then the cosine terms, we obtain four independent
equations:
50 6 CAPTIVE MEASUREMENTS
(m−Y
˙ v
)
−
aω
2
2
(1 + cos ψ) − (83)
Y
v
−
aω
2
sin ψ +
(mU −Y
r
)
−
aω
2l
sin ψ +
(mx
G
−Y
˙ r
)
−
aω
2
2l
(cos ψ −1) = F
a
cos θ
a
+ F
b
cos θ
b
(m−Y
˙ v
)
−
aω
2
2
(−sin ψ) − (84)
Y
v
−
aω
2
(1 + cos ψ) +
(mU −Y
r
)
−
aω
2l
(cos ψ −1) +
(mx
G
−Y
˙ r
)
−
aω
2
2l
(−sin ψ) = −F
a
sin θ
a
−F
b
sin θ
b
(I
zz
−N
˙ r
)
−
aω
2
2l
(cos ψ −1) + (85)
(mx
G
U −N
r
)
−
aω
2l
sin ψ +
(mx
G
−N
˙ v
)
−
aω
2
2
(1 + cos ψ) −
N
v
−
aω
2
sin ψ = l(F
b
cos θ
b
−F
a
cos θ
a
)
(I
zz
−N
˙ r
)
−
aω
2
2l
(−sin ψ) + (86)
(mx
G
U −N
r
)
−
aω
2l
(cos ψ −1) +
(mx
G
−N
˙ v
)
−
aω
2
2
(−sin ψ) −
N
v
−
aω
2
(1 + cos ψ) = l(−F
b
sin θ
b
+ F
a
sin θ
a
)
6.3 PlanarMotion Mechanism 51
In this set of four equations, we know from the imposed motion the values
[U, ψ, a, ω]. From the experiment, we obtain [F
a
, F
b
, θ
a
, θ
b
], and from the rigid
body model we have [m, I
zz
, x
G
]. It turns out that the two cases of ψ = 0
(pure sway motion) and ψ = 180
o
(pure yaw motion) yield a total of eight
independent equations, exactly what is required to ﬁnd the eight coeﬃcients
[Y
˙ v
, Y
v
, Y
˙ r
, Y
r
, N
˙ v
, N
v
, N
˙ r
, N
r
]. Remarkably, we can write the eight solutions
directly: For ψ = 0,
(m−Y
˙ v
)
−
aω
2
2
(2) = F
a
cos θ
a
+ F
b
cos θ
b
(87)
−Y
v
−
aω
2
(2) = −F
a
sin θ
a
−F
b
sin θ
b
(mx
G
−N
˙ v
)
−
aω
2
2
(2) = l(F
b
cos θ
b
−F
a
cos θ
a
)
−N
v
−
aω
2
(2) = l(−F
b
sin θ
b
+ F
a
sin θ
a
),
to be solved respectively for [Y
˙ v
, Y
v
, N
˙ v
, N
v
]. For ψ = 180
o
, we have
(mx
G
−Y
˙ r
)
−
aω
2
2l
(−2) = F
a
cos θ
a
+ F
b
cos θ
b
(88)
(mU −Y
r
)
−
aω
2l
(−2) = −F
a
sin θ
a
−F
b
sin θ
b
(I
zz
−N
˙ r
)
−
aω
2
2l
(−2) = l(F
b
cos θ
b
−F
a
cos θ
a
)
(mx
G
U −N
r
)
−
aω
2l
(−2) = l(−F
b
sin θ
b
+ F
a
sin θ
a
),
to be solved for [Y
˙ r
, Y
r
, N
˙ r
, N
r
]. Thus, the eight linear coeﬃcients for a surface
vessel maneuvering, for a given speed, can be deduced from two tests with
a planar motion mechanism. We note that the nonlinear terms will play a
signiﬁcant role if the motions are too large, and that some curve ﬁtting will be
needed in any event. The PMM can be driven with more complex trajectories
which will target speciﬁc nonlinear terms.
52 7 STANDARD MANEUVERING TESTS
7 STANDARD MANEUVERING TESTS
This section describes some of the typical maneuvering tests which are per
formed on fullscale vessels, to assess stability and performance.
7.1 Dieudonn´e Spiral
1. Achieve steady speed and direction for one minute. No changes in speed
setting are made after this point.
2. Turn rudder quickly by 15
◦
, and keep it there until steady yaw rate is
maintained for one minute.
3. Reduce rudder angle by 5
◦
, and keep it there until steady yaw rate is
maintained for one minute.
4. Repeat in decrements of 5
◦
, to 15
◦
.
5. Proceed back up to 15
◦
.
The Dieudonn´e maneuver has the vessel path following a growing spiral, and
then a contracting spiral in the opposite direction. The test reveals if the vessel
has a memory eﬀect, manifested as a hysteresis in yaw rate r. For example,
suppose that the ﬁrst 15
◦
rudder deﬂection causes the vessel to turn right, but
that the yaw rate at zero rudder, on the way negative, is still to the right. The
vessel has gotten “stuck” here, and will require a negative rudder action to
pull out of the turn. But if the corrective action causes the vessel to turn left
at all, the same memory eﬀect may occur. It is easy to see that the rudder in
this case has to be used excessively driving the vessel back and forth. We say
that the vessel is unstable, and clearly a poor design.
7.2 ZigZag Maneuver
1. With zero rudder, achieve steady speed for one minute.
2. Deﬂect the rudder to 20
◦
, and hold until the vessel turns 20
◦
.
3. Deﬂect the rudder to 20
◦
, and hold until the vessel turns to 20◦ with
respect to the starting heading.
4. Repeat.
7.3 Circle Maneuver 53
This maneuver establishes several important characteristics of the yaw re
sponse. These are: the response time (time to reach a given heading), the yaw
overshoot (amount the vessel exceeds ±20
◦
when the rudder has turned the
other way), and the total period for the 20
◦
oscillations. Of course, similar
tests can be made with diﬀerent rudder angles and diﬀerent threshold vessel
headings.
7.3 Circle Maneuver
From a steady speed, zero yaw rate condition, the rudder is moved to a new
setting. The vessel responds by turning in a circle. After steady state is
reached again, parameters of interest are the turning diameter, the drift angle
β, the speed loss, and the angle of heel ψ.
7.3.1 Drift Angle
The drift angle is the equivalent to angle of attack for lifting surfaces, and
describes how the vessel “skids” during a turn. If the turning circle has radius
R (measured from the vessel origin), then the speed tangential to the circle
is U = rR. The vesselreference velocity components are thus u = U cos β
and v = −U sin β. A line along the vessel centerline reaches closest to the
true center of the turning circle at a point termed the turning center. At
this location, which may or may not exist on the physical vessel, there is no
apparent lateral velocity, and it appears to an observer there that the vessel
turns about this point.
7.3.2 Speed Loss
Speed loss occurs primarily because of drag induced by the drift angle. A
vessel which drifts very little may have very little speed loss.
7.3.3 Heel Angle
Heel during turning occurs as a result of the intrinsic coupling of sway, yaw,
and roll caused by the center of gravity. In a surface vessel, the ﬂuid forces act
below the waterline, but the center of gravity is near the waterline or above.
When the rudder is ﬁrst deﬂected, inertial terms dominate (Phase 1) and the
sway equation is
54 7 STANDARD MANEUVERING TESTS
(m−Y
˙ v
) ˙ v −(Y
˙ r
−mx
G
) ˙ r = Y
δ
δ. (89)
The coeﬃcients for ˙ r are quite small, and thus the vessel ﬁrst rolls to starboard
(positive) for a positive rudder action.
When steady turning conditions are reached (Phase 3), hydrodynamic forces
equalize the centrifugal force mUr and the rudder force Y
δ
δ. The sway equation
is
−Y
v
v + (mU −Y
r
)r = Y
δ
δ, (90)
with Y
r
small, v < 0 when r > 0 for most vessels, and [Y
v
v[ > [Y
δ
δ[. Because
the centrifugal force acts above the waterline, the vessel ultimately rolls to
port (negative) for a positive rudder action.
The transition between the inertiallydominated and steadyturning regimes
includes an overshoot: in the above formulas, the vessel overshoots the ﬁnal
port roll angle as the vessel slows. From the sway equation, we see that if
the rudder is straightened out at this point, the roll will momentarily be even
worse!
In summary, the vessel rolls into the turn initially, but then out of the turn in
the steady state.
7.3.4 Heeling in Submarines with Sails
Submarines typically roll into a turn during all phases. Unlike surface vessels,
which have the rigid mass center above the center of ﬂuid forcing, submarines
have the mass center below the rudder action point, and additionally feel the
eﬀects of a large sail above both. The inertial equation
(m−Y
˙ v
) ˙ v −(Y
˙ r
−mx
G
) ˙ r = Y
δ
δ (91)
is dominated by m˙ v (acting low), −Y
˙ v
˙ v (acting high), and Y
δ
δ (intermediate).
Because [Y
˙ v
[ > m, the vessel rolls under the sail, the keel out of the turn. In
the steady state,
−Y
v
v + (mU −Y
r
)r = Y
δ
δ. (92)
The drift angle β keeps the Y
v
force, acting high, toward the center of the turn,
and again centrifugal force mUr causes the bottom of the submarine to move
out of the turn. Hence, the roll angle of a submarine with a sail is always into
7.3 Circle Maneuver 55
the turn, both initially and in the steady state. The heel angle declines as the
speed of the submarine drops.
56 8 STREAMLINED BODIES
8 STREAMLINED BODIES
8.1 Nominal Drag Force
A symmetric streamlined body at zero angle of attack experiences only a drag
force, which has the form
F
A
= −
1
2
ρC
A
A
o
U
2
. (93)
The drag coeﬃcient C
A
has both pressure and skin friction components, and
hence area A
o
is usually that of the wetted surface. Note that the Asubscript
will be used to denote zero angle of attack conditions; also, the sign of F
A
is
negative, because it opposes the vehicle’s xaxis.
8.2 Munk Moment
Any shape other than a sphere generates a moment when inclined in an invis
cid ﬂow. d’Alembert’s paradox predicts zero net force, but not necessarily a
zero moment. This Munk moment arises for a simple reason, the asymmetric
location of the stagnation points, where pressure is highest on the front of
the body (decelerating ﬂow) and lowest on the back (accelerating ﬂow). The
Munk moment is always destabilizing, in the sense that it acts to turn the
vehicle perpendicular to the ﬂow.
Consider a symmetric body with added mass components A
xx
along the vehicle
(slender) xaxis (forward), and A
zz
along the vehicle’s zaxis z (up). We will
limit the present discussion to the vertical plane, but similar arguments can
be used to describe the horizontal plane. Let α represent the angle of attack,
taken to be positive with the nose up – this equates to a negative pitch angle
φ in vehicle coordinates, if it is moving horizontally. The Munk moment is:
M
m
= −
1
2
(A
zz
−A
xx
)U
2
sin 2α (94)
· −(A
zz
−A
xx
)U
2
α.
A
zz
> A
xx
for a slender body, and the negative sign indicates a negative
pitch with respect to the vehicle’s pitch axis. The added mass terms A
zz
and
A
xx
can be estimated from analytical expressions (available only for regular
shapes such as ellipsoids), from numerical calculation, or from slender body
approximation (to follow).
8.3 Separation Moment 57
8.3 Separation Moment
In a viscous ﬂuid, ﬂow over a streamlined body is similar to that of potential
ﬂow, with the exceptions of the boundary layer, and a small region near the
trailing end. In this latter area, a helical vortex may form and convect down
stream. Since vortices correlate with low pressure, the eﬀect of such a vortex
is stabilizing, but it also induces drag. The formation of the vortex depends on
the angle of attack, and it may cover a larger area (increasing the stabilizing
moment and drag) for a larger angle of attack. For a small angle of attack, the
transverse force F
n
can be written in the same form as for control surfaces:
F
n
=
1
2
ρC
n
A
o
U
2
(95)
·
1
2
ρ
∂C
n
∂α
αA
o
U
2
.
With a positive angle of attack, this force is in the positive zdirection. The
zeroα drag force F
A
is modiﬁed by the vortex shedding:
F
a
= −
1
2
ρC
a
A
o
U
2
, where (96)
C
a
= C
A
cos
2
φ.
The last relation is based on writing C
A
(U cos φ)
2
as (C
A
cos
2
φ)U
2
, i.e., a
decomposition using apparent velocity.
8.4 Net Eﬀects: Aerodynamic Center
The Munk moment and the moment induced by separation are competing,
and their magnitudes determine the stability of a hullform. First we simplify:
F
a
= −γ
a
F
n
= γ
n
α
M
m
= −γ
m
α.
Each constant γ is taken as positive, and the signs reﬂect orientation in the
vehicle reference frame, with a noseup angle of attack. The Munk moment is a
58 8 STREAMLINED BODIES
pure couple which does not depend on a reference point. We pick a temporary
origin O for F
n
however, and write the total pitch moment about O as:
M = M
m
+ F
n
l
n
(97)
= (−γ
m
+ γ
n
l
n
)α.
where l
n
denotes the (positive) distance between O and the application point
of F
n
. The net moment about O is zero if we select
l
n
=
γ
m
γ
n
, (98)
and the location of O is then called the aerodynamic center or AC.
The point AC has an intuitive explanation: it is the location on the hull where
F
n
would act to create the total moment. Hence, if the vehicle’s origin lies in
front of AC, the net moment is stabilizing. If the origin lies behind AC, the
moment is destabilizing. For selfpropelled vehicles, the mass center must be
forward of AC for stability. Similarly, for towed vehicles, the towpoint must
be located forward of AC. In many cases with very streamlined bodies, the
aerodynamic center is signiﬁcantly ahead of the nose, and in this case, a rigid
sting would have to extend at least to AC in order for stable towing. As a
ﬁnal note, since the Munk moment persists even in inviscid ﬂow, AC moves
inﬁnitely far forward as viscosity eﬀects diminish.
8.5 Role of Fins in Moving the Aerodynamic Center
Control surfaces or ﬁxed ﬁns are often attached to the stern of a slender vehicle
to enhance directional stability. Fixed surfaces induce lift and drag on the
body:
L =
1
2
ρA
f
U
2
C
l
(α) · γ
L
α (99)
D = −
1
2
ρA
f
U
2
C
d
· −γ
D
(constant)
These forces act somewhere on the ﬁn, and are signed again to match the
vehicle frame, with γ > 0 and α > 0. The summed forces on the body are
thus:
8.5 Role of Fins in Moving the Aerodynamic Center 59
X = F
a
−[D[ cos α +[L[ sin α (100)
· −γ
a
−γ
D
+ γ
L
α
2
Z = F
n
+[L[ cos α +[D[ sin α
· γ
n
α + γ
L
α + γ
D
α.
All of the forces are pushing the vehicle up. If we say that the ﬁns act a
distance l
f
behind the temporary origin O, and that the moment carried by
the ﬁns themselves is very small (compared to the moment induced by Ll
f
)
the total moment is as follows:
M = (−γ
m
+ γ
n
l
n
)α + (γ
L
+ γ
D
)l
f
α. (101)
The moment about O vanishes if
γ
m
= γ
n
l
n
+ l
f
(γ
L
+ γ
D
). (102)
The Munk moment γ
m
opposes the aggregate eﬀects of vorticity lift γ
n
and
the ﬁns’ lift and drag γ
L
+γ
D
. With very large ﬁns, this latter term is large, so
that l
f
might be very small; this is the case of AC moving aft toward the ﬁns.
A vehicle with excessively large ﬁns will be diﬃcult to turn and maneuver.
Equation 102 contains two length measurements, referenced to an arbitrary
body point O. To solve it explicitly, let l
fn
denote the (positive) distance that
the ﬁns are located behind F
n
; this is likely a small number, since both eﬀects
usually act near the stern. We solve for l
f
:
l
f
=
γ
m
+ γ
n
l
fn
γ
n
+ γ
L
+ γ
D
. (103)
This is the distance that AC is located forward of the ﬁns, and thus AC can
be referenced to any other ﬁxed point easily. Without ﬁns, it can be recalled
that
l
n
=
γ
m
γ
n
.
Hence, the ﬁns act directly in the denominator to shorten l
f
. Note that if
the ﬁns are located forward of the vortex shedding force F
n
, i.e., l
fn
< 0, l
f
is
reduced, but since AC is referenced to the ﬁns, there is no net gain in stability.
60 8 STREAMLINED BODIES
8.6 Aggregate Eﬀects of Body and Fins
Since all of the terms discussed so far have the same dependence on α, it is
possible to group them into a condensed form. Setting
ˆ
F
a
and
ˆ
F
n
to account
for the fuselage and ﬁns, we have
X =
ˆ
F
a
· −
1
2
ρ
ˆ
C
a
ˆ
A
o
U
2
(104)
Z =
ˆ
F
n
·
1
2
ρ
ˆ
C
n
ˆ
A
o
U
2
α
M = −
ˆ
F
n
x
AC
· −
1
2
ˆ
C
n
ˆ
A
o
U
2
x
AC
α.
8.7 Coeﬃcients Z
w
, M
w
, Z
q
, and M
q
for a Slender Body
The angle of attack α is related to the crossbody velocity w as follows:
α = −tan
−1
w
u
(105)
· −
w
U
for U >> w.
We can then write several linear hydrodynamic coeﬃcients easily:
Z
w
= −
1
2
ρ
ˆ
C
n
ˆ
A
o
U (106)
M
w
=
1
2
ρ
ˆ
C
n
ˆ
A
o
Ux
AC
.
The rotation of the vessel involves complex ﬂow, which depends on both w and
q, as well as their derivatives. To start, we consider the contribution of the ﬁns
only – slender body theory, introduced shortly, provides good results for the
hull. The ﬁn center of pressure is located a distance l
f
aft of the body origin,
and we assume that the vehicle is moving horizontally, with an instantaneous
pitch angle of θ. The angle of attack seen by the ﬁn is a combination of a part
due to θ and a part linear with q:
α
f
· −θ +
l
f
q
U
(107)
8.7 Coeﬃcients Z
w
, M
w
, Z
q
, and M
q
for a Slender Body 61
and so lateral force and moment derivatives (for the ﬁn alone) emerge as
Z
q
= −
1
2
ρC
l
A
f
Ul
f
(108)
M
q
= −
1
2
ρC
l
A
f
Ul
2
f
.
62 9 SLENDERBODY THEORY
9 SLENDERBODY THEORY
9.1 Introduction
Consider a slender body with d << L, that is mostly straight. The body
could be asymmetric in crosssection, or even ﬂexible, but we require that
the lateral variations are small and smooth along the length. The idea of
the slenderbody theory, under these assumptions, is to think of the body as
a longitudinal stack of thin sections, each having an easilycomputed added
mass. The eﬀects are integrated along the length to approximate lift force and
moment. Slenderbody theory is accurate for small ratios d/L, except near
the ends of the body.
As one example, if the diameter of a body of revolution is d(s), then we can
compute δm
a
(x), where the nominal added mass value for a cylinder is
δm
a
= ρ
π
4
d
2
δx. (109)
The added mass is equal to the mass of the water displaced by the cylinder.
The equation above turns out to be a good approximation for a number of
twodimensional shapes, including ﬂat plates and ellipses, if d is taken as the
width dimension presented to the ﬂow. Many formulas for added mass of two
dimensional sections, as well as for simple threedimensional bodies, can be
found in the books by Newman and Blevins.
9.2 Kinematics Following the Fluid
The added mass forces and moments derive from accelerations that a ﬂuid
particles experience when they encounter the body. We use the notion of a
ﬂuid derivative for this purpose: the operator d/dt indicates a derivative taken
in the frame of the passing particle, not the vehicle. Hence, this usage has an
indirect connection with the derivative described in our previous discussion of
rigidbody dynamics.
For the purposes of explaining the theory, we will consider the twodimensional
heave/surge problem only. The local geometry is described by the location of
the centerline; it has vertical location (in body coordinates) of z
b
(x, t), and
local angle α(x, t). The timedependence indicates that the conﬁguration is
free to change with time, i.e., the body is ﬂexible. Note that the curvilinear
coordinate s is nearly equal to the bodyreference (linear) coordinate x.
The velocity of a ﬂuid particle normal to the body at x is w
n
(t, x):
9.3 Derivative Following the Fluid 63
w
n
=
∂z
b
∂t
cos α −U sin α. (110)
The ﬁrst component is the time derivative in the body frame, and the second
due to the deﬂection of the particle by the inclined body. If the body reference
frame is rotated to the ﬂow, that is, if w = 0, then ∂z
b
/∂t will contain w. For
small angles, sin α · tan α = ∂z
b
/∂x, and we can write
w
n
·
∂z
b
∂t
−U
∂z
b
∂x
.
The ﬂuid derivative operator in action is as follows:
w
n
=
dz
b
dt
=
∂
∂t
−U
∂
∂x
z
b
.
9.3 Derivative Following the Fluid
A more formal derivation for the ﬂuid derivative operator is quite simple. Let
µ(x, t) represent some property of a ﬂuid particle.
d
dt
[µ(x, t)] = lim
δt→0
1
δt
[µ(x + δx, t + δt) −µ(x, t)]
=
¸
∂µ
∂t
−U
∂µ
∂x
¸
.
The second equality can be veriﬁed using a Taylor series expansion of µ(x +
δx, t + δt):
µ(x + δx, t + δt) = µ(x, t) +
∂µ
∂t
δt +
∂µ
∂x
δx + h.o.t.,
and noting that δx = −Uδt. The ﬂuid is convected downstream with respect
to the body.
64 9 SLENDERBODY THEORY
9.4 Diﬀerential Force on the Body
If the local transverse velocity is w
n
(x, t), then the diﬀerential inertial force on
the body here is the derivative (following the ﬂuid) of the momentum:
δF = −
d
dt
[m
a
(x, t)w
n
(x, t)] δx. (111)
Note that we could here let the added mass vary with time also – this is the
case of a changing crosssection! The lateral velocity of the point z
b
(x) in the
bodyreference frame is
∂z
b
∂t
= w −xq, (112)
such that
w
n
= w −xq −Uα. (113)
Taking the derivative, we have
δF
δx
= −
∂
∂t
−U
∂
∂x
[m
a
(x, t)(w(t) −xq(t) −Uα(x, t))] .
We now restrict ourselves to a rigid body, so that neither m
a
nor α may change
with time.
δF
δx
= m
a
(x)(− ˙ w + x ˙ q) + U
∂
∂x
[m
a
(x)(w −xq −Uα)] . (114)
9.5 Total Force on a Vessel
The net lift force on the body, computed with strip theory is
Z =
x
N
x
T
δFdx (115)
where x
T
represents the coordinate of the tail, and x
N
is the coordinate of the
nose. Expanding, we have
9.6 Total Moment on a Vessel 65
Z =
x
N
x
T
m
a
(x) [− ˙ w + x ˙ q] dx + U
x
N
x
T
∂
∂x
[m
a
(x)(w −xq −Uα)] dx
= −m
33
˙ w −m
35
˙ q + Um
a
(x)(w −xq −Uα)[
x=x
N
x=x
T
.
We made use here of the added mass deﬁnitions
m
33
=
x
N
x
T
m
a
(x)dx
m
35
= −
x
N
x
T
xm
a
(x)dx.
Additionally, for vessels with pointed noses and ﬂat tails, the added mass m
a
at the nose is zero, so that a simpler form occurs:
Z = −m
33
˙ w −m
35
˙ q −Um
a
(x
T
)(w −x
T
q −Uα(x
T
)). (116)
In terms of the linear hydrodynamic derivatives, the strip theory thus provides
Z
˙ w
= −m
33
Z
˙ q
= −m
35
Z
w
= −Um
a
(x
T
)
Z
q
= Ux
T
m
a
(x
T
)
Z
α(x
T
)
= U
2
m
a
(x
T
).
It is interesting to note that both Z
w
and Z
α(x
T
)
depend on a nonzero base
area. In general, however, potential ﬂow estimates do not create lift (or drag)
forces for a smooth body, so this should come as no surprise. The two terms are
clearly related, since their diﬀerence depends only on how the body coordinate
system is oriented to the ﬂow. Another noteworthy fact is that the lift force
depends only on α at the tail; α could take any value(s) along the body, with
no eﬀect on Z.
9.6 Total Moment on a Vessel
A similar procedure can be applied to the moment predictions from slender
body theory (again for small α):
66 9 SLENDERBODY THEORY
M = −
x
N
x
T
xδFdx
=
x
N
x
T
x
∂
∂t
−U
∂
∂x
[m
a
(x)(w −xq −Uα)] dx
=
x
N
x
T
xm
a
(x)( ˙ w −x ˙ q)dx −U
x
N
x
T
x
∂
∂x
[m
a
(x)(w −xq + Uα)] dx.
Then we make the further deﬁnition
m
55
=
x
N
x
T
x
2
m
a
(x)dx,
(note that m
35
= m
53
) and use integration by parts to obtain
M = −m
35
˙ w −m
55
˙ q −Uxm
a
(x)(w −xq −Uα)[
x=x
N
x=x
T
+
U
x
N
x
T
m
a
(x)(w −xq −Uα)dx.
The integral above contains the product m
a
(x)α(x), which must be calculated
if α changes along the length. For simplicity, we now assume that α is in fact
constant on the length, leading to
M = −m
35
˙ w −m
55
˙ q + Ux
T
m
a
(x
T
)(w −x
T
q −Uα) +
Um
33
w + Um
35
q −U
2
m
33
α.
Finally, the linear hydrodynamic moment derivatives are
M
˙ w
= −m
35
M
˙ q
= −m
55
M
w
= Ux
T
m
a
(x
T
) + Um
33
M
q
= −Ux
2
T
m
a
(x
T
) + Um
35
M
α
= −U
2
x
T
m
a
(x
T
) −U
2
m
33
.
The derivative M
w
is closelyrelated to the Munk moment, whose linearization
would provide M
w
= (m
33
−m
11
)U. The Munk moment (an exact result) may
therefore be used to make a correction to the second term in the slenderbody
approximation above of M
w
. As with the lift force, M
w
and M
α
are closely
related, depending only on the orientation of the body frame to the ﬂow.
9.7 Relation to Wing Lift 67
9.7 Relation to Wing Lift
There is an important connection between the slender body theory terms in
volving added mass at the tail (m
a
(x
T
)), and low aspectratio wing theory.
The lift force from the latter is of the form L = −
1
2
ρUAC
l
w, where A = cs,
the product of chord (long) and span (short).
The lift coeﬃcient slope is approximated by (Hoerner)
C
l
≈
1
2
π(AR), (117)
where (AR) is the aspect ratio. Inserting this approximation into the lift
formula, we obtain
L = −
π
4
ρs
2
Uw. (118)
Now we look at a slender body approximation of the same force: The added
mass at the tail is m
a
(x
T
) = ρs
2
π/4, and using the slenderbody estimate for
Z
w
, we calculate for lift:
Z = −m
a
(x
T
)Uw
= −
π
4
ρs
2
Uw.
Slenderbody theory is thus able to recover exactly the lift of a lowaspect
ratio wing. Where does the slenderbody predict the force will act? Recalling
that M
w
= Um
33
+Ux
T
m
a
(x
T
), and since m
33
= 0 for a frontback symmetric
wing, the estimated lift force acts at the trailing edge. This location will tend
to stabilize the wing, in the sense that it acts to orient the wing parallel to
the incoming ﬂow.
9.8 Convention: Hydrodynamic Mass Matrix A
Hydrodynamic derivatives that depend on accelerations are often written as
components of a mass matrix A. By listing the bodyreferenced velocities
in the order s = [u, v, w, p, q, r], we write (M + A)
˙
s =
F, where M is the
mass matrix of the material vessel and F is a generalized force. Therefore
A
33
= −Z
˙ w
, A
5,3
= −M
˙ w
, and so on.
68 10 PRACTICAL LIFT CALCULATIONS
10 PRACTICAL LIFT CALCULATIONS
10.1 Characteristics of LiftProducing Mechanisms
At a small angle of attack, a slender body experiences transverse force due to:
helical body vortices, the blunt trailing end, and ﬁns. The helical body vortices
are stable and symmetric in this condition, and are convected continuously into
the wake. The low pressure associated with the vortices provides the suction
force, usually toward the stern of the vehicle. The blunt trailing end induces
lift as a product of added mass eﬀects, and can be accurately modeled with
slender body theory. A blunt trailing edge also induces some drag, which itself
is stabilizing. The ﬁns can often be properly modeled using experimental data
parametrized with aspect ratio and several other geometric quantities.
As the angle of attack becomes larger, the approximations in the ﬁn and
slenderbody analysis will break down. The helical vortices can become bigger
while remaining stable, but eventually will split randomly. Some of it convects
downstream, and the rest peels away from the body; this shedding is nonsym
metric, and greatly increases drag by widening the wake. In the limit of a 90
◦
angle of attack, vorticity sheds as if from a bluﬀbody, and there is little axial
convection.
10.2 Jorgensen’s Formulas
There are some heuristic and theoretical formulas for predicting transverse
force and moment on a body at various angles of attack, and we now present
one of them due to Jorgensen. The formulas provide a good systematic proce
dure for design, and are best suited to vehicles with a blunt trailing edge. We
call the area of the stern the base area.
Let the body have length L, and reference area A
r
. This area could be the
frontal projected area, the planform area, or the wetted area. The body travels
at speed U, and angle of attack α. The normal force, axial force, and moment
coeﬃcients are deﬁned as follows:
C
N
=
F
N
1
2
ρU
2
A
r
(119)
C
A
=
F
A
1
2
ρU
2
A
r
10.2 Jorgensen’s Formulas 69
C
M
=
M
xm
1
2
ρU
2
A
r
L
.
The moment M
xm
is taken about a point x
m
, measured back from the nose; this
location is arbitrary, and appears explicitly in the formula for C
m
. Jorgensen
gives the coeﬃcients as follows:
C
N
=
A
b
A
r
sin 2αcos
α
2
+
A
p
A
r
C
dn
sin
2
α (120)
C
A
= C
Ao
cos
2
α (121)
C
M
= −
∇−A
b
(L −x
m
)
A
r
L
sin 2αcos
α
2
−C
dn
A
p
A
r
x
m
−x
c
L
sin
2
α.(122)
We have listed only the formulas for the special case of circular crosssection,
although the complete equations do account for more complex shapes. Further,
we have assumed that L >> D, which is also not a constraint in the complete
equations. The parameters used here are
• A
b
: stern base area. A
b
= 0 for a body that tapers to a point at the
stern.
• C
dn
: crossﬂow drag coeﬃcient; equivalent to that of an inﬁnite circular
cylinder. If “normal” Reynolds number Re
n
= U sin αD/ν,
– C
d
≈ 1.2, Re
n
< 3 10
5
– C
d
≈ 0.3, 3 10
5
< Re
n
< 7 10
5
– C
d
≈ 0.6, 7 10
5
< Re
n
.
• A
p
: planform area.
• C
Ao
: axial drag at zero angle of attack, both frictional and form. C
Ao
·
0.002 − 0.006 for slender streamlined bodies, based on wetted surface
area. It depends on Re = UL/ν.
• ∇: body volume.
• x
c
: distance from the nose backwards to the center of the planform area.
70 10 PRACTICAL LIFT CALCULATIONS
In the formula for normal force, we see that if A
b
= 0, only drag forces act
to create lift, through a sin
2
αdependence. Similarly, the axial force is simply
the zeroα result, modiﬁed by cos
2
α. In both cases, scaling of U
2
into the
body principle directions is all that is required.
There are several terms that match exactly the slenderbody theory approxi
mations for small α. These are the ﬁrst term in the normal force (C
N
), and
the entire ﬁrst term in the moment (C
M
), whether or not A
b
= 0. Finally, we
note that the second term in C
M
disappears if x
m
= x
c
, i.e., if the moment is
referenced to the center of the planform area. The idea here is that the fore
and aft components of crossﬂow drag cancel out.
The aerodynamic center (again referenced toward the stern, from the nose)
can be found after the coeﬃcients are computed:
x
AC
= x
m
+
C
M
C
N
L. (123)
As written, the moment coeﬃcient is negative if the moment destabilizes the
body, while C
N
is always positive. Thus, the moment seeks to move the AC
forward on the body, but the eﬀect is moderated by the lift force.
10.3 Hoerner’s Data: Notation
An excellent reference for experimental data is the twovolume set by S. Ho
erner. It contains a large amount of aerodynamic data from many diﬀerent
types of vehicles, wings, and other common engineering shapes. A few nota
tions are used throughout the books, and are described here.
First, dynamic pressure is given as q =
1
2
ρU
2
, such that two typical body lift
coeﬃcients are:
C
Y
=
Y
DLq
C
Y
d
=
Y
D
2
q
.
The ﬁrst version uses the rectangular planform area as a reference, while the
second uses the square frontal area. Hence, C
Y
= C
Y
d
D/L. Two moment
coeﬃcients are:
10.4 SlenderBody Theory vs. Experiment 71
C
N
=
N
LD
2
q
C
N
1
=
N
1
LD
2
q
,
where N is the moment taken about the body midpoint, and N
1
is taken
about the nose. Note that the reference area for moment is the square frontal
projection, and the reference length is body length L. The following relation
holds for these deﬁnitions
C
N
1
= C
N
+
1
2
C
Y
d
.
The lift and moment coeﬃcients are strongly dependent on angle of attack;
Hoerner uses the notation
C
nb
=
∂C
N
∂b
C
n·b
=
∂C
N
1
∂b
C
yb
=
∂C
y
∂b
,
and so on, where b is the angle of attack, usually in degrees. It follows from
above that C
n·b
= C
nb
+ C
ydb
/2.
10.4 SlenderBody Theory vs. Experiment
In an experiment, the net moment is measured, comprising both the desta
bilizing part due to the potential ﬂow, and the stabilizing part due to vortex
shedding or a blunt tail. Comparison of the measurements and the theory
allows us to place the action point of the suction force. This section gives the
formula for this location in Hoerner’s notation, and gives two further examples
of how well the slenderbody theory matches experiments.
For L/D > 6, the slenderbody (pure added mass) estimates give
˜
C
nb
·
−0.015/deg, acting to destabilize the vehicle. The value compares well with
−0.027/deg for a long cylinder and −0.018/deg for a long ellipsoid; it also
72 10 PRACTICAL LIFT CALCULATIONS
reduces to −0.009 for L/D = 4. Note that the negative sign here is consistent
with Hoerner’s convention that destabilizing moments have negative sign.
The experimental lift force is typically given by C
Y b
· 0.003/deg; this acts at
a point on the latter half of the vehicle, stabilizing the angle. Because this
coeﬃcient scales roughly with wetted area, proportional to LD, it changes
little with L/D. It can be compared with a lowaspect ratio wing, which
achieves an equivalent lift of π(AR)/2 = 0.0027 for (AR) = 10 · D/L.
The point at which the viscous forces act can then be estimated as the following
distance aft of the nose:
x
L
=
C
n·b
−
˜
C
nb
C
ydb
(124)
The calculation uses experimental values of C
ydb
and C
n·b
, the moment slope
referenced to the nose. In the table following are values from Hoerner (p. 132,
Figure 2) for a symmetric and a bluntended body.
symmetric blunt
L/D 6.7 6.7
˜
C
nb
0.012 0.012
C
yb
0.0031 0.0037
C
ydb
0.021 0.025
C
n·b
0.0012 (stable) 0.0031 (stable)
C
nb
0.0093 (unstable) 0.0094 (unstable)
x/L 0.63 0.60
In comparing the two body shapes, we see that the moment at the nose is
much more stable (positive) for the body with a blunt trailing edge. At the
body midpoint, however, both vehicles are equally unstable. The blunttailed
geometry has a much larger lift force, but it acts too close to the midpoint to
add any stability there.
The lift force dependence on the blunt tail is not diﬃcult to see, using slender
body theory. Consider a body, with trailing edge radius r. The slenderbody
lift force associated with this end is simply the product of speed U and local
added mass m
a
(x
T
) (in our previous notation). It comes out to be
Z =
1
2
ρU
2
(πr
2
)(2α), (125)
such that the ﬁrst term in parentheses is an eﬀective area, and the second is
a lift coeﬃcient. With respect to the area πr
2
, the lift curve slope is therefore
10.5 SlenderBody Approximation for Fin Lift 73
2/rad. Expressed in terms of the Hoerner reference area D
2
, the equivalent
lift coeﬃcient is C
ydb
= 0.0044/deg, where we made the assumption here that
2r/D = 0.4 for the data in the table. This lift diﬀerence, due solely to the
blunt end condition, is consistent with measurements.
10.5 SlenderBody Approximation for Fin Lift
Let us now consider two ﬁns of span s each, acting at the tail end of the
vehicle. This is the case if the vehicle body tapers to a point where the ﬁns
have their trailing edge. The slenderbody approximation of lift as a result of
bluntend conditions is
Z = πs
2
ρU
2
α. (126)
Letting the aspect ratio be (AR) = (2s)
2
/A
f
, where A
f
is the total area of the
ﬁn pair, substitution will give a lift curve slope of
C
l
=
π
2
(AR).
This is known as Jones’ formula, and is quite accurate for (AR) · 1. It is
inadequate for higheraspect ratio wings however, overestimating the lift by
about 30% when (AR) = 2, and worsening further as (AR) grows.
74 11 FINS AND LIFTING SURFACES
11 FINS AND LIFTING SURFACES
Vessels traveling at signiﬁcant speed typically use rudders, elevators, and other
streamlined control surfaces to maneuver. Their utility arises mainly from the
high lift forces they can develop, with little drag penalty. Lift is always deﬁned
to act in a direction perpendicular to the ﬂow, and drag in the same direction
as the ﬂow.
11.1 Origin of Lift
A lifting surface is nominally an extrusion of a streamlined crosssection: the
crosssection has a rounded leading edge, sharp trailing edge, and a smooth
surface. The theory of lifting surfaces centers on the Kutta condition, which
requires that ﬂuid particle streamlines do not wrap around the trailing edge
of the surface, but instead rejoin with streamlines from the other side of the
wing at the trailing edge. This fact is true for a nonstalled surface at any
angle of attack.
Since the separation point on the front of the section rotates with the angle
of attack, it is clear that the ﬂuid must travel faster over one side of the
surface than the other. The reduced Bernoulli pressure this induces can then
be thought of as the liftproducing mechanism. More formally, lift arises from
circulation Γ:
Γ =
V ds. (127)
and then L = −ρUΓ. Circulation is the integral of velocity around the cross
section, and a lifting surface requires circulation in order to meet the Kutta
condition.
11.2 ThreeDimensional Eﬀects: Finite Length
Since all practical lifting surfaces have ﬁnite length, the ﬂow near the ends
may be threedimensional. Prandtl’s inviscid theory provides some insight.
Since bound circulation cannot end abruptly at the wing end, it continues on
in the ﬂuid, leading to socalled wingtip vortices. This continuation causes
induced velocities at the tips, and some induced drag. Another description for
the wingtip vortices is that the pressure diﬀerence across the surface simply
causes ﬂow around the end.
11.3 Ring Fins 75
A critical parameter which governs the extent of threedimensional eﬀects is
the aspect ratio:
AR =
span
chord
=
span
2
area
. (128)
The second representation is useful for nonrectangular control surfaces. The
eﬀective span is taken to be the length between the free ends of a symmetric
wing. If the wing is attached to a wall, the eﬀective span is twice the physical
value, by reﬂection, and in this case the eﬀective aspect ratio is therefore twice
the physical value.
The aspect ratio is a strong determinant of wing performance: for a given
angle of attack, a larger aspect ratio achieves a higher lift value, but also stalls
earlier.
Lift is written as
L =
1
2
ρU
2
AC
l
, (129)
where A is the singleside area of the surface. For angles of attack α below
stall, the lift coeﬃcient C
l
is nearly linear with α: C
l
= C
l
α, where C
l
is called
the lift coeﬃcient slope, and has one empirical description
C
l
=
1
1
2π¯ α
+
1
π(AR)
+
1
2π(AR)
2
, (130)
where α is in radians, ¯ α · 0.90, and AR is the eﬀective aspect ratio. When
AR →∞, the theoretical and maximum value for C
l
is 2π.
The lift generated on a surface is the result of a distributed pressure ﬁeld; this
fact creates both a net force and a net moment. A single equivalent force acts
at a socalled center of action x
A
, which depends mainly on the aspect ratio.
For high AR, x
A
· c/4, measured back from the front of the wing. For low
AR, x
A
· c/2.
11.3 Ring Fins
Ring ﬁns are useful when space allows, since they are omnidirectional, and
structurally more robust than cantilevered plane surfaces. The eﬀective aspect
ratio for a ring of diameter d is given as
AR =
4d
πc
. (131)
76 11 FINS AND LIFTING SURFACES
The eﬀective area of the ring is taken as
A
e
=
π
2
dc, (132)
and we thus have L =
1
2
ρU
2
A
e
C
l
α, where one formula for C
l
is
C
l
=
1
0.63 +
1
π(AR)
. (133)
77
12 PROPELLERS AND PROPULSION
12.1 Introduction
We discuss in this section the nature of steady and unsteady propulsion. In
many marine vessels and vehicles, an engine (diesel or gas turbine, say) or an
electric motor drives the propeller through a linkage of shafts, reducers, and
bearings, and the eﬀects of each part are important in the response of the net
system. Large, commercial surface vessels spend the vast majority of their time
operating in openwater and at constant speed. In this case, steady propulsion
conditions are generally optimized for fuel eﬃciency. An approximation of the
transient behavior of a system can be made using the quasistatic assumption.
In the second section, we list several loworder models of thrusters, which have
recently been used to model and simulate truly unsteady conditions.
12.2 Steady Propulsion of Vessels
The notation we will use is as given in Table 1, and there are two diﬀerent ﬂow
conditions to consider. Selfpropelled conditions refer to the propeller being
installed and its propelling the vessel; there are no additional forces or moments
on the vessel, such as would be caused by a towing bar or hawser. Furthermore,
the ﬂow around the hull interacts with the ﬂow through the propeller. We use
an sp subscript to indicate speciﬁcally selfpropulsion conditions. Conversely,
when the propeller is run in open water, i.e., not behind a hull, we use an o
subscript; when the hull is towed with no propeller we use a t subscript. When
subscripts are not used, generalization to either condition is implied. Finally,
because of similitude (using diameter D in place of L when the propeller is
involved), we do not distinguish between the magnitude of forces in model and
fullscale vessels.
12.2.1 Basic Characteristics
In the steady state, force balance in selfpropulsion requires that
R
sp
= T
sp
. (134)
The gear ratio λ is usually large, indicating that the propeller turns much
more slowly than the driving engine or motor. The following relations deﬁne
the gearbox:
78 12 PROPELLERS AND PROPULSION
R
sp
N hull resistance under selfpropulsion
R
t
N towed hull resistance (no propeller attached)
T N thrust of the propeller
n
e
Hz rotational speed of the engine
n
m
Hz maximum value of n
e
n
p
Hz rotational speed of the propeller
λ gear ratio
Q
e
Nm engine torque
Q
p
Nm propeller torque
η
g
gearbox eﬃciency
P
e
W engine power
P
p
W propeller shaft power
D m propeller diameter
U m/s vessel speed
U
p
m/s water speed seen at the propeller
Q
m
Nm maximum engine torque
f kg/s fuel rate (or energy rate in electric motor)
f
m
kg/s maximum value of f
Table 1: Nomenclature
n
e
= λn
p
(135)
Q
p
= η
g
λQ
e
,
and power follows as P
p
= η
g
P
e
, for any ﬂow condition. We call J = U
p
/n
p
D
the advance ratio of the prop when it is exposed to a water speed U
p
; note
that in the wake of the vessel, U
p
may not be the same as the speed of the
vessel U. A propeller operating in open water can be characterized by two
nondimensional parameters which are both functions of J:
K
T
=
T
o
ρn
2
p
D
4
(thrust coeﬃcient) (136)
K
Q
=
Q
po
ρn
2
p
D
5
(torque coeﬃcient). (137)
12.2 Steady Propulsion of Vessels 79
0.2 0.4 0.6 0.8 1.0
0.0
0.2
0.4
0.6
0.8
1.0 0.10
0.08
0.06
0.04
0.02
0.00
0.0
K
η
0
T
K
Q
η
0
K
T
J
K
Q
Figure 4: Typical thrust and torque coeﬃcients.
The open propeller eﬃciency can be written then as
η
o
=
T
o
U
2πn
p
Q
po
=
J(U)K
T
2πK
Q
. (138)
This eﬃciency divides the useful thrust power by the shaft power. Thrust and
torque coeﬃcients are typically nearly linear over a range of J, and therefore
ﬁt the approximate form:
K
T
(J) = β
1
−β
2
J (139)
K
Q
(J) = γ
1
−γ
2
J.
As written, the four coeﬃcients [β
1
, β
2
, γ
1
, γ
2
] are usually positive, as shown in
the ﬁgure.
We next introduce three factors useful for scaling and parameterizing our
mathematical models:
• U
p
= U(1 − w); w is referred to as the wake fraction. A typical wake
fraction of 0.1, for example, indicates that the incoming velocity seen by
the propeller is only 90% of the vessel’s speed. The propeller is operating
in a wake.
In practical terms, the wake fraction comes about this way: Suppose the
open water thrust of a propeller is known at a given U and n
p
. Behind
80 12 PROPELLERS AND PROPULSION
a vessel moving at speed U, and with the propeller spinning at the same
n
p
, the prop creates some extra thrust. w scales U at the prop and thus
J; w is then chosen so that the open water thrust coeﬃcient matches
what is observed. The wake fraction can also be estimated by making
direct velocity measurements behind the hull, with no propeller.
• R
t
= R
sp
(1 − t). Often, a propeller will increase the resistance of the
vessel by creating lowpressure on its intake side (near the hull), which
makes R
sp
> R
t
. In this case, t is a small positive number, with 0.2 as a
typical value. t is called the thrust deduction even though it is used to
model resistance of the hull; it is obviously speciﬁc to both the hull and
the propeller(s), and how they interact.
The thrust deduction is particularly useful, and can be estimated from
published values, if only the towed resistance of a hull is known.
• Q
po
= η
R
Q
psp
. The rotative eﬃciency η
R
, which may be greater than
one, translates selfpropelled torque to open water torque, for the same
incident velocity U
p
, thrust T, and rotation rate n
p
. η
R
is meant to
account for spatial variations in the wake of the vessel which are not
captured by the wake fraction, as well as the turbulence induced by the
hull. Note that in comparison with the wake fraction, rotative eﬃciency
equalizes torque instead of thrust.
A common measure of eﬃciency, the quasipropulsive eﬃciency, is based on
the towed resistance, and the selfpropelled torque.
η
QP
=
R
t
U
2πn
p
Q
psp
(140)
=
T
o
(1 −t)U
p
η
R
2πn
p
(1 −w)Q
po
= η
o
η
R
(1 −t)
(1 −w)
.
T
o
and Q
po
are values for the inﬂow speed U
p
, and thus that η
o
is the open
propeller eﬃciency at this speed. It follows that T
o
(U
p
) = T
sp
, which was used
to complete the above equation. The quasipropulsive eﬃciency can be greater
than one, since it relies on the towed resistance and in general R
t
> R
sp
. The
ratio (1 − t)/(1 − w) is often called the hull eﬃciency, and we see that a
12.2 Steady Propulsion of Vessels 81
small thrust deduction t and a large wake fraction w are beneﬁcial eﬀects, but
which are in competition. A high rotative eﬃciency and open water propeller
eﬃciency (at U
p
) obviously contribute to an eﬃcient overall system.
12.2.2 Solution for Steady Conditions
The linear form of K
T
and K
Q
(Equation 140) allows a closedform solution
for the steadyoperating conditions. Suppose that the towed resistance is of
the form
R
t
=
1
2
ρC
r
A
w
U
2
, (141)
where C
r
is the resistance coeﬃcient (which will generally depend on Re and
Fr), and A
w
is the wetted area. Equating the selfpropelled thrust and resis
tance then gives
T
sp
= R
sp
T
o
= R
t
/(1 −t)
K
T
(J(U
p
))ρn
2
p
D
4
(1 −t) =
1
2
ρC
r
A
w
U
2
(β
1
−β
2
J(U
p
))ρn
2
p
D
4
(1 −t) =
1
2
ρC
r
A
w
U
2
p
(1 −w)
2
β
1
−β
2
J(U
p
) =
C
r
A
w
2D
2
(1 −t)(1 −w)
2
. .. .
δ
J(U
p
)
2
J(U
p
) =
−β
2
+
β
2
2
+ 4β
1
δ
2δ
. (142)
The last equation predicts the steadystate advance ratio of the vessel, depend
ing only on the propeller open characteristics, and on the hull. The vessel speed
can be computed by recalling that J(U) = U/n
p
D and U
p
= U(1 −w), but it
is clear that we need now to ﬁnd n
p
. This requires a torque equation, which
necessitates a model of the drive engine or motor.
12.2.3 Engine/Motor Models
The torquespeed maps of many engines and motors ﬁt the form
82 12 PROPELLERS AND PROPULSION
1
2
3
f
f
f
m
Q /Q
e
n /n
e m
Figure 5: Typical gas turbine engine torquespeed characteristic for increasing
fuel rates f
1
, f
2
, f
3
.
Q
e
= Q
m
F(f/f
m
, n
e
/n
m
), (143)
where F() is the characteristic function. For example, gas turbines roughly ﬁt
the curves shown in the ﬁgure (Rubis). More speciﬁcally, if F() has the form
F(f/f
f
m, n
e
/n
m
) = −
a
f
f
m
+ b
n
e
n
m
+
c
f
f
m
+ d
(144)
= −α
1
n
p
n
m
/λ
+ α
2
.
then a closedform solution for n
e
(and thus n
p
) can be found. The manipula
tions begin by equating the engine and propeller torque:
Q
po
(J(U
p
)) = η
R
Q
psp
(J(U
p
))
ρn
2
p
D
5
K
Q
(J(U
p
)) = η
R
η
g
λQ
e
ρn
2
p
D
5
(γ
1
−γ
2
J(U
p
)) = η
R
η
g
λQ
m
F(f/f
m
, n
e
/n
m
)
n
2
p
(n
m
/λ)
2
=
η
R
η
g
λQ
m
ρD
5
(γ
1
−γ
2
J(U
p
))(n
m
/λ)
2
. .. .
−α
1
n
p
(n
m
/λ)
+ α
2
n
p
(n
m
/λ)
=
−α
1
+
2
α
2
1
+ 4α
2
2
. (145)
12.3 Unsteady Propulsion Models 83
Note that the fuel rate enters through both α
1
and α
2
.
The dynamic response of the coupled propulsion and ship systems, under the
assumption of quasistatic propeller conditions, is given by
(m + m
a
) ˙ u = T
sp
−R
sp
(146)
2πI
p
˙ n
p
= η
g
λQ
e
−Q
psp
.
Making the necessary substitutions creates a nonlinear model with f as the
input; this is left as a problem for the reader.
12.3 Unsteady Propulsion Models
When accurate positioning of the vehicle is critical, the quasistatic assumption
used above does not suﬃce. Instead, the transient behavior of the propulsion
system needs to be considered. The problem of unsteady propulsion is still in
development, although there have been some very successful models in recent
years. It should be pointed out that the models described below all pertain to
openwater conditions and electric motors, since the positioning problem has
been central to bluﬀ vehicles with multiple electric thrusters.
We use the subscript m to denote a quantity in the motor, and p for the
propeller.
12.3.1 OneState Model: Yoerger et al.
The torque equation at the propeller and the thrust relation are
I
p
˙ ω
p
= λQ
m
−K
ω
ω
p
[ω
p
[ (147)
T = C
t
ω
p
[ω
p
[. (148)
where I
p
is the total (material plus ﬂuid) inertia reﬂected to the prop;the
propeller spins at ω
p
radians per second. The diﬀerential equation in ω
p
pits
the torque delivered by the motor against a quadraticdrag type loss which
depends on rotation speed. The thrust is then given as a static map directly
from the rotation speed.
This model requires the identiﬁcation of three parameters: I
p
, K
ω
, and C
t
.
It is a ﬁrstorder, nonlinear, lowpass ﬁlter from Q
m
to T, whose bandwidth
depends directly on Q
m
.
84 12 PROPELLERS AND PROPULSION
12.3.2 TwoState Model: Healey et al.
The twostate model includes the velocity of a mass of water moving in the
vicinity of the blades. It can accommodate a tunnel around the propeller,
which is very common in thrusters for positioning. The torque equation, sim
ilarly to the above, is referenced to the motor and given as
I
m
˙ ω
m
= −K
ω
ω
m
+ K
v
V −Q
p
/λ. (149)
Here, K
ω
represents losses in the motor due to spinning (friction and resistive),
and K
v
is the gain on the input voltage (so that the current ampliﬁer is included
in K
v
). The second dynamic equation is for the ﬂuid velocity at the propeller:
ρALγ
˙
U
p
= −ρA∆β(U
p
−U)[U
p
−U[ + T. (150)
Here A is the disc area of the tunnel, or the propeller disk diameter if no tunnel
exists. L denotes the length of the tunnel, and γ is the eﬀective added mass
ratio. Together, ρALγ is the added mass that is accelerated by the blades; this
mass is always nonzero, even if there is no tunnel. The parameter ∆β is called
the diﬀerential momentum ﬂux coeﬃcient across the propeller; it may be on
the order of 0.2 for propellers with tunnels, and up to 2.0 for open propellers.
The thrust and torque of the propeller are approximated using wing theory,
which invokes lift and drag coeﬃcients, as well as an eﬀective angle of attack
and the propeller pitch. However, these formulae are static maps, and therefore
introduce no new dynamics. As with the onestate model of Yoerger et al., this
version requires the identiﬁcation of the various coeﬃcients from experiments.
This model has the advantage that it creates a thrust overshoot for a step
input, which is in fact observed in experiments.
85
13 TOWING OF VEHICLES
Vehicles which are towed have some similarities to the vehicles that have been
discussed so far. For example, towed vehicles are often streamlined, and usu
ally need good directional stability. Some towed vehicles might have active
lifting surfaces or thrusters for attitude control. On the other hand, if they
are to be supported by a cable, towed vehicles may be quite heavy in water,
and do not have to be selfpropelled. The cable itself is an important factor in
the behavior of the complete towed system, and in this section, we concentrate
on cable mechanics more than vehicle characteristics, which can generally be
handled with the same tools as other vehicles, i.e., slenderbody theory, wing
theory, linearization, etc.. Some basic guidelines for vehicle design are given
at the end of this section.
Modern cables can easily exceed 5000m in length, even a heavy steel cable with
2cm diameter. The cables are generally circular in cross section, and may
carry power conductors and multiple communication channels (ﬁber optic).
The extreme L/D ratio for these cables obviates any bending stiﬀness eﬀects.
Cable systems come in a variety of conﬁgurations, and one main division may
be made simply of the density of the cable. Lighttether systems are character
ized by neutrallybuoyant (or nearly so) cables, with either a minimal vehicle
at the end, as in a towed array, or a vehicle capable of maneuvering itself, such
as a remotelyoperated vehicle. The towed array is a relatively highvelocity
system that nominally streams out horizontally behind the vessel. An ROV,
on the other hand, operates at low speed, and must have large propulsors to
control the tether if there are currents. Heavy systems, in contrast, employ a
heavy cable and possibly a heavy weight; the rationale is that gravity will tend
to keep the cable vertical and make the deployment robust against currents
and towing speed. The heavy systems will generally transmit surface motions
and tensions to the towed vehicle much more easily than lighttether systems.
We will not discuss light systems speciﬁcally here, but rather look at heavy
systems. Most of the analysis can be adapted to either case, however.
13.1 Statics
13.1.1 Force Balance
For the purposes of deriving the static conﬁguration of a cable in a ﬂow, we
assume for the moment that that it is inextensible. Tension and hydrostatic
86 13 TOWING OF VEHICLES
n
T
ds
φ
T+dT
ds W
ds R
n
ds R
t
s+
U
x
y
pressure will elongate a cable, but the eﬀect is usually a small percentage of
the total length.
We employ the curvilinear axial coordinate s, which we take to be zero at the
bottom end of the cable; upwards along the cable is the positive direction.
The freebody diagram shown has the following components:
• W
n
: net inwater weight of the cable per unit length.
• R
n
(s): external normal force, per unit length.
• R
t
(s): external tangential force, per unit length.
• T(s): local tension.
• φ(s): local inclination angle.
Force balance in the tangential and normal coordinates gives two coupled
equations for T and φ:
dT
ds
amp; = amp; W
n
sin φ −R
t
(151)
T
dφ
ds
amp; = amp; W
n
cos φ + R
n
. (152)
The external forces are primarily ﬂuid drag; the tangential drag is controlled
by a frictional drag coeﬃcient C
t
, and the normal drag scales with a crossﬂow
drag coeﬃcient C
n
. In both cases, the ﬂuid velocity vector, U horizontal
toward the left, is to be projected onto the relevant axes, leading to
13.1 Statics 87
R
t
= −
1
2
ρC
t
dU
2
cos
2
φ (153)
R
n
= −
1
2
ρC
n
dU
2
sin
2
φ. (154)
Note that we simpliﬁed the drag laws from a usual form v[v[ to v
2
, since as
drawn, 0 ≤ φ ≤ π/2.
The equations for T and φ can be integrated along the cable coordinate s to
ﬁnd the cable’s static conﬁguration. Two boundary conditions are needed, and
the common case is that a force balance on the vehicle, dominated by drag,
weight, and the cable tension, provides both T(0) and φ(0). For example, a
very heavy but lowdrag vehicle will impose φ(0) · π/2, with T(0) equal to
the inwater weight of the vehicle.
With regard to Cartesian coordinates x, y, the cable conﬁguration follows
dx
ds
amp; = amp; cos φ (155)
dy
ds
amp; = amp; sin φ. (156)
The simultaneous integration of all four equations (T, φ, x, y) deﬁnes the cable
conﬁguration, and current dependency may be included, say U is a function
of y.
13.1.2 Critical Angle
For very deep systems, the total weight of cable will generally exceed that the
vehicle. This gives rise to a conﬁguration in which the cable is straight for
a majority of its length, but turns as necessary at the vehicle end, to meet
the bottom boundary condition. In the straight part of the cable, normal
weight and drag components are equalized. The uniform angle is called the
critical angle φ
c
, and can be approximated easily. Let the relative importance
of weight be given as
δ =
W
n
ρC
n
dU
2
,
88 13 TOWING OF VEHICLES
85
13 TOWING OF VEHICLES
Vehicles which are towed have some similarities to the vehicles that have been
discussed so far. For example, towed vehicles are often streamlined, and usu
ally need good directional stability. Some towed vehicles might have active
lifting surfaces or thrusters for attitude control. On the other hand, if they
are to be supported by a cable, towed vehicles may be quite heavy in water,
and do not have to be selfpropelled. The cable itself is an important factor in
the behavior of the complete towed system, and in this section, we concentrate
on cable mechanics more than vehicle characteristics, which can generally be
handled with the same tools as other vehicles, i.e., slenderbody theory, wing
theory, linearization, etc.. Some basic guidelines for vehicle design are given
at the end of this section.
Modern cables can easily exceed 5000min length, even a heavy steel cable with
2cm diameter. The cables are generally circular in cross section, and may
carry power conductors and multiple communication channels (ﬁber optic).
The extreme L/D ratio for these cables obviates any bending stiﬀness eﬀects.
Cable systems come in a variety of conﬁgurations, and one main division may
be made simply of the density of the cable. Lighttether systems are character
ized by neutrallybuoyant (or nearly so) cables, with either a minimal vehicle
at the end, as in a towed array, or a vehicle capable of maneuvering itself, such
as a remotelyoperated vehicle. The towed array is a relatively highvelocity
system that nominally streams out horizontally behind the vessel. An ROV,
on the other hand, operates at low speed, and must have large propulsors to
control the tether if there are currents. Heavy systems, in contrast, employ a
heavy cable and possibly a heavy weight; the rationale is that gravity will tend
to keep the cable vertical and make the deployment robust against currents
and towing speed. The heavy systems will generally transmit surface motions
and tensions to the towed vehicle much more easily than lighttether systems.
We will not discuss light systems speciﬁcally here, but rather look at heavy
systems. Most of the analysis can be adapted to either case, however.
13.1 Statics
13.1.1 Force Balance
For the purposes of deriving the static conﬁguration of a cable in a ﬂow, we
assume for the moment that that it is inextensible. Tension and hydrostatic
so that the condition dφ/ds = 0 requires from the force balance
δ cos φ
c
−
1
2
sin
2
φ
c
= 0.
We are considering the case of 0 < φ
c
< π/2. Substituting sin
2
φ
c
= 1−cos
2
φ
c
,
we solve a quadratic equation and keep only the positive solution:
cos φ
c
=
√
δ
2
+ 1 −1. (157)
In the case of a very heavy cable, δ is large, and the linear approximation of
the square root
√
1 + ≈ 1 + /2 gives
cos φ
c
amp; · amp;
1
2δ
−→
φ
c
amp; · amp;
π
2
−
1
2δ
. (158)
For a very light cable, δ is small; the same approximation gives
13.2 Linearized Dynamics 89
cos φ
c
· 1 −δ −→φ
c
·
√
2δ.
The table below gives some results of the exact solution, and the approxima
tions.
δ amp; exact amp; δ >> 1 amp; δ << 1
0.1 amp; 0.44 amp;  amp; 0.45
0.2 amp; 0.61 amp;  amp; 0.63
0.5 amp; 0.91 amp; 0.57 amp; 1.00
1.0 amp; 1.14 amp; 1.07 amp; 1.41
2.0 amp; 1.33 amp; 1.32 amp; 
5.0 amp; 1.47 amp; 1.47 amp; 
13.2 Linearized Dynamics
13.2.1 Derivation
The most direct procedure for deriving useful linear dynamic equations for a
planar cable problem is to consider the total tension and angle as made up of
static parts summed with dynamic parts:
T(s, t) =
¯
T(s) +
˜
T(s, t)
φ(s, t) =
¯
φ(s) +
˜
φ(s, t).
We also write the axial deﬂection with respect to the static conﬁguration as
p(s, t), and the lateral deﬂection q(s, t). It follows that
˜
φ = ∂q/∂s. Now
augment the two static conﬁguration equations with inertial components:
m
∂
2
p
∂t
2
amp; = amp;
∂
¯
T
∂s
+
∂
˜
T
∂s
−W
n
sin(
¯
φ +
˜
φ) −
1
2
ρC
t
d
U cos φ +
∂p
∂t
2
(m + m
a
)
∂
2
q
∂t
2
amp; = amp; (
¯
T +
˜
T)
∂
¯
φ
∂s
+
∂
˜
φ
∂s
−W
n
cos(
¯
φ +
˜
φ) +
amp; amp;
1
2
ρC
n
d
U sin φ −
∂q
∂t
2
.
90 13 TOWING OF VEHICLES
Here the material mass of the cable per unit length is m, and its transverse
added mass is m
a
. Note that avoiding the drag law form v[v[ again, we have
implicitly assumed that U cos φ > [∂p/∂t[ and U sin φ > [∂q/∂t[. If it is not
the case, say U = 0, then equivalent linearization can be used for the quadratic
drag.
Now we perform the trigonometry substitutions in the weight terms, let φ ·
¯
φ
for the calculation of drag, and substitute the constitutive (Hooke’s) law
∂
˜
T
∂s
= EA
∂
2
p
∂s
2
.
The static solution cancels out of both governing equations, and keeping only
linear terms we obtain
m
∂
2
p
∂t
2
amp; = amp; EA
∂
2
p
∂s
2
−W
n
cos
¯
φ
∂q
∂s
−ρC
t
dU cos
¯
φ
∂p
∂t
(m + m
a
)
∂
2
q
∂t
2
amp; = amp;
¯
T
∂
2
q
∂s
2
+ EA
∂p
∂s
∂
¯
φ
∂s
+ W
n
sin
¯
φ
∂q
∂s
−ρC
n
dU sin
¯
φ
∂q
∂t
.
The axial dynamics (p) couples with the lateral equation through the weight
term −W
n
cos
¯
φ
˜
φ. The lateral dynamics (q) couples with the axial through the
term
˜
T∂
¯
φ/∂s. An additional weight term W
n
sin
¯
φ∂q/∂s also appears. The
uncoupled dynamics are both in the form of damped wave equations
m
∂
2
p
∂t
2
+ b
t
∂p
∂t
amp; = amp; EA
∂
2
p
∂s
2
(m + m
a
)
∂
2
q
∂t
2
+ b
n
∂q
∂t
amp; = amp;
¯
T
∂
2
q
∂s
2
+ W
n
sin
¯
φ
∂q
∂s
,
where we made the substitution b
t
= ρC
t
dU cos
¯
φ and b
n
= ρC
n
dU sin
¯
φ. To a
linear approximation, the outofplane vibrations of a cable are also governed
by the second equation above.
Because of light damping in the tangential direction, heavy cables easily trans
mit motions and tensions along their length, and can develop longitudinal
resonant conditions (next section). In contrast, the lateral cable motions are
heavily damped, such that disturbances only travel a few tens or hundreds of
meters before they dissipate. The nature of the lateral response, in and out
13.2 Linearized Dynamics 91
of the towing plane, is a very slow, damped nonlinear ﬁlter. Highfrequency
vessel motions in the horizontal plane are completely missed by the vehicle,
while lowfrequency motions occur sluggishly, and only after a signiﬁcant delay
time.
amp; axial amp; lateral
wave speed amp;
EA
m
amp;
¯
T(s)
m+ma
amp; FAST amp; SLOW
natural frequency amp;
nπ
L
EA
m
amp; O
nπ
L
¯
T(L/2)
m+ma
mode shape amp; sine/cosine amp; Bessel function
damping amp; C
t
· O(0.01) amp; C
n
· O(1)
disturbances amp; amp;
travel down amp; YES amp; NO
cable? amp; amp;
13.2.2 Damped Axial Motion
Mode Shape. The axial direction is of particular interest, since it is lightly
damped and forced by the heaving of vessels in seas. Consider a long cable
governed by the damped wave equation
m¨ p + b
t
˙ p = EAp
(159)
We use overdots to indicate time derivatives, and primes to indicate spatial
derivatives. At the surface, we impose the motion
p(L, t) = P cos ωt, (160)
while the towed vehicle, at the lower end, is an undamped mass responding to
the local tension variations:
EA
∂p(0, t)
∂s
= M¨ p(0, t). (161)
These top and bottom behaviors comprise the boundary conditions for the
wave equation. We let p(s, t) = ˜ p(s) cos ωt, so that
92 13 TOWING OF VEHICLES
˜ p
+
mω
2
−iωb
t
EA
˜ p = 0. (162)
This admits the solution ˜ p(s) = c
1
cos ks + c
2
sin ks, where
k =
mω
2
−iωb
t
EA
. (163)
Note that k is complex when b
t
= 0. The top and bottom boundary conditions
give, respectively,
P amp; = amp; c
1
cos kL + c
2
sin kL
0 amp; = amp; c
1
+ δc
2
,
where δ = EAk/ω
2
M. These can be combined to give the solution
˜ p = P
δ cos ks −sin ks
δ cos kL −sin kL
. (164)
In the case that M → 0, the scalar δ → ∞, simplifying the result to ˜ p =
P cos ks/ cos kL.
Dynamic Tension. It is possible to compute the dynamic tension via
˜
T =
EA˜ p
. We obtain
˜
T = −EAPk
δ sin ks + cos ks
δ cos kL −sin kL
. (165)
There are two dangerous situations:
• The maximum tension is
¯
T +[
˜
T[ and must be less than the working load
of the cable. This is normally problematic at the top of the cable, where
the static tension is highest.
• If [
˜
T[ >
¯
T, the cable will unload completely and then reload with ex
tremely high impulsive forces. This is known as snap loading; it occurs
primarily at the vehicle, where
¯
T is low.
Natural Frequency. The natural frequency can be found by letting b
t
= 0,
and investigating the singularity of ˜ p, for which δ cos kL = sin kL. In general,
13.2 Linearized Dynamics 93
kL << 1, but we ﬁnd that a ﬁrstorder approximation yields ω =
EA/LM,
which is only a correct answer if M >> mL, i.e., the system is dominated by
the vehicle mass. Some higher order terms need to be kept. We start with
better approximations for sin() and cos():
δ
1 −
(kL)
2
2
= kL
1 −
(kL)
2
6
.
Employing the deﬁnition for δ, and recalling that ω
2
= k
2
EA/m, we arrive at
mL
M
1 −
(kL)
2
2
= (kL)
2
1 −
(kL)
2
6
.
If we match up to second order in kL, then
ω =
EA/L
M + mL/2
.
This has the familiar form of the square root of a stiﬀness divided by a mass:
the stiﬀness of the cable is EA/L, and the mass that is oscillating is M+mL/2.
In very deep water, the eﬀects of mL/2 dominate; if ρ
c
= m/A is the density
of the cable, we have the approximation
ω ·
1
L
2E
ρ
c
.
A few examples are given below for a steel cable with E = 200 10
9
Pa, and
ρ
c
= 7000kg/m
3
. The natural frequencies near wave excitation at the surface
vessel must be taken into account in any design or deployment. Even if a
cable can withstand the eﬀects of resonance, it may be undesirable to expose
the vehicle to these motions. Some solutions in use today are: stable vessels
(e.g., SWATH), heave compensation through an active crane, a clump weight
below which a light cable is employed, and an Sshaped length of cable at the
bottom formed with ﬂotation balls.
94 13 TOWING OF VEHICLES
L = 500m amp; ω
n
= 15.0rad/s
1000m amp; 7.6rad/s
2000m amp; 3.7rad/s
5000m amp; 1.5rad/s
13.3 Cable Strumming
Cable strumming causes a host of problems, including obvious fatigue when
the amplitudes and frequencies are high. The most noteworthy issue with
towing is that the vibrations may cause the normal drag coeﬃcient C
n
to
increase dramatically – from about 1.2 for a nonoscillating cable to as high
as 3.5. This drag penalty decreases the critical angle of towing, so that larger
lengths of cable are needed to reach a given depth, and the towed system lags
further and further behind the surface vessel. The static tension will increase
accordingly as well.
Strumming of cables is caused by the proximity of a preferred vortex forma
tion frequency ω
S
to the natural frequency of the structure ω
n
. This latter
frequency can be obtained as a zero of the lightlydamped Bessel function
solution of the lateral dynamics equation above. The preferred frequency of
vortex formation is given by the empirical relation ω
S
= 2πSU/d, where S
is the Strouhal number, about 0.160.20 for a large range of Re. Strumming
of amplitude d/2 or greater can occur for 0.6 < ω
S
/ω
n
< 2.0. The book by
Blevins is a good general reference.
13.4 Vehicle Design
The physical layout of a towed vehicle is amenable to the analysis tools of
selfpropelled vehicles, with the main exceptions that the towpoint presents
a large mean force as well as some disturbances, and that the vehicle can be
quite heavy in water. Here are basic guidelines to be considered:
1. The towpoint must be located above the vehicle center of inwater weight,
for basic roll and pitch stability.
2. The towpoint should be forward of the aerodynamic center, for towing
stability reasons.
3. The combined center of mass (material and added mass) should be longi
tudinally between the towpoint and the aerodynamic center, and nearer
13.4 Vehicle Design 95
the towpoint. This will ensure that highfrequency disturbances do not
induce excessive pitching.
4. The towpoint should be longitudinally forward of the center of inair
weight, so that the vehicle enters the water ﬁns ﬁrst, and selfstabilizes
with U > 0.
5. The center of buoyancy should be behind the inwater center of weight,
so that the vehicle pitches downward at small U, and hence the net lift
force is downward, away from the surface.
Meeting all of these criteria simultaneously is no small feat, and the perfor
mance of the device is very sensitive to small perturbations in the geometry.
For this reason, fullscale experiments are commonly used in the design pro
cess.
96 14 TRANSFER FUNCTIONS & STABILITY
14 TRANSFER FUNCTIONS & STABILITY
The reader is referred to Laplace Transforms in the section MATH FACTS for
preliminary material on the Laplace transform. Partial fractions are presented
here, in the context of control systems, as the fundamental link between pole
locations and stability.
14.1 Partial Fractions
Solving linear timeinvariant systems by the Laplace Transform method will
generally create a signal containing the (factored) form
Y (s) =
K(s + z
1
)(s + z
2
) (s + z
m
)
(s + p
1
)(s + p
2
) (s + p
n
)
. (166)
Although for the moment we are discussing the signal Y (s), later we will see
that dynamic systems are described in the same format: in that case we call
the impulse response G(s) a transfer function. A system transfer function is
identical to its impulse response, since L(δ(t)) = 1.
The constants −z
i
are called the zeros of the transfer function or signal, and
−p
i
are the poles. Viewed in the complex plane, it is clear that the magnitude
of Y (s) will go to zero at the zeros, and to inﬁnity at the poles.
Partial fraction expansions alter the form of Y (s) so that the simple transform
pairs can be used to ﬁnd the timedomain output signals. We must have
m < n; if this is not the case, then we have to divide the numerator by the
denominator as necessary to ﬁnd a simple form.
14.2 Partial Fractions: Unique Poles
Under the condition m < n, it is a fact that Y (s) is equivalent to
Y (s) =
a
1
s + p
1
+
a
2
s + p
2
+
a
n
s + p
n
, (167)
in the special case that all of the poles are unique and real. The coeﬃcient
a
i
is termed the residual associated with the i’th pole, and once all these are
found it is a simple matter to go back to the transform table and look up the
timedomain responses.
14.3 Example: Partial Fractions with Unique Real Poles 97
How to ﬁnd a
i
? A simple rule applies: multiply the righthand sides of the
two equations above by (s + p
i
), evaluate them at s = −p
i
, and solve for a
i
,
the only one left.
14.3 Example: Partial Fractions with Unique Real Poles
G(s) =
s(s + 6)
(s + 4)(s −1)
e
−2s
.
Since we have a pure delay and m = n, we can initially work with G(s)/se
−2s
.
We have
s + 6
(s + 4)(s −1)
=
a
1
s + 4
+
a
2
s −1
, giving
a
1
=
(s+6)(s+4)
(s+4)(s−1)
s=−4
= −
2
5
a
2
=
(s+6)(s−1)
(s+4)(s−1)
s=1
=
7
5
Thus
L
−1
(G(s)/se
−2s
) = −
2
5
e
−4t
+
7
5
e
t
−→
g(t) = δ(t −2) +
8
5
e
−4(t−2)
+
7
5
e
t−2
.
The impulse response is needed to account for the step change at t = 2. Note
that in this example, we were able to apply the derivative operator s after
expanding the partial fractions. For cases where a second derivative must be
taken, i.e., m ≥ n + 1, special care should be used when accounting for the
signal slope discontinuity at t = 0. The more traditional method, exempliﬁed
by Ogata, may prove easier to work through.
The case of repeated real roots may be handled elegantly, but this condition
rarely occurs in applications.
98 14 TRANSFER FUNCTIONS & STABILITY
14.4 Partial Fractions: ComplexConjugate Poles
A complexconjugate pair of poles should be kept together, with the following
procedure: employ the form
Y (s) =
b
1
s + b
2
(s + p
1
)(s + p
2
)
+
a
3
s + p
3
+ , (168)
where p
1
= p
∗
2
(complex conjugate). As before, multiply through by (s +
p
1
)(s + p
2
), and then evaluate at s = −p
1
.
14.5 Example: Partial Fractions with Complex Poles
G(s) =
s + 1
s(s + j)(s −j)
=
b
1
s + b
2
(s + j)(s −j)
+
a
3
s
:
¸
s + 1
s
s=−j
= [b
1
s + b
2
]
s=−j
−→
1 + j = −b
1
j + b
2
−→
b
1
= −1
b
2
= 1; also
¸
s + 1
(s + j)(s −j)
¸
s=0
= a
3
= 1.
Working out the inverse transforms from the table of pairs, we have simply
(noting that ζ = 0)
g(t) = −cos t + sin t + 1(t).
14.6 Stability in Linear Systems
In linear systems, exponential stability occurs when all the real exponents of e
are strictly negative. The signals decay within an exponential envelope. If one
exponent is 0, the response never decays or grows in amplitude; this is called
marginal stability. If at least one real exponent is positive, then one element
of the response grows without bound, and the system is unstable.
14.7 Stability ⇐⇒ Poles in LHP 99
14.7 Stability ⇐⇒ Poles in LHP
In the context of partial fraction expansions, the relationship between stability
and pole locations is especially clear. The unit step function 1(t) has a pole at
zero, the exponential e
−at
has a pole at −a, and so on. All of the other pairs
exhibit the same property: A system is stable if and only if all of the poles
occur in the left half of the complex plane. Marginally stable parts correlate
with a zero real part, and unstable parts to a positive real part.
14.8 General Stability
There are two deﬁnitions, which apply to systems with input u(t) and output
y(t).
1. Exponential. If u(t) =
0 and y(0) = y
o
, then [y
i
(t)[ < αe
−γt
, for ﬁnite
α and γ > 0. The output asymptotically approaches zero, within a
decaying exponential envelope.
2. BoundedInput BoundedOutput (BIBO). If y(0) =
0, and [f
i
(t)[ <
γ, γ > 0 and ﬁnite, then [y
i
(t)[ < α, α > 0 and ﬁnite.
In linear timeinvariant systems, the two deﬁnitions are identical. Exponential
stability is easy to check for linear systems, but for nonlinear systems, BIBO
stability is usually easier to achieve.
100 15 CONTROL FUNDAMENTALS
15 CONTROL FUNDAMENTALS
15.1 Introduction
15.1.1 Plants, Inputs, and Outputs
Controller design is about creating dynamic systems that behave in useful
ways. Many target systems are physical; we employ controllers to steer ships,
ﬂy jets, position electric motors and hydraulic actuators, and distill alcohol.
Controllers are also applied in macroeconomics and many other important,
nonphysical systems. It is the fundamental concept of controller design that
a set of input variables acts through a given “plant” to create an output.
Feedback control then uses sensed plant outputs to apply corrective inputs:
Plant Inputs Outputs Sensors
Jet aircraft elevator, rudder, etc. altitude, hdg altimeter, GPS
Marine vessel rudder angle heading gyrocompass
Hydraulic robot valve position tip position joint angle
U.S. economy fed interest rate, etc. prosperity inﬂation, M1
Nuclear reactor cooling, neutron ﬂux power level temp., pressure
15.1.2 The Need for Modeling
Eﬀective control system design usually beneﬁts from an accurate model of
the plant, although it must be noted that many industrial controllers can be
tuned up satisfactorily with no knowledge of the plant. Ziegler and Nichols, for
example, developed a general recipe which we detail later. In any event, plant
models simply do not match realworld systems exactly; we can only hope to
capture the basic components in the form of diﬀerential or integrodiﬀerential
equations.
Beyond prediction of plant behavior based on physics, the process of system
identiﬁcation generates a plant model from data. The process is often prob
lematic, however, since the measured response could be corrupted by sensor
noise or physical disturbances in the system which cause it to behave in unpre
dictable ways. At some frequency high enough, most systems exhibit eﬀects
that are diﬃcult to model or reproduce, and this is a limit to controller per
formance.
15.2 Representing Linear Systems 101
15.1.3 Nonlinear Control
The bulk of this subject is taught using the tools of linear systems analysis.
The main reason for this restriction is that nonlinear systems are diﬃcult
to model, diﬃcult to design controllers for, and diﬃcult overall! Within the
paradigm of linear systems, there are many sets of powerful tools available.
The reader interested in nonlinear control is referred to the book by Slotine
and Li.
15.2 Representing Linear Systems
Except for the most heuristic methods of tuning up simple systems, control
system design depends on a model of the plant. The transfer function de
scription of linear systems has already been described in the discussion of the
Laplace transform. The statespace form is an entirely equivalent timedomain
representation that makes a clean extension to systems with multiple inputs
and multiple outputs, and opens the way to standard tools from linear algebra.
15.2.1 Standard StateSpace Form
We write a linear system in a statespace form as follows
˙ x = Ax + Bu + Gw (169)
y = Cx + Du + v
where
• x is a state vector, with as many elements as there are orders in the
governing diﬀerential equations.
• A is a matrix mapping x to its derivative; A captures the natural dy
namics of the system without external inputs.
• B is an input gain matrix for the control input u.
• G is a gain matrix for unknown disturbance w; w drives the state just
like the control u.
• y is the observation vector, comprised mainly of a linear combination of
states Cx (where C is a matrix.
102 15 CONTROL FUNDAMENTALS
• Du is a direct map from input to output (usually zero for physical sys
tems).
• v is an unknown sensor noise which corrupts the measurement.
u
w
B
G
A
1/s C
v
D
y
+
+
+
+
+
+
x x’
15.2.2 Converting a StateSpace Model into a Transfer Function
There are a number of canonical statespace forms available, which can create
the same transfer function. In the case of no disturbances or noise, the transfer
function (or transfer matrix) can be written as
G(s) =
y(s)
u(s)
= C(sI −A)
−1
B + D, (170)
where I is the identity matrix with the same size as A. A similar equation
holds for y(s)/w(s), and clearly y(s)/v(s) = I.
15.2.3 Converting a Transfer Function into a StateSpace Model
It may be possible to write the corresponding diﬀerential equation along one
row of the state vector, and then cascade derivatives. For example, consider
the following system:
my
(t) + by
(t) + ky(t) = u
(t) + u(t) (massspringdashpot)
G(s) =
s + 1
ms
2
+ bs + k
15.3 PID Controllers 103
Setting x = [y
, y]
T
, we obtain the system
dx
dt
=
¸
−b/m −k/m
1 0
¸
x +
¸
1/m
0
¸
u
y = [1 1] x
Note speciﬁcally that dx
2
/dt = x
1
, leading to an entry of 1 in the oﬀdiagonal
of the second row in A. Entries in the Cmatrix are easy to write in this case
because of linearity; the system response to u
is the same as the derivative of
the system response to u.
15.3 PID Controllers
The most common type of industrial controller is the proportionalintegral
derivative (PID) design. If u is the output from the controller, and e is the
error signal it receives, this control law has the form
u(t) = k
p
e(t) + k
i
t
0
e(τ)dτ + k
d
e
(t),
C(s) =
U(s)
E(s)
= k
p
+
k
i
s
+ k
d
s (171)
= k
p
¸
1 +
1
τ
i
s
+ τ
d
s
,
where the last line is written using the conventions of one overall gain k
p
, plus
a time characteristic to the integral part (τ
i
) and and time characteristic to
the derivative part (τ
d
).
In words, the proportional part of this control law will create a control action
that scales linearly with the error – we often think of this as a springlike action.
The integrator is accumulating the error signal over time, and so the control
action from this part will continue to grow as long as an error exists. Finally,
the derivative action scales with the derivative of the error. The controller will
retard motion toward zero error, which helps to reduce overshoot.
The common variations are: P, PD, PI, PID.
15.4 Example: PID Control
Consider the case of a mass (m) sliding on a frictionless table. It has a per
fect thruster that generates force u(t), but is also subject to an unknown
104 15 CONTROL FUNDAMENTALS
disturbance d(t). If the linear position of the mass is y(t), and it is perfectly
measured, we have the plant
my
(t) = u(t) + d(t).
Suppose that the desired condition is simply y(t) = 0, with initial conditions
y(0) = y
o
and y
(0) = 0.
15.4.1 Proportional Only
A proportional controller alone invokes the control law u(t) = −k
p
y(t), so that
the closedloop dynamics follow
my
(t) = −k
p
y(t) + d(t).
In the absence of d(t), we see that y(t) = y
o
cos
kp
m
t, a marginally stable
response that is undesirable.
15.4.2 ProportionalDerivative Only
Let u(t) = −k
p
y(t) −k
d
y
(t), and it follows that
my
(t) = −k
p
y(t) −k
d
y
(t) + d(t).
The system now resembles a secondorder massspringdashpot system where
k
p
plays the part of the spring, and k
d
the part of the dashpot. With an
excessively large value for k
d
, the system would be overdamped and very slow
to respond to any command. In most applications, a small amount of overshoot
is employed because the response time is shorter. The k
d
value for critical
damping in this example is 2
mk
p
, and so the rule is k
d
< 2
mk
p
. The
result, easily found using the Laplace transform, is
y(t) = y
o
e
−k
d
2m
t
¸
cos ω
d
t +
k
d
2mω
d
sin ω
d
t
¸
,
15.5 Heuristic Tuning 105
where ω
d
=
4mk
p
−k
2
d
/2m. This response is exponentially stable as desired.
Note that if the mass had a very large amount of natural damping, a negative
k
d
could be used to cancel some of its eﬀect and speed up the system response.
Now consider what happens if d(t) has a constant bias d
o
: it balances exactly
the proportional control part, eventually settling out at y(t = ∞) = d
o
/k
p
. To
achieve good rejection of d
o
with a PD controller, we would need to set k
p
very
large. However, very large values of k
p
will also drive the resonant frequency
ω
d
up, which is unacceptable.
15.4.3 ProportionalIntegralDerivative
Now let u(t) = −k
p
y(t) −k
i
t
0
y(τ)dτ −k
d
y
(t): we have
my
(t) = −k
p
y(t) −k
i
t
0
y(τ)dτ −k
d
y
(t) + d(t).
The control system has now created a thirdorder closedloop response. If
d(t) = d
o
, a time derivative leads to
my
(t) + k
p
y
(t) + k
i
y(t) + k
d
y
(t) = 0,
so that y(t = ∞) = 0, as desired, provided the roots are stable.
15.5 Heuristic Tuning
For many practical systems, tuning of a PID controller may proceed without
any system model. This is especially pertinent for plants which are openloop
stable, and can be safely tested with varying controllers. One useful approach
is due to Ziegler and Nichols (1942), which transforms the basic characteristics
of a step response (e.g., the input is 1(t)) into a reasonable PID design. The
idea is to approximate the response curve by a ﬁrstorder lag (gain k and time
constant τ) and a pure delay T:
G(s) ·
ke
−Ts
τs + 1
(172)
The following rules apply only if the plant contains no dominating, lightly
damped complex poles, and has no poles at the origin:
106 15 CONTROL FUNDAMENTALS
P k
p
= 1.0τ/T
PI k
p
= 0.9τ/T k
i
= 0.27τ/T
2
PID k
p
= 1.2τ/T k
i
= 0.60τ/T
2
k
d
= 0.60τ
Note that if no pure time delay exists (T = 0), this recipe suggests the pro
portional gain can become arbitrarily high! Any characteristic other than a
true ﬁrstorder lag would therefore be expected to cause a measurable delay.
15.6 Block Diagrams of Systems
15.6.1 Fundamental Feedback Loop
The topology of a feedback system can be represented graphically by consid
ering each dynamical system element to reside within a box, having an input
line and an output line. For example, the plant used above (a simple mass)
has transfer function P(s) = 1/ms
2
, which relates the input, force u(s), into
the output, position y(s). In turn, the PDcontroller has transfer function
C(s) = k
p
+ k
d
s; its input is the error signal E(s) = −y(s), and its output is
force u(s). The feedback loop in block diagram form is shown below.
u(s) y(s)
C(s) P(s)
15.6.2 Block Diagrams: General Case
The simple feedback system above is augmented in practice by three external
inputs. The ﬁrst is a process disturbance, which can be taken to act at the
input of the physical plant, or at the output. In the former case, it is additive
with the control action, and so has some physical meaning. In the second case,
the disturbance has the same units as the plant output.
Another external input is the reference command or setpoint, used to create a
more general error signal e(s) = r(s) − y(s). Note that the feedback loop, in
trying to force e(s) to zero, will necessarily make y(s) approximate r(s).
The ﬁnal input is sensor noise, which usually corrupts the feedback signal y(s),
causing some error in the evaluation of e(s), and so on. Sensors with very poor
15.6 Block Diagrams of Systems 107
noise properties can ruin the performance of a control system, no matter how
perfectly understood are the other components.
y
r
e
u
C
u
d
y
d
n
P
+
+
+
+
+
+

+
15.6.3 Primary Transfer Functions
Some algebra shows that
e
r
=
1
1 + PC
= S
y
r
=
PC
1 + PC
= T
u
r
=
C
1 + CP
= U.
e/r = S relates the reference input and noise to the error, and is known as the
sensitivity function. We would generally like S to be small at low frequencies,
so that the tracking error there is small. y/r = T is called the complementary
sensitivity function. Note that S + T = 1, implying that these two functions
must always trade oﬀ; they cannot both be small or large at the same time.
Other systems we encounter again later are the (forward) loop transfer function
PC, the loop transfer function broken between C and P: CP, and
e
d
u
=
−P
1 + PC
y
d
u
=
P
1 + PC
u
d
u
=
−CP
1 + CP
e
d
y
=
−1
1 + PC
= −S
108 15 CONTROL FUNDAMENTALS
y
d
y
=
1
1 + PC
= S
u
d
y
=
−C
1 + CP
= −U
e
n
=
−1
1 + PC
= −S
y
n
=
−PC
1 + PC
= −T
u
n
=
−C
1 + CP
= −U.
If the disturbance is taken at the plant output, then the three functions S, T,
and U (control action) completely describe the system. This will in fact be
the procedure when we address loopshaping.
109
16 MODAL ANALYSIS
16.1 Introduction
The evolution of states in a linear system occurs through independent modes,
which can be driven by external inputs, and observed through plant output.
This section provides the basis for modal analysis of systems. Throughout, we
use the statespace description of a system with D = 0:
˙
x = Ax + Bu
y = Cx.
16.2 Matrix Exponential
16.2.1 Deﬁnition
In the instance of an unforced response to initial conditions, consider the
system
˙
x = Ax, x(t = 0) = χ.
In the scalar case, the response is x(t) = χe
at
, giving a decaying exponential
if a < 0. The same notation holds for the case of a vector x, and matrix A:
x(t) = e
At
χ, where
e
At
= I + At +
(At)
2
2!
+
e
At
is usually called the matrix exponential.
16.2.2 Modal Canonical Form
Introductory material on the eigenvalue problem and modal decomposition
can be found in the MATH FACTS section. This modal decomposition of A
leads to a very useful statespace representation. Namely, since A = V ΛV
−1
,
a transformation of state variables can be made, x = V z, leading to
110 16 MODAL ANALYSIS
˙
z = Λz + V
−1
Bu (173)
y = CV z.
This is called the modal canonical form, since the states are simply the modal
amplitudes. These states are uncoupled in Λ, but may be coupled through the
input (V
−1
B) and output (CV ) mappings. The modal form is numerically
robust for computations.
16.2.3 Modal Decomposition of Response
Now we are ready to look at the matrix exponential e
At
in terms of its con
stituent modes. Employing the above form for A, we ﬁnd that
e
At
= I + At +
(At)
2
2!
+
= V
I + Λt +
(Λt)
2
2!
+
W
T
= V e
Λt
W
T
=
n
¸
i=1
e
λ
i
t
v
i
w
T
i
.
In terms of the response to an initial condition χ, we have
x(t) =
n
¸
i=1
e
λ
i
t
v
i
( w
T
i
χ).
The product w
T
i
χ is a scalar, the projection of the initial conditions onto the
i’th mode. If χ is perpendicular to w
i
, then the product is zero and the i’th
mode does not respond. Otherwise, the i’th mode does participate in the
response. The projection of the i’th mode onto the states x is through the
right eigenvector v
i
.
For stability of the system, the eigenvalues of A, that is, λ
i
, must have neg
ative real parts; they are in fact the poles of the equivalent transfer function
description.
16.3 Forced Response and Controllability 111
16.3 Forced Response and Controllability
Now consider the system with an external input u:
˙
x = Ax + Bu, x(t = 0) = χ.
Taking the Laplace transform of the system, taking into account the initial
condition for the derivative, we have
sx(s) − χ = Ax(s) + Bu(s) −→
x(s) = (sI −A)
−1
χ + (sI −A)
−1
Bu(s).
Thus (sI − A)
−1
can be recognized as the Laplace transform of the matrix
exponential e
At
. In the time domain, the second term then has the form of a
convolution of the matrix exponential and the net input Bu:
x(t) =
t
0
e
A(t−τ)
Bu(τ)dτ
=
n
¸
i=1
t
0
e
λ
i
(t−τ)
v
i
w
T
i
Bu(τ)dτ.
Suppose now that there are m inputs, such that B = [
b
1
,
b
2
, ,
b
m
]. Then
some rearrangement will give
x(t) =
n
¸
i=1
v
i
m
¸
k=1
( w
T
i
b
k
)
t
0
e
λ
i
(t−τ)
u
k
(τ)dτ.
The product w
T
i
b
k
, a scalar, represents the projection of the k’th control chan
nel onto the i’th mode. We say that the i’th mode is controllable from the
k’th input if the product is nonzero. If a given mode i has w
T
i
b
k
= 0 for all
input channels k, then the mode is uncontrollable.
In normal applications, controllability for the entire system is checked using
the following test: Construct the socalled controllability matrix:
M
c
= [B, AB, , A
n−1
B]. (174)
This matrix has size n (nm), where m is the number of input channels. If
M
c
has rank n, then the system is controllable, i.e., all modes are controllable.
112 16 MODAL ANALYSIS
16.4 Plant Output and Observability
We now turn to a related question: can the complete state vector of the system
be observed given only the output measurements y, and the known control u?
The response due to the external input is easy to compute deterministically,
through the convolution integral. Consider the part due to initial conditions
χ. We found above
x(t) =
n
¸
i=1
e
λ
i
t
v
i
w
T
i
χ.
The observation is y = Cx (r channels of output), and writing
C =
c
T
1
c
T
r
¸
¸
¸ .
the k’th channel of the output is
y
k
(t) =
n
¸
i=1
(c
T
k
v
i
)e
λ
i
t
( w
T
i
χ).
The i’th mode is observable in the k’th output if the product c
T
k
v
i
= 0. We
say that a system is observable if every mode can be seen in at least one output
channel. The usual test for system observability requires computation of the
observability matrix:
M
o
=
C
T
, A
T
C
T
, , (A
T
)
n−1
C
T
. (175)
This matrix has size n (rn); the system is observable if M
o
has rank n.
113
17 CONTROL SYSTEMS – LOOPSHAPING
17.1 Introduction
This section formalizes the notion of loopshaping for linear control system
design. The loopshaping approach is inherently twofold. First, we shape the
openloop transfer function (or matrix) P(s)C(s), to meet performance and
robustness speciﬁcations. Once this is done, then the compensator must be
computed, from from knowing the nominal product P(s)C(s), and the nominal
plant P(s).
Most of the analysis here is given for singleinput, singleoutput systems, but
the link to multivariable control is not too diﬃcult. In particular, absolute
values of transfer functions are replaced with the maximum singular values of
transfer matrices. Design based on singular values is the idea of L
2
control, or
LQG/LTR, to be presented in the next lectures.
17.2 Roots of Stability – Nyquist Criterion
We consider the SISO feedback system with reference trajectory r(s) and plant
output y(s), as given previously. The tracking error signal is deﬁned as e(s) =
r(s) −y(s), thus forming the negative feedback loop. The sensitivity function
is written as
S(s) =
e(s)
r(s)
=
1
1 + P(s)C(s)
,
where P(s) represents the plant transfer function, and C(s) the compensator.
The closedloop characteristic equation, whose roots are the poles of the closed
loop system, is 1 + P(s)C(s) = 0, equivalent to P(s)C(s) + P(s)C(s) =
0, where the underline and overline denote the denominator and numerator,
respectively. The Nyquist criterion allows us to assess the stability properties
of a system based on P(s)C(s) only. This method for design involves plotting
the complex loci of P(s)C(s) for the range ω = [−∞, ∞]. There is no explicit
calculation of the closedloop poles, and in this sense the design approach is
quite diﬀerent from the rootlocus method (see Ogata).
114 17 CONTROL SYSTEMS – LOOPSHAPING
17.2.1 Mapping Theorem
We impose a reasonable assumption from the outset: The number of poles in
P(s)C(s) exceeds the number of zeros. It is a reasonable constraint because
otherwise the loop transfer function could pass signals with inﬁnitely high
frequency. In the case of a PID controller (two zeros) and a secondorder zero
less plant, this constraint can be easily met by adding a highfrequency rolloﬀ
to the compensator, the equivalent of lowpass ﬁltering the error signal.
Let F(s) = 1 + P(s)C(s). The heart of the Nyquist analysis is the mapping
theorem, which answers the following question: How do paths in the splane
map into paths in the Fplane? We limit ourselves to closed, clockwise(CW)
paths in the splane, and the remarkable result of the mapping theorem is
Every zero of F(s) enclosed in the splane generates exactly one CW encir
clement of the origin in the F(s)plane. Conversely, every pole of F(s) enclosed
in the splane generates exactly one CCW encirclement of the origin in the
F(s)plane. Since CW and CCW encirclements of the origin may cancel, the
relation is often written Z −P = CW.
The trick now is to make the trajectory in the splane enclose all unstable poles,
i.e., the path encloses the entire righthalf plane, moving up the imaginary
axis, and then proceeding to the right at an arbitrarily large radius, back to
the negative imaginary axis.
Since the zeros of F(s) are in fact the poles of the closedloop transfer function,
e.g., S(s), stability requires that there are no zeros of F(s) in the righthalf
splane. This leads to a slightly shorter form of the above relation:
P = CCW. (176)
In words, stability requires that the number of unstable poles in F(s) is equal
to the number of CCW encirclements of the origin, as s sweeps around the
entire righthalf splane.
17.2.2 Nyquist Criterion
The Nyquist criterion now follows from one translation. Namely, encirclements
of the origin by F(s) are equivalent to encirclements of the point (−1 + 0j)
by F(s) −1, or P(s)C(s). Then the stability criterion can be cast in terms of
the unstable poles of P(s)C(s), instead of those of F(s):
P = CCW ←→ closedloop stability (177)
17.2 Roots of Stability – Nyquist Criterion 115
This is in fact the complete Nyquist criterion for stability. It is a necessary and
suﬃcient condition that the number of unstable poles in the loop transfer func
tion P(s)C(s) must be matched by an equal number of CCW encirclements
of the critical point (−1 + 0j).
There are several details to keep in mind when making Nyquist plots:
• If neither the plant nor the controller have unstable modes, then the loci
of P(s)C(s) must not encircle the critical point at all.
• Because the path taken in the splane includes negative frequencies (i.e.,
the negative imaginary axis), the loci of P(s)C(s) occur as complex
conjugates – the plot is symmetric about the real axis.
• The requirement that the number of poles in P(s)C(s) exceeds the num
ber of zeros means that at high frequencies, P(s)C(s) always decays such
that the loci go to the origin.
• For the multivariable (MIMO) case, the procedure of looking at individ
ual Nyquist plots for each element of a transfer matrix is unreliable and
outdated. Referring to the multivariable deﬁnition of S(s), we should
count the encirclements for the function [det(I +P(s)C(s)) −1] instead
of P(s)C(s). The use of gain and phase margin in design is similar to
the SISO case.
17.2.3 Robustness on the Nyquist Plot
The question of robustness in the presence of modelling errors is central to
control system design. There are two natural measures of robustness for the
Nyquist plot, each having a very clear graphical representation. The loci need
to stay away from the critical point; how close the loci come to it can be
expressed in terms of magnitude and angle.
• When the angle of P(s)C(s) is −180
◦
, the magnitude [P(s)C(s)[ should
not be near one.
• When the magnitude [P(s)C(s)[ = 1, its angle should not be −180
◦
.
These notions lead to deﬁnition of the gain margin k
g
and phase margin γ for
a design. As the ﬁgure shows, the deﬁnition of k
g
is diﬀerent for stable and
unstable P(s)C(s). Rules of thumb are as follows. For a stable plant, k
g
≥ 2
and γ ≥ 30
◦
; for an unstable plant, k
g
≤ 0.5 and γ ≥ 30
◦
.
116 17 CONTROL SYSTEMS – LOOPSHAPING
g
Re(s)
1/k
Re(s)
Im(s) Im(s) P(s)C(s)
P(s)C(s)
γ
γ
g
1/k
Stable P(s)C(s) Unstable P(s)C(s)
17.3 Design for Nominal Performance
Performance requirements of a feedback controller, using the nominal plant
model, can be cast in terms of the Nyquist plot. We restrict the discussion to
the scalar case in the following sections.
Since the sensitivity function maps reference input r(s) to tracking error e(s),
we know that [S(s)[ should be small at low frequencies. For example, if one
percent tracking is to be maintained for all frequencies below ω = λ, then
[S(s)[ < 0.01, ∀ω < λ. This can be formalized by writing
[W
1
(s)S(s)[ < 1, (178)
where W
1
(s) is a stable weighting function of frequency. To force S(s) to be
small at low ω, W
1
(s) should be large in the same range. The requirement
[W
1
(s)S(s)[ < 1 is equivalent to [W
1
(s)[ < [1 + P(s)C(s)[, and this latter
condition can be interpreted as: The loci of P(s)C(s) must stay outside the
disk of radius W
1
(s), which is to be centered on the critical point (−1 + 0j).
The disk is to be quite large, possibly inﬁnitely large, at the lower frequencies.
17.4 Design for Robustness
It is a ubiquitous observation that models of plants degrade with increasing
frequency. For example, the DC gain and slow, lightlydamped modes or zeros
are easy to observe, but higherfrequency components in the response may
17.4 Design for Robustness 117
be hard to capture or even excite repeatably. Higherfrequency behavior may
have more nonlinear properties as well.
The eﬀects of modeling uncertainty can be considered to enter the nominal
feedback system as a disturbance at the plant output, d
y
. One of the most
useful descriptions of model uncertainty is the multiplicative uncertainty:
˜
P(s) = (1 + ∆(s)W
2
(s))P(s). (179)
Here, P(s) represents the nominal plant model used in the design of the control
loop, and
˜
P(s) is the actual, perturbed plant. The perturbation is of the
multiplicative type, ∆(s)W
2
(s)P(s), where ∆(s) is an unknown but stable
function of frequency for which [∆(s)[ ≤ 1. The weighting function W
2
(s)
scales ∆(s) with frequency; W
2
(s) should be growing with increasing frequency,
since the uncertainty grows. However, W
2
(s) should not grow any faster than
necessary, since it will turn out to be at the cost of nominal performance.
In the scalar case, the weight can be estimated as follows: since
˜
P/P − 1 =
∆W
2
, it will suﬃce to let [
˜
P/P −1[ < [W
2
[.
Example: Let
˜
P = k/(s −1), where k is in the range 2–5. We need to create
a nominal model P = k
0
/(s−1), with the smallest possible value of W
2
, which
will not vary with frequency in this case. Two equations can be written using
the above estimate, for the two extreme values of k, yielding k
0
= 7/2, and
W
2
= 3/7.
For constructing the Nyquist plot, we observe that
˜
P(s)C(s) = (1+∆(s)W
2
(s))P(s)C(s). The path of the perturbed plant could
be anywhere on a disk of radius [W
2
(s)P(s)C(s)[, centered on the nominal loci
P(s)C(s). The robustness condition is that this disk should not intersect the
critical point. This can be written as
[1 + PC[ > [W
2
PC[ ←→
1 >
[W
2
PC[
[1 + PC[
←→
1 > [W
2
T[, (180)
where T is the complementary sensitivity function. The last inequality is thus
a condition for robust stability in the presence of multiplicative uncertainty
parametrized with W
2
.
118 17 CONTROL SYSTEMS – LOOPSHAPING
17.5 Robust Performance
The condition for good performance with plant uncertainty is a combination
of the above two conditions. Graphically, the disk at the critical point, with
radius [W
1
[, should not intersect the disk of radius [W
2
PC[, centered on the
nominal locus PC. This is met if
[W
1
S[ +[W
2
T[ < 1. (181)
The robust performance requirement is related to the magnitude [PC[ at dif
ferent frequencies, as follows:
1. At low frequency, [W
1
S[ · [W
1
/PC[, since [PC[ is large. This leads
directly to the performance condition [PC[ > [W
1
[ in this range.
2. At high frequency, W
2
T[ · [W
2
PC[, since [PC[ is small. We must
therefore have [PC[ < 1/[W
2
[, for robustness.
17.6 Implications of Bode’s Integral
The loop transfer function PC cannot roll oﬀ too rapidly in the crossover
region. The simple reason is that a steep slope induces a large phase loss,
which in turn degrades the phase margin. To see this requires a short foray
into Bode’s integral. For a transfer function H(s), the crucial relation is
angle(H(jω
0
)) =
1
π
∞
−∞
d
dν
(ln[H(jω)) ln(coth([ν[/2))dν, (182)
where ν = ln(ω/ω
0
). The integral is hence taken over the log of a frequency
normalized with ω
0
. It is not hard to see how the integral controls the angle:
the function ln(coth([ν[/2)) is nonzero only near ν = 0, implying that the
angle depends only on the local slope d(ln[H[)/dν. Thus, if the slope is large,
the angle is large.
Example: Suppose H(s) = ω
n
0
/s
n
, i.e., it is a simple function with n poles at
the origin, and no zeros; ω
0
is a ﬁxed constant. It follows that [H[ = ω
n
0
/ω
n
,
and ln[H[ = −nln(ω/ω
0
), so that d(ln[H[)/dν = −n. Then we have just
angle(H) = −
n
π
∞
−∞
ln(coth([ν[/2))dν = −
nπ
2
. (183)
17.7 The Recipe for Loopshaping 119
This integral is trivial to look up or compute. Each pole at the origin clearly
induces 90
◦
of phase loss. In the general case, each pole not at the origin
induces 90
◦
of phase loss for frequencies above the pole. Each zero at the
origin adds 90
◦
phase lead, while zeros not at the origin at 90
◦
of phase lead
for frequencies above the zero. In the immediate neighborhood of these poles
and zeros, the phase may vary signiﬁcantly with frequency.
The Nyquist loci are clearly susceptible to these variations is phase, and
the phase margin can be easily lost if the slope of PC at crossover (where
the magnitude is unity) is too steep. The slope can safely be ﬁrstorder
(−20dB/decade, equivalent to a single pole), and may be secondorder
(−40dB/decade) if an adequate phase angle can be maintained near crossover.
17.7 The Recipe for Loopshaping
In the above analysis, we have extensively described what the open loop trans
fer function PC should look like, to meet robustness and performance speci
ﬁcations. We have said very little about how to get the compensator C, the
critical component. For clarity, let the designed loop transfer function be re
named, L = PC. We will use concepts from optimal linear control for the
MIMO case, but in the scalar case, it suﬃces to just pick
C = L/P. (184)
This extraordinarily simple step involves a plant inversion.
The overall idea is to ﬁrst shape L as a stable transfer function meeting the
requirements of stability and robustness, and then divide through by the plant.
• When the plant is stable and has stable zeros (minimumphase), the
division can be made directly.
• One caveat for the wellbehaved plant is that lightlydamped poles or
zeros should not be cancelled verbatim by the compensator, because
the closedloop response will be sensitive to any slight change in the
resonant frequency. The usual procedure is to widen the notch or pole
in the compensator, through a higher damping ratio.
• Nonminimum phase or unstable behavior in the plant can usually be
handled by performing the loopshaping for the closest stable model, and
then explicitly considering the eﬀects of adding the unstable parts. In
120 17 CONTROL SYSTEMS – LOOPSHAPING
the case of unstable zeros, we ﬁnd that they impose an unavoidable
frequency limit for the crossover. In general, the troublesome zeros must
be faster than the closedloop frequency response.
In the case of unstable poles, the converse is true: The feedback system
must be faster than the corresponding frequency of the unstable mode.
When a control system involves multiple inputs and outputs, the ideas from
scalar loopshaping can be adapted using the singular value. We list below
some basic properties of the singular value decomposition, which is analogous
to an eigenvector, or modal, analysis. Useful properties and relations for the
singular value are found in the section MATH FACTS.
The condition for MIMO robust performance can be written in many ways,
including a direct extension of our scalar condition
σ(W
1
S) + σ(W
2
T) < 1. (185)
The openloop transfer matrix L should be shaped accordingly. In the follow
ing sections, we use the properties of optimal state estimation and control to
perform the plant inversion for MIMO systems.
121
18 LINEAR QUADRATIC REGULATOR
18.1 Introduction
The simple form of loopshaping in scalar systems does not extend directly
to multivariable (MIMO) plants, which are characterized by transfer matrices
instead of transfer functions.
The notion of optimality is closely tied to MIMO control system design. Op
timal controllers, i.e., controllers that are the best possible, according to some
ﬁgure of merit, turn out to generate only stabilizing controllers for MIMO
plants. In this sense, optimal control solutions provide an automated design
procedure – we have only to decide what ﬁgure of merit to use. The lin
ear quadratic regulator (LQR) is a wellknown design technique that provides
practical feedback gains.
18.2 FullState Feedback
For the derivation of the linear quadratic regulator, we assume the plant to be
written in statespace form ˙ x = Ax + Bu, and that all of the n states x are
available for the controller. The feedback gain is a matrix K, implemented as
u = −K(x −x
desired
). The system dynamics are then written as:
˙ x = (A −BK)x + Kx
desired
. (186)
x
desired
represents the vector of desired states, and serves as the external input
to the closedloop system. The “Amatrix” of the closed loop system is (A −
BK), and the “Bmatrix” of the closedloop system is K. The closedloop
system has exactly as many outputs as inputs: n. The column dimension
of B equals the number of channels available in u, and must match the row
dimension of K. Poleplacement is the process of placing the poles of (A−BK)
in stable, suitablydamped locations in the complex plane.
18.3 Dynamic Programming
There are at least two conventional derivations for the LQR; we present here
one based on dynamic programming, due to R. Bellman. The key observation
is best given through a loose example:
Suppose that we are driving from Point A to Point C, and we ask what is
the shortest path in miles. If A and C represent Los Angeles and Boston,
122 18 LINEAR QUADRATIC REGULATOR
A
B
B
1
B
2
3
1
C
C
2
3
C
D
n = 3
s = 2
for example, there are many paths to choose from! Assume that one way or
another we have found the best path, and that a Point B lies along this path,
say Las Vegas. Let X be an arbitrary point east of Las Vegas. If we were to
now solve the optimization problem for getting from only Las Vegas to Boston,
this same arbitrary point X would be along the new optimal path as well.
The point is a subtle one: the optimization problem from Las Vegas to Boston
is easier than that from Los Angeles to Boston, and the idea is to use this
property backwards through time to evolve the optimal path, beginning in
Boston.
Example: Nodal Travel. We now add some structure to the above experi
ment. Consider now traveling from point A (Los Angeles) to Point D (Boston).
Suppose there are only three places to cross the Rocky Mountains, B
1
, B
2
, B
3
,
and three places to cross the Mississippi River, C
1
, C
2
, C
3
.
3
By way of no
tation, we say that the path from A to B
1
is AB
1
. Suppose that all of the
paths (and distances) from A to the Bnodes are known, as are those from the
Bnodes to the Cnodes, and the Cnodes to the terminal point D. There are
nine unique paths from A to D.
A bruteforce approach sums up the total distance for all the possible paths,
and picks the shortest one. In terms of computations, we could summarize that
this method requires nine additions of three numbers, equivalent to eighteen
additions of two numbers. The comparison of numbers is relatively cheap.
The dynamic programming approach has two steps. First, from each Bnode,
pick the best path to D. There are three possible paths from B
1
to D, for
example, and nine paths total from the Blevel to D. Store the best paths
as B
1
D[
opt
, B
2
D[
opt
, B
3
D[
opt
. This operation involves nine additions of two
3
Apologies to readers not familiar with American geography.
18.4 Dynamic Programming and FullState Feedback 123
numbers.
Second, compute the distance for each of the possible paths from A to D,
constrained to the optimal paths from the Bnodes onward: AB
1
+ B
1
D[
opt
,
AB
2
+ B
2
D[
opt
, or AB
3
+ B
3
D[
opt
. The combined path with the shortest
distance is the total solution; this second step involves three sums of two
numbers, and total optimization is done in twelve additions of two numbers.
Needless to say, this example gives only a mild advantage to the dynamic
programming approach over brute force. The gap widens vastly, however, as
one increases the dimensions of the solution space. In general, if there are s
layers of nodes (e.g., rivers or mountain ranges), and each has width n (e.g., n
river crossing points), the brute force approach will take (sn
s
) additions, while
the dynamic programming procedure involves only (n
2
(s −1) + n) additions.
In the case of n = 5, s = 5, brute force requires 6250 additions; dynamic
programming needs only 105.
18.4 Dynamic Programming and FullState Feedback
We consider here the regulation problem, that is, of keeping x
desired
= 0.
The closedloop system thus is intended to reject disturbances and recover
from initial conditions, but not necessarily follow ytrajectories. There are
several necessary deﬁnitions. First we deﬁne an instantaneous penalty function
l(x(t), u(t)), which is to be greater than zero for all nonzero x and u. The cost
associated with this penalty, along an optimal trajectory, is
J =
∞
0
l(x(t), u(t))dt, (187)
i.e., the integral over time of the instantaneous penalty. Finally, the optimal
return is the cost of the optimal trajectory remaining after time t:
V (x(t), u(t)) =
∞
t
l(x(τ), u(τ))dτ. (188)
.
We have directly from the dynamic programming principle
V (x(t), u(t)) = min
u
¦l(x(t), u(t))δt + V (x(t + δt), u(t + δt))¦ . (189)
The minimization of V (x(t), u(t)) is made by considering all the possible con
trol inputs u in the time interval (t, t + δt). As suggested by dynamic pro
gramming, the return at time t is constructed from the return at t + δt, and
124 18 LINEAR QUADRATIC REGULATOR
the diﬀerential component due to l(x, u). If V is smooth and has no explicit
dependence on t, as written, then
V (x(t + δt), u(t + δt)) = V (x(t), u(t)) +
∂V
∂x
dx
dt
δt + h.o.t. −→ (190)
= V (x(t), u(t)) +
∂V
∂x
(Ax(t) + Bu(t))δt.
Now control input u in the interval (t, t + δt) cannot aﬀect V (x(t), u(t)), so
inserting the above and making a cancellation gives
0 = min
u
l(x(t), u(t)) +
∂V
∂x
(Ax(t) + Bu(t))
¸
. (191)
We next make the assumption that V (x, u) has the following form:
V (x, u) =
1
2
x
T
Px, (192)
where P is a symmetric matrix, and positive deﬁnite.
45
It follows that
∂V
∂x
= x
T
P −→ (193)
0 = min
u
l(x, u) + x
T
P(Ax + Bu)
¸
.
We ﬁnally specify the instantaneous penalty function. The LQR employs the
special quadratic form
l(x, u) =
1
2
x
T
Qx +
1
2
u
T
Ru, (194)
where Q and R are both symmetric and positive deﬁnite. The matrices Q
and R are to be set by the user, and represent the main “tuning knobs” for
the LQR. Substitution of this form into the above equation, and setting the
derivative with respect to u to zero gives
4
Positive deﬁniteness means that x
T
Px > 0 for all nonzero x, and x
T
Px = 0 if x = 0.
5
This suggested form for the optimal return can be conﬁrmed after the optimal controller
is derived.
18.5 Properties and Use of the LQR 125
0 = u
T
R + x
T
PB (195)
u
T
= −x
T
PBR
−1
u = −R
−1
B
T
Px.
The gain matrix for the feedback control is thus K = R
−1
B
T
P. Inserting
this solution back into equation 194, and eliminating u in favor of x, we have
0 =
1
2
x
T
Qx −
1
2
x
T
PBR
−1
B
T
P + x
T
PAx.
All the matrices here are symmetric except for PA; since x
T
PAx = x
T
A
T
Px,
we can make its eﬀect symmetric by letting
x
T
PAx =
1
2
x
T
PAx +
1
2
x
T
A
T
Px,
leading to the ﬁnal matrixonly result
0 = Q + PA + A
T
P −PBR
−1
B
T
P. (196)
This equation is the matrix algebraic Riccati equation (MARE), whose solution
P is needed to compute the optimal feedback gain K. The MARE is easily
solved by standard numerical tools in linear algebra.
18.5 Properties and Use of the LQR
Static Gain. The LQR generates a static gain matrix K, which is not a
dynamical system. Hence, the order of the closedloop system is the same as
that of the plant.
Robustness. The LQR achieves inﬁnite gain margin: k
g
= ∞, implying that
the loci of (PC) (scalar case) or (det(I + PC) − 1) (MIMO case) approach
the origin along the imaginary axis. The LQR also guarantees phase margin
γ ≥ 60 degrees. This is in good agreement with the practical guidelines for
control system design.
126 18 LINEAR QUADRATIC REGULATOR
Output Variables. In many cases, it is not the states x which are to be
minimized, but the output variables y. In this case, we set the weighting
matrix Q = C
T
Q
C, since y = Cx, and the auxiliary matrix Q
weights the
plant output.
Behavior of ClosedLoop Poles: Expensive Control. When R >>
C
T
Q
C, the cost function is dominated by the control eﬀort u, and so the
controller minimizes the control action itself. In the case of a completely stable
plant, the gain will indeed go to zero, so that the closedloop poles approach
the openloop plant poles in a manner consistent with the scalar root locus.
The optimal control must always stabilize the closedloop system, however,
so there should be some account made for unstable plant poles. The expen
sive control solution puts stable closedloop poles at the mirror images of the
unstable plant poles.
Behavior of ClosedLoop Poles: Cheap Control. When R << C
T
Q
C,
the cost function is dominated by the output errors y, and there is no penalty
for using large u. There are two groups of closedloop poles. First, poles
are placed at stable plant zeros, and at the mirror images of the unstable
plant zeros. This part is akin to the highgain limiting case of the root locus.
The remaining poles assume a Butterworth pattern, whose radius increases to
inﬁnity as R becomes smaller and smaller.
The Butterworth pattern refers to an arc in the stable lefthalf plane, as shown
in the ﬁgure. The angular separation of n closedloop poles on the arc is
constant, and equal to 180
◦
/n. An angle 90
◦
/n separates the most lightly
damped poles from the imaginary axis.
18.5 Properties and Use of the LQR 127
n = 4
φ = 45 degrees
X
X
X
X
φ
φ/2
φ
φ
φ/2
128 19 KALMAN FILTER
19 KALMAN FILTER
19.1 Introduction
In the previous section, we derived the linear quadratic regulator as an optimal
solution for the fullstate feedback control problem. The inherent assumption
was that each state was known perfectly. In real applications, the measure
ments are subject to disturbances, and may not allow reconstruction of all the
states. This state estimation is the task of a modelbased estimator having
the form:
˙
ˆ x = Aˆ x + Bu + H(y −Cˆ x) (197)
The vector ˆ x represents the state estimate, whose evolution is governed by
the nominal A and B matrices of the plant, and a correction term with the
estimator gain matrix H. H operates on the estimation error mapped to
the plant output y, since Cˆ x = ˆ y. Given statistical properties of real plant
disturbances and sensor noise, the Kalman Filter designs an optimal H.
+

y
x
+
+
+
1/s
A
H
C
x
y
Bu
19.2 Problem Statement
We consider the statespace plant model given by:
˙ x = Ax + Bu + W
1
(198)
y = Cx + W
2
.
There are n states, m inputs, and l outputs, so that A has dimension nn, B
is n m, and C is l n. The plant subject to two random input signals, W
1
19.3 Step 1: An Equation for
˙
Σ 129
and W
2
. W
1
represents disturbances to the plant, since it drives ˙ x directly;
W
2
denotes sensor noise, which corrupts the measurement y.
An important assumption of the Kalman Filter is that W
1
and W
2
are each
vectors of unbiased, independent white noise, and that all the n + l channels
are uncorrelated. Hence, if E() denotes the expected value,
E(W
1
(t
1
)W
1
(t
2
)
T
) = V
1
δ(t
1
−t
2
) (199)
E(W
2
(t
1
)W
2
(t
2
)
T
) = V
2
δ(t
1
−t
2
) (200)
E(W
1
(t)W
2
(t)
T
) = 0
n×l
. (201)
Here δ(t) represents the impulse (or delta) function. V
1
is an n n diagonal
matrix of intensities, and V
2
is an l l diagonal matrix of intensities.
The estimation error may be deﬁned as e = x−ˆ x. It can then be veriﬁed that
˙ e = [Ax + Bu + W
1
] −[Aˆ x + Bu + H(y −Cˆ x)] (202)
= (A −HC)e + (W
1
−HW
2
).
The eigenvalues of the matrix A−HC thus determine the stability properties
of the estimation error dynamics. The second term above, W
1
+ HW
2
is
considered an external input.
The Kalman ﬁlter design provides H that minimizes the scalar cost function
J = E(e
T
We), (203)
where W is an unspeciﬁed symmetric, positive deﬁnite weighting matrix. A
related matrix, the symmetric error covariance, is deﬁned as
Σ = E(ee
T
). (204)
There are two main steps for deriving the optimal gain H.
19.3 Step 1: An Equation for
˙
Σ
The evolution of Σ follows from some algebra and the convolution form of e(t).
We begin with
130 19 KALMAN FILTER
˙
Σ = E( ˙ ee
T
+ e ˙ e
T
) (205)
= E[(A −HC)ee
T
+ (W
1
−HW
2
)e
T
+ ee
T
(A
T
−C
T
H
T
) +
e(W
T
1
−W
T
2
H
T
)].
The last term above can be expanded, using the property that
e(t) = e
(A−HC)t
e(0) +
t
0
e
(A−HC)(t−τ)
(W
1
(τ) −HW
2
(τ)) dτ.
We have
E(e(W
T
1
−W
T
2
H
T
)) = e
(A−HC)t
E(e(0)(W
T
1
−W
T
2
H
T
)) +
t
0
e
(A−HC)(t−τ)
E((W
1
(τ) −HW
2
(τ))
(W
T
1
(t) −W
T
2
(t)H
T
)) dτ
=
t
0
e
(A−HC)(t−τ)
E((W
1
(τ) −HW
2
(τ))
(W
T
1
(t) −W
T
2
(t)H
T
)) dτ
=
t
0
e
(A−HC)(t−τ)
(V
1
δ(t −τ) + HV
2
H
T
δ(t −τ)) dτ
=
1
2
V
1
+
1
2
HV
2
H
T
.
To get from the ﬁrst righthand side to the second, we note that the initial
condition e(0) is uncorrelated with W
T
1
−W
T
2
H
T
. The fact that W
1
and HW
2
are uncorrelated leads to the third line, and the ﬁnal result follows from
t
0
δ(t −τ)dτ =
1
2
,
i.e., the written integral includes only half of the impulse.
The ﬁnal expression for E(e(W
T
1
− W
T
2
H
T
)) is symmetric, and therefore ap
pears in Equation 205 twice, leading to
˙
Σ = (A −HC)Σ + Σ(A
T
−C
T
H
T
) + V
1
+ HV
2
H
T
. (206)
This equation governs propagation of the error covariance. It is independent
of the initial condition e(0), and depends on the (as yet) unknown estimator
gain matrix H.
19.4 Step 2: H as a Function of Σ 131
19.4 Step 2: H as a Function of Σ
We now make the connection between Σ = E(ee
T
) (a matrix) and J =
E(e
T
We) (a scalar). The trace of a matrix is the sum of its diagonal ele
ments, and it can be veriﬁed that
J = E(e
T
We) = trace(ΣW). (207)
We now introduce an auxiliary cost function deﬁned as
J
= trace(ΣW + ΛF), (208)
where F is an n n matrix of zeros, and Λ is an n n matrix of unknown
Lagrange multipliers. Note that since F is zero, J
= J, so minimizing J
solves
the same problem. Lagrange multipliers provide an ingenious mechanism for
drawing constraints into the optimization; the constraint we invoke is the
evolution of Σ, Equation 206:
J
= trace
ΣW + Λ(−
˙
Σ + AΣ −HCΣ + ΣA
T
−ΣC
T
H
T
+ V
1
+ HV
2
H
T
)
(209)
If J
is an optimal cost, it follows that ∂J
/∂H = 0, i.e., the correct choice of
H achieves an extremal value. We need the following lemmas, which give the
derivative of a trace with respect to a constituent matrix:
∂
∂H
trace(−ΛHCΣ) = −Λ
T
ΣC
T
∂
∂H
trace(−ΛΣC
T
H
T
) = −ΛΣC
T
∂
∂H
trace(ΛHV
2
H
T
) = Λ
T
HV
2
+ ΛHV
2
.
Proofs of the ﬁrst two are given at the end of this section; the last lemma uses
the chain rule, and the previous two lemmas. Next, we enforce Λ = Λ
T
, since
the values are arbitrary. Then the condition ∂J
/∂H = 0 leads to
0 = 2Λ(−ΣC
T
+ HV
2
), satisﬁed if
H = ΣC
T
V
−1
2
. (210)
132 19 KALMAN FILTER
Duality of Linear Quadratic Regulator and Kalman Filter
Linear Quadratic Regulator Kalman Filter
˙ x = Ax + Bu ˙ x = Ax + Bu + W
1
y = Cx + W
2
u = −Kx
˙
ˆ x = Aˆ x + Bu + H(y −Cˆ x)
2J =
∞
0
(x
T
Qx + u
T
Ru)dt J = E(e
T
We)
Q ≥ 0, R > 0 V
1
≥ 0, V
2
> 0
K = R
−1
B
T
P H = ΣC
T
V
−1
2
PA + A
T
P + Q−PBR
−1
B
T
P = 0 ΣA
T
+ AΣ + V
1
−ΣC
T
V
−1
2
CΣ = 0
Hence the estimator gain matrix H can be written as a function of Σ. Inserting
this back into Equation 206, we obtain
˙
Σ = AΣ + ΣA
T
+ V
1
−ΣC
T
V
−1
2
CΣ. (211)
Equations 210 and 211 represent the practical solution to the Kalman ﬁltering
problem, which minimizes the squarednorm of the estimation error. The
evolution of Σ is always stable, and depends only on the constant matrices
[A, C, V
1
, V
2
]. Notice also that the result is independent of the weighting matrix
W, which might as well be the identity.
19.5 Properties of the Solution
The solution involves a matrix Riccati equation, like the LQR, suggesting a
duality with the LQR problem. This is in fact the case, and the same analysis
and numerical tools can be applied to both methodologies.
The steadystate solution for Σ is valid for timeinvariant systems, leading to
a more common MARE form of Equation 211:
0 = AΣ + ΣA
T
+ V
1
−ΣC
T
V
−1
2
CΣ. (212)
The Kalman Filter is guaranteed to create a stable nominal dynamics A−HC,
as long as the plant is fully stateobservable. This is dual to the stability
guarantee of the LQR loop, when the plant is statecontrollable. Furthermore,
like the LQR, the KF loop achieves 60
◦
phase margin, and inﬁnite gain margin,
for all the channels together or independently.
19.6 Combination of LQR and KF 133
The qualitative dependence of the estimator gain H = ΣC
T
V
−1
2
on the other
parameters can be easily seen. Recall that V
1
is the intensity matrix of the
plant disturbance, V
2
is the intensity of the sensor noise, and Σ is the error
covariance matrix.
• A large uncertainty Σ creates large H, placing emphasis on the corrective
action of the ﬁlter.
• A small disturbance V
1
, and large sensor noise V
2
creates a small H,
weighting the model dynamics Aˆ x + Bu more.
• A large disturbance V
1
, and small sensor noise V
2
creates a large H, so
that the ﬁlter’s correction is dominant.
The limiting closedloop poles of the Kalman ﬁlter are similar, and dual to
those of the LQR:
• V
2
<< V
1
: good sensors, large disturbance, H >> 1, dual to cheap
control problem. Some closedloop poles go to the stable plant zeros, or
the mirror image of unstable plant zeros. The remaining poles follow a
Butterworth pattern whose radius increases with increasing V
1
/V
2
.
• V
2
>> V
1
: poor sensors, small disturbance, H small, dual to expensive
control problem. Closedloop poles go to the stable plant poles, and the
mirror images of the unstable plant poles.
19.6 Combination of LQR and KF
An optimal output feedback controller is created through the use of a Kalman
ﬁlter coupled with an LQR fullstate feedback gain. This combination is usu
ally known as the Linear Quadratic Gaussian design, or LQG. For the plant
given as
˙ x = Ax + Bu + W
1
y = Cx + W
2
,
we put the Kalman Filter and controller gain G together as follows:
134 19 KALMAN FILTER
˙
ˆ x = Aˆ x + Bu + H(y −Cˆ x) (213)
u = −Kˆ x. (214)
H
C
K
B
+

+
+
φ y u
C(s)
φ = (sIA)
1
There are two central points to this construction:
1. Separation Principle: The eigenvalues of the nominal closedloop sys
tem are made of up the eigenvalues of (A−HC) and the eigenvalues of
(A −BK), separately. See proof below.
2. Output Tracking: This compensator is a standalone system that,
as written, tries to drive its input y to zero. It can be hooked up to
receive tracking error e(s) = r(s) − y(s) as an input instead, so that it
is not limited to the regulation problem alone. In this case, ˆ x no longer
represents an estimated state, but rather an estimated state tracking
error. We use the output error as a control input in the next section, on
loopshaping via loop transfer recovery.
19.7 Proofs of the Intermediate Results
19.7.1 Proof that E(e
T
We) = trace(ΣW)
E(e
T
We) = E
¸
n
¸
i=1
e
i
¸
n
¸
j=1
W
ij
e
j
¸
¸
=
n
¸
i=1
n
¸
j=1
Σ
ij
W
ji
,
19.7 Proofs of the Intermediate Results 135
the transpose of W being valid since it is symmetric. Now consider the diagonal
elements of the product ΣW:
ΣW =
Σ
11
W
11
+ Σ
12
W
21
+
Σ
21
W
12
+ Σ
22
W
22
+
¸
¸
¸ →
trace(ΣW) =
n
¸
i=1
n
¸
j=1
Σ
ij
W
ji
. 2
19.7.2 Proof that
∂
∂H
trace(−ΛHCΣ) = −Λ
T
ΣC
T
trace(AHB) = trace
n
¸
j=1
A
ij
l
¸
k=1
H
jk
B
kl
¸
¸
, the il
th element
=
n
¸
i=1
n
¸
j=1
A
ij
l
¸
k=1
H
jk
B
ki
,
where the second form is a sum over i of the ii’th elements. Now
∂
∂H
joko
trace(AHB) =
n
¸
i=1
A
ijo
B
koi
= (BA)
kojo
= (BA)
T
joko
−→
∂
∂H
trace(AHB) = (BA)
T
= A
T
B
T
. 2
19.7.3 Proof that
∂
∂H
trace(−ΛΣC
T
H
T
) = −ΛΣC
T
trace(AH
T
) = trace
n
¸
j=1
A
ij
H
T
jl
¸
¸
, the il
th element
= trace
n
¸
j=1
A
ij
H
lj
¸
¸
=
n
¸
i=1
n
¸
j=1
A
ij
H
ij
,
136 19 KALMAN FILTER
where the last form is a sum over the ii’th elements. It follows that
∂
∂H
iojo
trace(AH
T
) = A
iojo
−→
∂
∂H
trace(AH
T
) = A. 2
19.7.4 Proof of the Separation Principle
Without external inputs W
1
and W
2
, the closedloop system evolves according
to
d
dt
x
ˆ x
¸
=
¸
A −BK
HC A −BK −HC
¸
x
ˆ x
¸
.
Using the deﬁnition of estimation error e = x − ˆ x, we can write the above in
another form:
d
dt
x
e
¸
=
¸
A −BK BK
0 A −HC
¸
x
e
¸
.
If A
represents this compound Amatrix, then its eigenvalues are the roots of
det(sI − A
) = 0. However, the determinant of an upper triangular block
matrix is the product of the determinants of each block on the diagonal:
det(sI−A
) = det(sI−(A−BK))det(sI−(A−HC)), and hence the separation
of eigenvalues follows.
137
20 LOOP TRANSFER RECOVERY
20.1 Introduction
The Linear Quadratic Regulator(LQR) and Kalman Filter (KF) provide prac
tical solutions to the fullstate feedback and state estimation problems, respec
tively. If the sensor noise and disturbance properties of the plant are indeed
wellknown, then an LQG design approach, that is, combining the LQR and
KF into an output feedback compensator, may yield good results. The LQR
tuning matrices Q and R would be picked heuristically to give a reasonable
closedloop response.
There are two reasons to avoid this kind of direct LQG design procedure,
however. First, although the LQR and KF each possess good robustness
properties, there do exist plants for which there is no robustness guarantee
for an LQG compensator. Even if one could steer clear of such pathological
cases, a second problem is that this design technique has no clear equivalent
in frequency space. It cannot be directly mapped to the intuitive ideas of
loopshaping and the Nyquist plot, which are at the root of feedback control.
We now reconsider just the feedback loop of the Kalman ﬁlter. The KF has
openloop transfer function L(s) = Cφ(s)H, where φ(s) = (sI − A)
−1
. This
follows from the estimator evolution equation
˙
ˆ x = Aˆ x + Bu + H(y −Cˆ x)
and the ﬁgure. Note that we have not included the factor Bu as part of the
ﬁgure, since it does not aﬀect the error dynamics of the ﬁlter.
As noted previously, the KF loop has good robustness properties, speciﬁcally
to perturbations at the output ˆ y, and further is amenable to output tracking.
In short, the KF loop is an ideal candidate for a loopshaping design. Supposing
that we have an estimator gain H which creates an attractive loop function
L(s), we would like to ﬁnd the compensator C(s) that establishes
P(s)C(s) ≈ Cφ(s)H, or (215)
Cφ(s)BC(s) ≈ Cφ(s)H.
It will turn out that the LQR can be set up so that the an LQGtype com
pensator achieves exactly this result. The procedure is termed Loop Transfer
138 20 LOOP TRANSFER RECOVERY
H
y
x
C
φ (s)
^
Recovery (LTR), and has two main parts. First, one carries out a KF design for
H, so that the Kalman ﬁlter loop itself has good performance and robustness
properties. In this regard, the KF loop has sensitivity function S(s) = (I +
Cφ(s)H)
−1
and complementary sensitivity T(s) = (I + Cφ(s)H)
−1
Cφ(s)H.
The condition σ(W
1
(s)S(s)) + σ(W
2
(s)T(s)) < 1 is suﬃcient for robust per
formance with multiplicative plant uncertainty at the output. Secondly, we
pick suitable parameters of the LQR design, so that the LQG compensator
satisﬁes the approximation of Equation 215.
LTR is useful as a SISO control technique, but has a much larger role in
multivariable control.
20.2 A Special Property of the LQR Solution
Letting Q = C
T
C and R = ρI, where I is the identity matrix, we will show
(roughly) that
lim
ρ→0
(
√
ρK) = WC,
where K is the LQR gain matrix, and W is an orthonormal matrix, for which
W
T
W = I.
First recall the gain and Riccati equations for the LQR:
K = R
−1
B
T
P
0 = Q + PA + A
T
P −PBR
−1
B
T
P.
Now Q = C
T
C = C
T
W
T
WC = (WC)
T
WC. The Riccati equation becomes
0 = ρ(WC)
T
WC + ρPA + ρA
T
P −PBB
T
P = 0.
In the limit as ρ →0, it must be the case that P →0 also, and so in this limit
20.3 The Loop Transfer Recovery Result 139
ρ(WC)
T
WC ≈ PBB
T
P
= (B
T
P)
T
B
T
P
= (R
−1
B
T
P)
T
RR(R
−1
B
T
P)
= ρ
2
K
T
K −→
WC ≈
√
ρK. 2
Note that another orthonormal matrix W
could be used in separating K
T
from K in the last line. This matrix may be absorbed into W through a
matrix inverse, however, and so does not need to be written. The result of the
last line establishes that the plant must be square: the number of inputs (i.e.,
rows of K) is equal to the number of outputs (i.e., rows of C).
Finally, we note that the above property is true only for LQR designs with
minimumphase plants, i.e., those with only stable zeros (Kwakernaak and
Sivan).
20.3 The Loop Transfer Recovery Result
The theorem is stated as: If lim
ρ→0
(
√
ρK) = WC (the above result), with W
an orthonormal matrix, then the limiting LQG controller C(s) satisﬁes
lim
ρ→0
P(s)C(s) = Cφ(s)H.
The LTR method is limited by two conditions:
• The plant has an equal number of inputs and outputs.
• The design plant has no unstable zeros. The LTR method can be in fact
be applied in the presence of unstable plant zeros, but the recovery is
not to the Kalman ﬁlter loop transfer function. Instead, the recovered
function will exhibit reasonable limitations inherent to unstable zeros.
See Athans for more details and references on this topic.
The proof of the LTR result depends on some easy lemmas, given at the end
of this section. First, we develop C(s), with the deﬁnitions φ(s) = (sI −A)
−1
and X(s) = (φ
−1
(s) + HC)
−1
= (sI −A + HC)
−1
.
140 20 LOOP TRANSFER RECOVERY
C(s) = K(sI −A + BK + HC)
−1
H
= K(X
−1
(s) + BK)
−1
H, then use Lemma 2 →
= K(X(s) −X(s)B(I + KX(s)B)
−1
KX(s))H
= KX(s)H −KX(s)B(I + KX(s)B)
−1
KX(s)H
= (I −KX(s)B(I + KX(s)B)
−1
)KX(s)H, then use Lemma 3 →
= (I + KX(s)B)
−1
KX(s)H
= (
√
ρI +
√
ρKX(s)B)
−1
√
ρKX(s)H.
Next we invoke the result from the LQR design, with ρ → 0, to eliminate
√
ρK:
lim
ρ→0
C(s) = (WCX(s)B)
−1
WCX(s)H
= (CX(s)B)
−1
CX(s)H.
In the last expression, we used the assumption that W is square and invertible,
both properties of orthonormal matrices. Now we look at the product CX(s):
CX(s) = C(SI −A + HC)
−1
= C(φ
−1
(s) + HC)
−1
, then use Lemma 2 →
= C(φ(s) −φ(s)H(I + Cφ(s)H)
−1
Cφ(s))
= (I −Cφ(s)H(I + Cφ(s)H)
−1
)Cφ(s), then use Lemma 3 →
= (I + Cφ(s)H)
−1
Cφ(s).
This result, reintroduced into the limiting compensator, gives
lim
ρ→0
C(s) = ((I + Cφ(s)H)
−1
Cφ(s)B)
−1
(I + Cφ(s)H)
−1
Cφ(s)H
= (Cφ(s)B)
−1
Cφ(s)H
= P
−1
(s)Cφ(s)H.
Finally it follows that lim
ρ→0
P(s)C(s) = Cφ(s)H, as desired.
20.4 Usage of the Loop Transfer Recovery 141
20.4 Usage of the Loop Transfer Recovery
The idea of LTR is to “recover” a Kalman ﬁlter loop transfer function L(s) =
Cφ(s)H, by using the limiting cheapcontrol LQR design, with Q = C
T
C and
R = ρI. The LQR design step is thus trivial.
Some speciﬁc techniques are useful.
• Scale the plant outputs (and references), so that one unit of error in one
channel is as undesirable as one unit of error in another channel. For
example, in depth and pitch control of a large submarine, one meter of
depth error cannot be compared directly with one radian of pitch error.
• Scale the plant inputs in the same way. One Newton of propeller thrust
cannot be compared with one radian of rudder angle.
• Design for crossover frequency. The bandwidth of the controller is
roughly equal to the frequency at which the (recovered) loop transfer
function crosses over 0dB. Often, the bandwidth of is a more intuitive
design parameter than is, for example, the highfrequency multiplicative
weighting W
2
. Quantitative uncertainty models are usually at the cost
of a lengthy identiﬁcation eﬀort.
• Integrators should be part of the KF loop transfer function, if no steady
state error is to be allowed. Since the Kalman ﬁlter loop has only as many
poles as the plant, the plant input channels must be augmented with the
necessary additional poles (at the origin). Then, once the KF design is
completed, and the compensator C(s) is constructed, the integrators are
moved from the plant over to the input side of the compensator. The
tracking errors will accrue as desired.
20.5 Three Lemmas
Lemma 1: Matrix Inversion
(A + BCD)
−1
= A
−1
−A
−1
B(C
−1
+ DA
−1
B)
−1
DA
−1
.
Proof:
142 20 LOOP TRANSFER RECOVERY
(A + BCD)(A + BCD)
−1
= I
A(A + BCD)
−1
= I −BCD(A + BCD)
−1
(A + BCD)
−1
= A
−1
−A
−1
BCD(A + BCD)
−1
= A
−1
−A
−1
BCD(I + A
−1
BCD)
−1
A
−1
= A
−1
−A
−1
B(D
−1
C
−1
+ A
−1
B)
−1
A
−1
= A
−1
−A
−1
B(C
−1
+ DA
−1
B)
−1
DA
−1
. 2
Lemma 2: Short Form of Lemma 1
(X
−1
+ BD)
−1
= X −XB(I + DXB)
−1
DX
Proof: substitute A = X
−1
and C = I into Lemma 1.
Lemma 3
I −A(I + A)
−1
= (I + A)
−1
Proof:
I −A(I + A)
−1
= (I + A)(I + A)
−1
−A(I + A)
−1
= (I + A −A)(I + A)
−1
= (I + A)
−1
. 2
143
21 SYSTEM IDENTIFICATION
21.1 Introduction
Modelbased controller design techniques, such as the LQR and LTR, require
a plant model. The process of generating workable models from observed
data is the goal of system identiﬁcation. Good controllers usually have some
reasonable robustness guarantees, which motivates identiﬁcation with simple
methods. We discuss in this section four fundamental but useful techniques
for approximate system identiﬁcation of singleinput, singleoutput plants. It
should be noted that the area of system identiﬁcation is a very rich one, and
that the methods are only a small subset of what is available.
Except for the last approach, timedomain simulation, the methods are limited
to linear models. If diﬀerent inputs give diﬀerent linear model coeﬃcients, then
it is likely that nonlinear terms are playing a role. The user then has the choice
of ignoring the nonlinearity, for example if the operating point is controlled
very closely, or developing a controller which takes speciﬁc account of the it.
In any event, simulations with the nonlinear plant should always be performed
to assess the robustness of the control strategy.
21.2 Visual Output from a Simple Input
For loworder plants which can tolerate impulse or step input, a great amount
can be learned through step and impulse responses. The basic idea is to
express what is observed as a time signal whose Laplacedomain equivalent
can be recognized. As an example, consider the plant transfer function P(s) =
k/(τs +1), a ﬁrstorder lag with gain k, and time constant τ. We have y(s) =
k/s(τs + 1) for the step input u(s) = 1/s, and therefore y(t) = k(1 − e
−t/τ
).
The gain k is evident as the maximum value taken by the measured output.
The time constant τ is equal to the time required for y(t) to reach the value
k(1 −e
−1
) = 0.632k. Similar estimates can be made for secondorder systems,
especially with the help of step functions parametrized on damping ratio ζ.
Systems with order three or higher will usually be more diﬃcult to assess with
this visual technique.
Example: Two raw step responses in heading were recorded for two dif
ferent vessels. The ﬁrst vessel was strongly unstable and had to be powered
down abruptly after one second; however, the smoothed data can still be ﬁt
ted to the curve y(t) = 0.58(e
t
−1). Give estimates of the two plant transfer
144 21 SYSTEM IDENTIFICATION
0 0.2 0.4 0.6 0.8 1
−0.5
0
0.5
1
1.5
seconds
0 1 2 3 4 5
0
0.2
0.4
0.6
0.8
1
seconds
functions P(s), assuming that the step input was of magnitude one.
The time trace of the ﬁrst system looks like the Laplace transform pair:
y(t) =
1
b −a
(e
−at
−e
−bt
) ↔y(s) =
1
(s + a)(s + b)
for the values b = 0 and a = −1. Thus for the experimental data,
y(s) =
0.58
s(s −1)
.
Since y(s) =
ˆ
P(s)u(s), where u(s) = 1/s, we have
ˆ
P(s) = 0.58/(s − 1). The
second trace looks like a secondorder response of the form
y(s)
u(s)
=
kω
2
n
s
2
+ 2ζω
n
s + ω
2
n
,
where gain k, undamped frequency ω
n
, and ζ are to be determined. First,
we note that the steady value of y(t) is about 0.5, so let k = 0.5. Next,
with respect to the steady value, the ﬁrst overshoot is about 0.3, and the
second overshoot (same side) is about 0.1. The ratio is often written as the
logarithmic decrement δ = ln(0.3/0.1) · 1.10, so that the damping ratio is
simply ζ = δ/2π · 0.175. Finally, the damped natural period is about 1.7s,
leading to the damped natural frequency ω
d
= 2π/1.7 · 3.70rad/s. The
undamped natural frequency ω
n
is related to ω
d
as follows:
21.3 Transfer Function Estimation – Sinusoidal Input 145
ω
n
=
ω
d
√
1 −ζ
2
· 3.75rad/s.
Inserting these into the template above, we have
ˆ
P(s) =
7.0
s
2
+ 1.28s + 14.1
.
21.3 Transfer Function Estimation – Sinusoidal Input
The main idea of transfer function estimation is that P(s) can be estimated by
simply dividing the measured output by the measured input, in the frequency
domain:
ˆ
P(s) =
y(s)
u(s)
. (216)
In applications, the input signal u(s) is known quite accurately because it is
generated by a computer. Two sources of error can corrupt the output signal
y(s), however: real disturbances and sensor noise. In some cases, disturbances
may be bandlimited (e.g., water waves), and if these occur in a frequency
range far away from the dominant dynamics, the estimated transfer function
approach will succeed. A similar argument holds for sensor noise, which in
many devices is negligible for low frequencies. A sensor which has noise in the
frequency range of the system dynamics is problematic for obvious reasons.
This section discusses the use of periodic inputs to create a Bode plot, while
the next section generalizes to broadband input. Bode plots are ﬁgures of
transfer function magnitude and phase as functions of frequency, which can
be either parameterized in terms of poles, zeros, and a gain, or used directly
in a loopshaping or Nyquist plot approach.
Certain plants which cannot admit a step or impulsive input will tolerate a
sinusoidal input. The idea then is to drive the plant with a narrowband, i.e.,
periodic, input signal u(s). For each such test at a speciﬁc frequency, compute
the magnitude and phase relating the input to the output. Conduct as many
tests as are necessary to build a Bode plot of the plant estimate
ˆ
P(s).
Example: Sinusoidal rudder angles trajectories, of 10deg amplitude ([u(t)[ =
10deg), were implemented on a vessel operating near its cruising speed. The
data are plotted, along with the magnitude and phase of the transfer function
146 21 SYSTEM IDENTIFICATION
ˆ
P(s) =
0.093
s
2
+ 0.83s + 0.21
,
which holds reasonably well at low frequencies. The phase angle of y(s) is taken
with respect to the input signal u(s). Note that the experimental magnitude
and phase in this example deteriorate at the higher frequencies. This is a
property of almost all physical systems, and an indicator that the plant model
cannot be trusted above a certain frequency range.
−40
−30
−20
−10
0

y
(
s
)
/
u
(
s
)

,
d
B
10
−3
10
−2
10
−1
10
0
10
1
10
2
−200
−150
−100
−50
0
a
n
g
l
e
(
y
(
s
)
)
,
d
e
g
rad/s
21.4 Transfer Function Estimation – Broadband Input
Many times one needs to deal with experimental data that is broadband, either
as the result of a closedloop run, or of signiﬁcant disturbances. In this case,
frequencydomain analysis can still be used, but elements of spectral analysis
are necessary.
21.4 Transfer Function Estimation – Broadband Input 147
21.4.1 Fourier Transform of Sampled Data
For the purpose of analyzing transfer functions, the discrete Fourier transform
(DFT) is a standard, optimized tool. It operates on a data vector x(n) of
length N:
X(m) =
N−1
¸
n=0
x(n)e
−j2πnm/N
, (217)
for m = [1, N]. We review several points for dealing with the DFT calculation:
• The m’th DFT point is a summation of complex unitmagnitude vectors
(e
−j2πnm/N
) times the original data (x(n)). Dividing the DFT result by
N returns the signal to its real magnitude.
• The DFT generates a vector of N complex points as outputs. These
correspond to the frequency range
ω
m
=
2π
dt
m−1
N
, (218)
i.e., the ﬁrst frequency is 0, and the last frequency is slightly less than
the sampling rate.
• The sampling theorem limits our usable frequency range to only onehalf
of the sampling rate, called the Nyquist rate. The points higher than
π/dt correspond to negative frequencies, and a complexconjugacy holds:
X(2) = X
∗
(m), X(3) = X
∗
(m−1), and so on. What happens near the
Nyquist rate depends on whether N is odd or even.
An additional multiplication of the DFT result by 2 will give peaks which
are of about the right magnitude to be compared with the timedomain
signal. With this scaling, the value X(1) is twice the true DC value.
• The Nyquist rate depends only on the sampling time step dt, but the
frequency vector accompanying the DFT can be made arbitrarily long
by increasing the number of data points. To improve the frequency
resolution of the DFT, a common approach is to zeropad the end of the
real data,
148 21 SYSTEM IDENTIFICATION
• The data x(k) should be multiplied by a smoothing window w(k), for
two reasons. First, if the window goes to a small value near its tails,
Gibb’s eﬀect (the spectral signature of a discontinuity between x(1) and
x(N)) can be minimized. The second rationale for windowing is based
on the discrete nature of the transform. Because the DFT provides
frequency data at only N/2 unique frequencies, there is the possibility
that a component in the real data lies in between one of these DFT
frequencies. The DFT magnitude at a speciﬁc ω(m) is in fact an average
of continuous frequencies from the neighborhood of the point. Smoothing
windows are generally chosen to achieve a tradeoﬀ of two conﬂicting
properties relating to the average: the width of the primary lobe (wide
primary lobes bring in frequencies that belong to the adjacent bins ω(m−
1) and ω(m + 1)), and the magnitude of the side lobes (tall side lobes
bring in frequencies that are far away from ω(m)).
There are many smoothing windows to choose from; no windowing at
all is usually referred to as applying a rectangular window. One of the
simplest is the Hann, or cosine, window:
w(k) =
1
2
¸
1 −cos
2πk
N
¸
• The DFT of a given signal is subject to bias and variance. These can
be reduced by taking separate DFT’s of subsections of data (perhaps
zeropadded to maintain frequency resolution), and then averaging them.
It is common to use segments which do not overlap, but it is not a
necessity. This averaging idea is especially important in transfer function
veriﬁcation; see below.
• If possible, the DFT should be run on a number of samples N which is
a multiple of a large power of two. If N is prime, the DFT will take a
long time to execute.
The above steps, all of which are available through the Matlab function
spectrum() for example, will ensure a fair spectral analysis of a time series.
21.4.2 Estimating the Transfer Function
In continuous time, when the plant input y(t) and output u(t) are transformed
into frequency space, the estimated transfer function follows from
21.4 Transfer Function Estimation – Broadband Input 149
ˆ
P(s) =
y(s)
u(s)
. (219)
As noted above, spectral analysis of a signal beneﬁts by using multiple seg
ments and averaging; we now want to include the same approach in our esti
mation of P(s). The procedure is:
• For the p’th segment of data, compute the transfer function estimate
ˆ
P
p
(m) =
Y
p
(m)Y
∗
p
(m)
U
p
(m)U
∗
p
(m)
.
• For the p’th segment of data, compute two covariances and a cross
covariance:
Γ
p
uu
(m) = U
p
(m)U
∗
p
(m)
Γ
p
yy
(m) = Y
p
(m)Y
∗
p
(m)
Γ
p
yu
(m) = Y
p
(m)U
∗
p
(m).
• Construct average values of the transfer function estimate and the other
quantities:
ˆ
P(m) = avg
p
(
ˆ
P
p
(m))
Γ
uu
(m) = avg
p
(Γ
p
uu
(m))
Γ
yy
(m) = avg
p
(Γ
p
yy
(m))
Γ
yu
(m) = avg
p
(Γ
p
yu
(m))
• Compute the coherence function, which assesses the quality of the ﬁnal
estimate
ˆ
P(m):
Coh(m) =
Γ
yu
(m)Γ
∗
yu
(m)
Γ
yy
(m)Γ
uu
(m)
.
150 21 SYSTEM IDENTIFICATION
If the coherence is near zero, then the segmental crosscovariances Γ
p
yu
(m)
are sporadic and have cancelled out; there is no clear relation between
the input and the output. This result could be caused by either dis
turbances or sensor noise, both of which are reasonably assumed to be
random processes, and uncoupled to the input signal. Alternatively, if
the coherence is near one, then the crosscovariances are in agreement
and a real inputoutput relationship exists. With real data, the coher
ence will deteriorate at high frequencies and also at any frequency where
disturbances or noise occur.
21.5 TimeDomain Simulation
The timedomain simulation approach tweaks the parameters of a simulation
so that its output matches the observed output. The method has its main
strength in the fact that it applies to any model that can be simulated, in
cluding those of high order and with signiﬁcant nonlinearities. On the other
hand, the method is computationally expensive and gives no guarantee of a
useful solution, or even of convergence.
At the outset, we need to come up with some structure of the plant model.
This can be based on physics in many cases. Consider, for example, the case
of a mass mounted on a spring and a dashpot, driven by the input force u(t).
A fair guess for the actual dynamics has the form my
+by
+ky = u(t), and
we plan to look through the threedimensional parameter space
θ = [m, b, k].
The simulation operation can be written this way: ˆ y(t) = G(
θ, u(t)), and the
system identiﬁcation problem is to minimize [[ˆ y(t) −y
obs
(t)[[, say, where [[ [[
here indicates the Euclidean norm. For a given parameter vector
θ, running the
simulation generates a new ˆ y, and computing the norm gives a scalar measure
of goodness.
Since the normed error is a complicated function of both u(t) and
θ, the mini
mization must proceed iteratively. The NelderMeade simplex method is easy
to use, and can be invoked with the Matlab function fmins. As an example,
the three programs listed comprise a working Matlab set for identiﬁcation of
a ﬁrstorder, nonlinear system. Some notes on use:
• In this example, the same simulation generates the “observed” data and
the simulated response. Since the program simulate always uses the
global variable theta as the parameters, we must be careful about setting
theta in the calling programs.
21.5 TimeDomain Simulation 151
• After a simulation run is complete, the data is interpolated to the same
time scale as the observed data, in order to compute the error.
• The initial guess for theta is a random vector; the Simplex method will
take over from this point. In many instances, however, theta is roughly
known, and a better starting value can be given.
• The Simplex method may head into invalid parameter space, e.g., nega
tive mass. The error calculation, however, can be easily augmented by
a term which penalizes invalid parameters, e.g.,
err = err + 1000*(1sign(mass)).
• There is no guarantee that a global minimum will be found or even exists.
Starting from diﬀerent initial guesses for θ may help ﬁnd better results,
but we are still at the mercy of the minimization algorithm, and a very
complicated function.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
clear all ; clear global ;
global u_obs t_obs y_obs dt theta ;
dt = .3 ; % time step
theta = [1 2] ; % true parameter vector
t_obs = 0:dt:20*dt ; % observed time vector
u_obs = ones(length(t_obs),1) ; % observed input
[t_raw, y_raw] = ode45(’simulate’, [0 max(t_obs)], 0) ;
y_obs = spline(t_raw, y_raw, t_obs) ; % observed output
[theta_final] = fmins(’get_err’, randn(2,1)) ;
disp(sprintf(’final theta(1): %g.’, theta_final(1))) ;
disp(sprintf(’final theta(2): %g.’, theta_final(2))) ;
theta = theta_final ;
[t_raw, y_raw] = ode45(’simulate’, [0 max(t_obs)], 0) ;
y_sim = spline(t_raw, y_raw, t_obs) ;
figure(1) ; clf ; hold off ;
plot(t_obs, y_obs, t_obs, y_sim, t_obs, u_obs) ;
152 21 SYSTEM IDENTIFICATION
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
function [err] = get_err(theta_arg) ;
global y_obs t_obs theta ;
theta = theta_arg ;
[t_raw, y_raw] = ode45(’simulate’, [0 max(t_obs)], 0) ;
y_sim = spline(t_raw, y_raw, t_obs) ;
err = norm(y_sim  y_obs) ;
disp(sprintf(’error: %f.’, err)) ;
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
function [ydot] = simulate(t,y) ;
global u_obs dt theta ;
k = theta(1) ;
tau = theta(2) ;
ind = floor(t/dt) + 1 ; % we have to choose which u_obs to use:
% a zeroorder hold as implemented.
ydot = ( k*u_obs(ind)  y^3 ) / tau ;
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
153
22 CARTESIAN NAVIGATION
The bulk of our discussion on maneuvering and control has assumed that the
necessary system states can be measured. The marine engineer is in fact faced
with choices between many diﬀerent basic sensor packages, notably compasses,
paddle wheels, inertial navigation units, rate gyros, and depth guages, for ex
ample. These listed sensors are selfcontained and rely primarily on the phys
ical properties of the natural environment. There is also a class of distributed
sensor systems; these generally involve an array of communicating elements, lo
cated remotely from the vehicle. We present the fundamental concepts behind
two methodologies in this second class: the global positioning system (GPS)
and acoustic navigation, both of which can provide highaccuracy absolute
Cartesian navigation.
22.1 Acoustic Navigation
Consider a transponder A, which can transmit an acoustic signal, and also
measure, with microsecond accuracy, the time to receive a reply. Next, place
a responder B at a distance R away from A; the job of B is just to transmit
a signal whenever it receives one, with a (short) predictable response time T
t
.
Thus, the elapsed time T between a tranmission and consequent reception at
A is
T = T
t
+ 2R/c
w
,
where c
w
is the speed of sound in water, about 1450m/s. The range R follows
by inversion:
R =
(T −T
t
)c
2
2
.
Suppose that the location of B is known; then a given measurement of R
places A on a sphere around B. It is a case of three unknowns (x, y, z) and one
equation:
(x
A
−x
B
)
2
+ (y
A
−y
B
)
2
+ (z
A
−z
B
)
2
= R
2
.
154 22 CARTESIAN NAVIGATION
The introduction of a another responder C (typically listening and responding
at a diﬀerent frequency than B) places A on the intersection of two spheres,
i.e., a circle. There are two equations, but still three unknowns. When A,
B, and C lie in a nearly horizontal plane, then the intersection circle lies in a
vertical plane; the addition of a depth sensor to our suite would allow us to pin
A’s location at one of two points on the circle. Finally, if we know which side A
is on, and do not allow for abrupt crossovers, then we have a functional set of
measurements for acoustic navigation with just two responders. The baseline is
the line connecting responders B and C; when A is near this baseline, positional
accuracy will be very poor since the two solution spheres are tangent.
Better and better performance can be obtained by increasing the number of
responders, and consequently of the baselines and spheres. With three re
sponders, for example, the intersection of a sphere (responder D) and a circle
(responders B and C) is two points. Here there are three equations and three
unknowns, but the nonlinearity of the equations leads to the nonuniqueness
in the solution. A fourth responder or a depth transducer would be needed to
completely constrain the solution.
The above discussion is a minimum conceptual explanation of acoustic naviga
tion. There are many other pieces to the approach, including an account of the
variation of sound speed c
w
with water depth, obtaining the Cartesian loca
tions of the responder network, and handling various geometric conﬁgurations
that give rise to poor or degenerate solutions.
There are two common conﬁgurations used for acoustic navigation, named for
the length of the baselines relative to the target (transponder A) range.
(Ultra) ShortBaseline The geometry of SBL or USBL puts very short
baselines between the responders compared to the target distance. For in
stance, the ﬁxed net is often attached to a vessel or other structure, with base
line lengths on the order of 10100m. Typical frequencies in use are around
100kHz, with a working range of 100500m to the target. The wavelength of a
100kHz signal is about 1cm and 5cm is a reasonable estimate of the accuracy
for these systems.
LongBaseline Longbaseline systems typically involve a larger responder
net, and the target ranges are similar to the baseline lengths. Very large
systems utilize frequencies of 1015kHz (for a placement accuracy around 2
5m), and may have tenkilometer baselines. Frequent sources of error in long
baseline systems are variations of c
w
, and also in multipath. In the latter
22.2 Global Positioning System (GPS) 155
B
C
A
3
A*
3
2
A
A*
2
A
1
C B
A
Figure 6: Left: General conﬁguration of a longbaseline acoustic system, with
target transponder A and ﬁxed responders B and C. The position A
1
is on
the baseline and has poor accuracy in the direction normal to the baseline; in
contrast, A
3
is wellposed. A
∗
3
is an additional solution to the tworesponder
problem. Right: General conﬁguration of a shortbaseline system.
condition, false signals can be caused by reﬂections oﬀ the seaﬂoor and the
surface; these signals are typically eliminated by rejecting those receptions
which are outside a very tight and slowly moving window on travel time T.
Sometimes, bottom topography can shield a direct acoustic path, and the only
receptions available are via multipath!
22.2 Global Positioning System (GPS)
The GPS system is the most powerful system publicly available for absolute
position reference above water. It is similar in concept to the acoustic naviga
tion systems described above, but with one fundamental diﬀerence: only the
oneway travel time from each satellite is measured at the target. The accu
racies achieved with GPS are therefore strongly dependent on the accuracy of
timekeeping.
156 22 CARTESIAN NAVIGATION
The core of the system is a network of 24 satellites arranged in six planes (4
per plane) inclined at 55 degrees from the equatorial plane, at an altitude of
about 20,000km. The altitude is chosen carefully to correspond with a 12hour
orbit; geosynchronous orbit is much higher, at around 40,000km. The orbit
lines over a nonrotating earth are thus nearsinusoids that reach lattitudes
of 55
◦
N and 55
◦
S, and the six lines are spaced 60
◦
apart in longitude. With
reference to a rotating Earth, the spatial frequency is doubled, so that the
longitudinal distance between tracks is 30 degrees, or 1800 nautical miles.
Each satellite transmits a regular signal which contains, among other items, the
time of its transmission, and the threedimensional location of the satellite (the
ephemeris). It is from these two pieces of information, from many satellites,
that triangulation and navigation are performed at the target.
For the ideal case that the time bases of the target and the satellites are
exactly aligned, triangulation of the type described for acoustic navigation is
possible. In practical terms, the speed of light c = 3 10
8
m/s implies that
a 1ns timing error will cause a 0.3m range error. For this reason, each of
the satellites carries four atomic clocks on board, good to 1ns per day. The
time base is continually monitored and updated from a ground station, and
accurate to within 1ns. The ephemeris of each satellite is also monitored and
updated from the ground station, using a combination of leastsquares analysis
of past data (1 week), and a Kalman ﬁlter to predict the future ephemeris.
The accuracy of satellite position is better than 10m.
We now come to the last thorny issue: how to make target receivers that don’t
require atomic clocks! The solution is very clever, and illustrated for the case
of twodimensions; the arguments for three dimensions are the same. In two
dimensions, the range measurements from two sources locate the target on one
of two points (the intersection of two circles). Suppose that the target clock
is too slow by an amount
˜
t, so that the range estimate is
r + ˜ r = c(t +
˜
t).
The estimated range circles are too large, and the estimated location of the
target too far from the baseline. Introduce a third satellite now so that the
intersections of the circle pairs now occur at six points. Three of these are
very close, and indicate the approximate true solution. The trick is to ﬁnd the
correction for
˜
t that puts the three close solutions onto a single point. This
single point is near the centroid of the three approximate points, and represents
the best position solution. The target time base can be kept up to date by
22.2 Global Positioning System (GPS) 157
performing this check for every set of signals. Hence, the twodimensional
timing problem is solved with three satellites; the threedimensional timing
problem is solved with four satellites.
The position speciﬁcation for GPS is 25m, at the 95’th percentile. This is a re
markable feat, given that it is an absolute measure over the entire surface and
atmosphere of our 6000kmradius Earth. Major sources of error include: clock
base and satellite navigation (≈ 2.5m), ionosphere and troposphere electro
magnetic variations (≈ 2.5m), and inaccuracies in the receiver and multipath
(≈ 2m). GPS is also subject to selective availability or S/A, the addition of
a slowvarying random component in the satellite ephemeris data. Selective
availability is controlled by the United States Department of Defense, and de
grades the position speciﬁcation to 100m (typical). The advent of diﬀerential
GPS solved most of the S/A concerns of American allies, by providing high
accuracy navigation in local areas. The idea here is that a stationary target
can detect the eﬀects of S/A (since it is not moving) and then transmit the
corrections over a small geographical area. Many current GPS receivers are
capable of decoding these local corrections, which are then applied to their
own satellite navigation processing. Diﬀerential GPS typically provides 12m
accuracy.
Finally, as with the acoustic navigation systems described, an independent
altitude measurement will enhance the accuracy of GPS, essentially reducing
a threedimensional to a twodimensional problem.
158 22 CARTESIAN NAVIGATION
159
23 REFERENCES
Short notes indicate how the reference is related to the text.
Dynamics of Mechanical and Electromechanical Systems, S.H. Crandall, D.C.
Karnopp, E.F. Kurtz, Jr., and D.C. PridmoreBrown, Krieger: Malabar, FL,
1985. Coordinate systems and rigidbody dynamics.
Optimal Spacecraft Rotational Maneuvers, J.L. Junkins and J.D. Turner, El
sevier: New York, 1986. Euler angles and quaternions.
Principles of Naval Architecture, 2nd Revision, E.V. Lewis, ed., Society of
Naval Architects and Marine Engineers: Jersey City, NJ, 1988. Classical treat
ment of stability, propulsion, and lifting surfaces.
Marine Hydrodynamics, J.N. Newman, MIT Press: Cambridge, 1997. General
reference.
Guidance and Control of Ocean Vehicles, T.I. Fossen, Wiley: New York, 1994.
Comprehensive modern treatment of marine dynamics and control.
Formulas for Natural Frequency and Mode Shape, R.D. Blevins, Krieger: Mal
abar, FL, 1984. Extensive added mass formulas.
A method for estimating static aerodynamic characteristics for slender bodies
of circular and noncircular cross section alone and with lifting surfaces at
angles of attack from 0
◦
to 90
◦
, L. Jorgensen, NASA Technical Note D7228,
April 1973.
Fluid Dynamic Lift, S.F. Hoerner and H.V. Borst, Hoerner Fluid Dynamics:
Bakersﬁeld, CA, 1975. Extensive data on lift of bodies and wings.
Fluid Dynamic Drag, S.F. Hoerner and H.V. Borst, Hoerner Fluid Dynamics:
Bakersﬁeld, CA, 1975. Data on drag of bodies and wings.
Marine Propellers and Propulsion, J.S. Carleton, ButterworthHeinemann:
Boston, 1994. Includes a chapter on transverse and azimuthing thrusters.
The inﬂuence of thruster dynamics on underwater vehicle behavior and their
160 23 REFERENCES
incorporation into control system design, D. R. Yoerger, J.G. Cooke, and J.
J. E. Slotine, IEEE J. Oceanic Engineering, 15:167178, 1990. Firstorder
nonlinear thruster model.
Toward an improved understanding of thruster dynamics for underwater ve
hicles, A.J. Healey, S.M. Rock, S. Cody, D. Miles, and J.P. Brown, Proc.
IEEE Symposium on Autonomous Underwater Vehicle Technology, pp 340
352. 1994. Secondorder nonlinear thruster model.
Accurate fourquadrant nonlinear dynamical model for marine thrusters: The
ory and experimental validation, R. Bachmayer, L.L. Whitcomb, and M.A.
Grosenbaugh, IEEE J. Oceanic Engineering, 25:146159, 2000. Stateofthe
art in thruster modeling.
Acceleration and steadystate propulsion dynamics of a gas turbine ship with
controllablepitch propeller, C.J. Rubis, Proc. SNAME Annual Meeting, No.
10, 1972.
Governing ship propulsion gas turbine engines, C.J. Rubis and T.R. Harper,
SNAME Trans., 94:283308, 1986.
FlowInduced Vibrations, 2nd Edition, R.D. Blevins, Van Nostrand Reinhold,
New York, 1990. Cable strumming.
Modern Control Engineering, K. Ogata, PrenticeHall: Englewood Cliﬀs, NJ,
1970. SISO linear control.
Circuits, Signals, and Systems, W.M. Siebert, MIT Press: Cambridge, MA,
1986. General reference on transforms.
Applied Nonlinear Control, J.J.E. Slotine and W. Li, PrenticeHall: Engle
wood Cliﬀs, NJ, 1991.
Feedback Control Theory, J.C. Doyle, B.A. Francis, and A.R. Tannenbaum,
Macmillan: New York, 1992. Loopshaping concepts.
Control Engineering: A Modern Approach, P.R. B´elanger, Sauders College
Publishing, Philadelphia, 1995. General control reference.
Introduction to Dynamic Systems, D.G. Luenberger, Wiley: New York, 1979.
161
Dynamic programming and LQR derivation.
Applied Optimal Estimation, A. Gelb, ed., MIT Press: Cambridge, 1988.
Kalman ﬁlter practical aspects.
Linear Optimal Control Systems, H. Kwakernaak and R. Sivan, Wiley: New
York, 1972. Kalman ﬁlter derivation and properties.
Multivariable feedback design: concepts for a classical/modern synthesis, J.C.
Doyle and G. Stein, IEEE Trans. Automatic Control, 26:416, 1981. Seminal
paper on LTR.
A tutorial on the LQG/LTR method, M. Athans, Proc. American Control
Conference, pp 12891296, 1986.
Multivariable control of a submersible using the LQG/LTR design methodol
ogy, R. Martin, L. Valavani, and M. Athans, Proc. American Control Confer
ence, pp 13131324, 1986. More tutorial LTR.
Optimal Control, F.L. Lewis and V.L. Syrmos, Wiley: New York, 1995. Gen
eral reference on LTR.
Numerical Recipes in C, 2nd Edition, W.H. Press, S.A. Teukolsky, W.T. Vet
terling, and B.P. Flannery, Cambridge University Press: Cambridge, UK,
1992. Function minimization.
Spectral Analysis in Engineering, G.E. Hearn and A.V. Metcalfe, Halsted:
New York, 1996. Transfer functions from spectral analysis.
NAVSTAR: Global positioning system – Ten years later, B.W. Parkinson, S.W.
Gilbert, Proc. IEEE, 71:11771186, 1983. The whole issue is in fact dedicated
to global positioning systems. There is another such issue: IEEE Proc. Vol.
87, No. 1, 1999.
162 23 REFERENCES
163
24 PROBLEMS
Many of the following problems use data presented by Fossen and in Principles
of Naval Architecture.
1. Inertial dynamics. The gas generator turbine of a LM2500 marine
engine rotates at p = 10, 000rpm (revolutions per minute), and has a
mass of approximately 200kg. The following question pertains to the
inertial forces experienced by this spinning rigid body, with I
xy
= I
xz
=
I
yz
= x
g
= z
g
= 0.
Because of high rotational speed, the turbine is exceedingly sensitive to
lateral balancing. Estimate the net force that the radial bearings would
have to support for a 1mm mass imbalance, say, y
g
= 1mm with respect
to the spin axis.
2. Kinematics. Although deadreckoning is not recommended as a method
for longterm navigation, it can be quite useful for preplanning trajec
tories. For example, consider a surface vessel heading proﬁle as follows:
t = 0 φ(t) = 0 rad
0 < t <= 60s φ(t) = t/60 π/4 rad
60s < t < 240s φ(t) = π/4 rad
240s < t < 300s φ(t) = (300 −t)/60 π/4 rad
t = 300s φ(t) = 0 rad
φ
π/4
t = 60s t = 0s t = 240s t = 300s
y
x
φ
What is the total distance traveled in the Cartesian x direction during
these ﬁve minutes? You may assume that the forward speed of the vessel
is u = 1m/s, and that there is no signiﬁcant sideslip. Hint: Break the
trajectory up into three parts, which can be evaluated separately.
164 24 PROBLEMS
3. Coeﬃcients and system modeling. Consider a weather vane in a
wind of velocity U
o
. If θ is the angle of the vane with respect to the wind
direction,
(a) Write the singledegree of freedom (N) linearized equations of mo
tion about the ﬁxed axis 0.
(b) Write N
θ
, N
˙
θ
, and N
¨
θ
in terms of N
v
, N
r
, N
˙ r
, etc..
(c) If we consider the diﬀerential equation
A¨ y(t) + B ˙ y(t) + Cy(t) = 0,
the condition for stability is that A, B, and C must have the same
sign. Express this requirement in terms of the derivatives in the
previous question. Give physical interpretations for what would
make such a device stable or unstable.
(d) Create a numerical model of this system, using the MATLAB ODE
solver ode45. The system equation can be written as two ﬁrstorder
equations:
d
dt
˙
θ
θ
¸
=
¸
−B/A −C/A
1 0
¸
˙
θ
θ
¸
.
Simulate the system response to nonzero initial conditions (e.g.,
θ(0) = 1,
˙
θ(0) = 0). Discuss, using several examples, response
sensitivity to B and C, which are related to the aerodynamic coef
ﬁcients. For example, look at the range ¦A, B, C¦ = ¦1, ±3, ±3¦.
U
ο
x
y
θ O
165
4. Hydrodynamics. The ﬁgure below shows some characteristic ﬂuid
force curves versus a motion parameter. Give the linear hydrodynamic
coeﬃcient at two diﬀerent operating conditions, origin O and A: is it
zero, small, ﬁnite positive, ﬁnite negative?
F F F
A A A O
O O
u u u
5. Secondorder system. Consider the vibration of a massspringdashpot
system:
m¨ x + b ˙ x + kx = 0,
wherein all the coeﬃcients are positive and real (i.e., physical).
(a) Set x(t) = e
st
, and ﬁnd two solutions for s in terms of m, b, and k,
using the quadratic formula. The solution pairs depend on whether
b exceeds a critical value: what is the critical value and how do the
solutions change?
(b) From the initial condition x(0) = x
o
, and ˙ x(0) = 0, ﬁnd the coeﬃ
cients for the general solution
x(t) = c
1
e
s
1
t
+ c
2
e
s
2
.
in terms of x
o
, s
1
, and s
2
. What do you think will happen when
s
1
= s
2
?
(c) Write this x(t) for subcritical b (that is, complex s) in terms of the
standard parameters
166 24 PROBLEMS
ω
n
=
k
m
(undamped natural frequency)
ζ =
b
2
√
km
(damping ratio)
ω
d
= ω
n
1 −ζ
2
(damped natural frequency).
Give your answer in terms of a real exponential multiplied by a
single sine or cosine, and make a sketch.
(d) Sketch the response x(t) for supercritical b, i.e., both s are real, for
the same initial conditions. What happens when the roots s
i
are
far apart vs. when they are close together?
(e) Consider now the case where an input acts to drive the mass from
zero initial conditions:
m¨ x + b ˙ x + kx = u.
This equation deﬁnes a system, with input u, and output x. Using
the Laplace transform, write the transfer function for this system,
i.e., the impulse response, both in frequency (Laplace) space, and
in the timedomain. Make a sketch of the timedomain result. Be
sure to cover both the subcritical and supercritical damping cases.
(f) What is the step response of this system, in both the subcritical
and supercritical damping cases? Include sketches.
6. Submarine roll dynamics and control. A submarine has weight
1200t (tons) and the center of gravity is 0.5m above the center of buoy
ancy (What can you conclude?). The rolling motion can be assumed
to be decoupled from the other motions. This submarine has anti
rolling ﬁns to ensure stability. The control hydrodynamic derivative is
K
δ
= −2.8tm per degree of ﬁn rotation δ, at a forward speed of 5m/s.
(a) Write the equation of motion for roll (K).
(b) If the automatic control law δ = k
1
ψ is used, where ψ is the roll
angle, what range of k
1
ensures stability?
167
(c) If the speed suddenly drops to 2.5m/s, how does the range of sta
bilizing k
1
change?
7. Steering. Draw curves of rudder angle δ vs. yaw rate r for a dynamically
stable surface vessel, and then for an unstable vessel. Indicate areas
where the vessel turns against the rudder action.
8. Submarine pitch/heave dynamics. Consider a submarine moving
forward at speed U, and restricted to small motions in the vertical (x−z)
plane. Assume that:
• The submarine is symmetric in the x − y plane and y
G
= 0. The
submarine is not symmetric in the x −z plane due to the sail, etc.,
but assume it is nearly symmetric, so that you can omit certain
hydrodynamic terms.
• At rest, the submarine is neutrally buoyant and stable with the x
and y axes horizontal. What does this tell us about the magnitude
of the buoyant force and where it acts with respect to the center of
weight?
(a) Derive the equations of motion for the submarine moving at speed
U.
(b) Derive the hydrostatic, restoring pitch moment for small pitch θ.
(c) Linearize the inertial terms in the equations of motion.
(d) Expand the ﬂuid forces and moments in terms of the motions. Omit
nonlinear and memory eﬀects, and be sure to include the hydrostatic
moment. Explain your choices.
(e) Write out the complete linearized equations of motion. Does surge
decouple from pitch and heave? Write out the coupled equations in
matrix form.
(f) Can the submarine be stable in pitch without feedback control?
Can the submarine be stable in depth without feedback control?
9. Submarine stability via slender body theory.
(a) An underwater vehicle hull has the shape of a circular cylinder,
except at the very ends, where it rapidly tapers down to a point. If
168 24 PROBLEMS
the radius is 0.5m, length is 17m, and speed is 2m/s, use slender
body theory to express the following coeﬃcients: Z
˙ w
, M
˙ q
, and M
w
,
all referenced to the body midpoint.
(b) It turns out that in order to match the lift and moment of the hull
in experiments, a ﬂat tail of radius 0.4m has to be included in the
slenderbody estimate. Determine the corrected value for M
w
, and
in addition compute Z
w
. Where is the aerodynamic center with
respect to the center of the hull, and is this hull stable in pitch?
Recall that the aerodynamic center is the equivalent location of the
observed lift force that creates the observed moment.
(c) A ﬁn with predetermined lift slope of 3.6/rad is to be applied at
the tail to make the device just stable in pitch. How big is the
required ﬁn area that brings the net M
w
to zero, as desired?
10. Lifting surfaces. Using the lifting surface formulas, estimate the coef
ﬁcient K
v
of a submarine sail as shown below, for U = 10m/s.
15m
10m
7m
11. Slender body theory. Consider a long ellipsoidal body of length l and
diameter d.
(a) With l/d = 7.0, approximate the crossbody added mass, using
slenderbody theory, and compare it with the exact results from
the table below (Blevins).
(b) Perform the slender body calculation also for a sphere, and compare
again to the exact result.
(c) What is the added mass in the inline direction?
169
Sphere added mass: 2ρπa
3
/3
Ellipsoid crossbody added mass: 4αρπab
2
/3
where a/b = 0.1: α = 0.075
0.2: 0.143
0.6: 0.355
1.0: 0.500
2.0: 0.704
5.0: 0.894
7.0: 0.933
10.0: 0.960
∞: 1.000
Ellipsoid longitudinal added mass: 4αρπab
2
/3
where a/b = 0.1: α = 6.148
0.2: 3.008
0.6: 0.908
1.0: 0.500
2.0: 0.210
5.0: 0.059
7.0: 0.036
10.0: 0.021
∞: 0.000
2a
2b
12. Submarine pitch stability with various methods. We will calculate
the pitch derivative due to an angle of attack, M
w
, for a small submarine,
using several diﬀerent methods.
The submarine is a rotationally symmetric ellipsoid with L/D = 7, L =
35m. We ﬁx the body origin to the midpoint as shown below. The body
170 24 PROBLEMS
is appended with two ﬁns as shown in the ﬁgure: each has an area of
1m
2
, geometric aspect ratio of 3, and has its longitudinal center of action
1m forward of the body stern.
(a) Apply wing theory to characterize M
w
due to the ﬁns alone.
(b) Apply slender body theory to estimate M
w
for the body alone.
(c) Linearize the Munk moment, and give a corrected slenderbody
value for M
w
of the body.
(d) Apply Jorgensen’s approximate formulas to estimate M
w
for the
body; make linearizations where necessary.
(e) Use the experimental data from Hoerner (p. 13.2) below to estimate
M
w
for the body.
(f) Use the Hoerner result and your result from wing theory to write a
net M
w
for the body plus ﬁns.
(g) What is the location of the net aerodynamic center? If it has an
unstable location forward of the midpoint, what increase in ﬁn area
would bring it back to the midpoint?
(h) What is the diameter of a ﬂat stern that will allow the slender body
theory to give the same moment as the Hoerner (experimental)
data, for the body alone?
(i) What is the diameter of a ﬂat stern that allows the approximate
formulas of Jorgensen to give the same moment as the Hoerner data,
for the body alone?
Hoerner p. 13.2:
Symmetric body of revolution with L/D = 6.7.
Force and moment referenced to body midpoint:
Z = −0.5ρU
2
D
2
C
ydb
tan(w/u). C
y
= 1.20.
M = 0.5ρU
2
D
2
LC
nb
tan(w/u). C
m
= 0.53. DESTABILIZING
171
1m
Each fin: 1m area
AR = 3
2
L = 35m
y
x
L/D = 7
13. Cable dynamics. A towing cable is steeljacketed, with diameter 3cm,
an eﬀective extensibility of E = 100 10
9
Pa, and a density of ρ
c
=
5000kg/m
3
. The vehicle it tows is streamlined, and has negligible mass
(M). If we are towing at low speed and L = 4000m depth,
(a) Calculate ω
1
, the ﬁrst undamped natural frequency in heave, with
mL >> M.
(b) If the vessel heaves with a P = 2m amplitude and a 4.5second
period, what is the dynamic tension amplitude at the top (s = L) of
the cable? The formula for this dynamic tension (with mL >> M)
is
˜
T = EAPk sin(ks)/cos(kL), where A is the crosssectional area
of the cable, k = ω
m/EA, and m is the cable mass per unit
length.
(c) Taking into account the static tension induced by inwater weight,
does the cable unload at the surface, for the heaving conditions
above?
14. Cable mechanics (hard!). You are asked to assess the operational
envelope of a cable/vehicle system which has been installed on a vessel.
The cable is steeljacketed, with diameter 3cm, an eﬀective extensibility
of 100 10
9
Pa, and a density of 5000kg/m. The vehicle is streamlined,
and has a mass (material plus added) of 100kg.
(a) The towing angle at the vessel cannot exceed 25deg from vertical,
for reasons of deck and crane geometry. If this angle is considered
equal to the critical angle φ
c
, what is the maximum speed for the
system? Assume the normal drag coeﬃcient of the cable is C
n
=
1.2.
172 24 PROBLEMS
(b) Considering the undamped, uncoupled axial dynamics, express the
natural frequency as a function of the vehicle depth, and sketch the
curve. If the fastest waves that the surface vessel responds to have
period 4s, what is the “threshold” operating depth?
(c) The undamped, uncoupled lateral dynamics follow the equation
(m + m
a
)
∂
2
q
∂t
2
=
¯
T
∂
2
q
∂s
+ w
n
sin
¯
φ
∂q
∂s
,
where m
a
is the added mass of the cable per unit length, and w
n
is
the inwater weight of the cable. Separation of variables q(s, t) =
˜ q(s) cos ωt gives
¯
T
∂
2
˜ q
∂s
2
+ w
n
sin
¯
φ
∂˜ q
∂s
+ w
n
γ˜ q = 0,
where γ = (m + m
a
)ω
2
/w
n
. Simplify this expression by writing
¯
T = w
n
s (the vehicle weight eﬀect is small), sin
¯
φ · 1 (the cable is
nearly vertical), and then use the substitution z = 2
√
γs to arrive
at the Bessel equation
z
2
∂
2
˜ q
∂z
2
+ z
∂˜ q
∂z
+ z
2
˜ q = 0.
The derivatives can be transformed from s to z coordinates via the
chain rule:
∂˜ q
ds
=
∂˜ q
∂z
∂z
∂s
∂
2
˜ q
ds
2
=
∂
2
˜ q
∂z
2
∂z
∂s
2
+
∂˜ q
∂z
∂
2
z
∂z∂s
∂z
∂s
.
(d) The equation above has a solution of the form ˜ q = cJ
0
(z), where
c is a constant to be found, and J
0
(z) is the zero’th order Bessel
function of the ﬁrst kind. If ˜ q(s = L) = Q, that is, we impose a
harmonic input at the top, ﬁnd c in terms of Q and z(s = L). What
condition makes ˜ q blow up, indicating a resonance frequency ω
n
?
173
(e) Recalling that strumming occurs when ω
n
· 0.2U/d, construct a
graph of towing velocities vs. deployment depths for which the
strumming would be centered. Include at least the three lowest
modes in your sketch, and discuss the trends for the higher modes.
J
0
(x) = 0 for for x · π(n −0.25), n = 0, 1, 2, , especially when x
is large. You should ﬁnd that strumming is hard to avoid in deep
towing applications!
15. Vessel heading control. We consider the yaw/sway dynamics of a
highspeed container ship. All the calculations below are to be made in
nondimensional coordinates, so you do not need to perform any trans
formations in this question.
(a) The nondimensional system with states x
= [v
, r
] evolves accord
ing to dx
/dt
= Ax
+ Bδ, where
A =
¸
−0.90 −0.42
−4.8 −2.3
¸
, B =
−0.13
1.4
¸
.
If the output is yaw rate r
, i.e., C = [0 1], write the transfer
function r
(s)/δ(s) = C(sI − A)
−1
B. Use matrix inversion for the
22 matrix (sI −A), perform the matrix multiplications, and then
use the quadratic formula if necessary to express your answer as
r
(s)
δ(s)
= K
s + z
1
(s + p
1
)(s + p
2
)
,
where K is a constant gain, −z
1
is the zero (there is one real zero),
and [−p
1
, −p
2
] are the two poles of the secondorder system. Are
both poles stable?
(b) For the purposes of autopilot design, the transfer function φ(s)/δ(s)
is crucial. Fortunately, all you have to do is divide r
(s)/δ(s) by an
other factor of s to account for the fact that dφ/dt
= r
. Sketch all
the poles and zeros of the plant transfer function P(s) = φ(s)/δ(s)
on a root locus plot.
174 24 PROBLEMS
(c) A PID controller C(s) will work to stabilize the heading angle,
and brings to P(s)C(s) one additional pole at the origin, and two
arbitrary zeros. Demonstrate a rough PID design by devising two
controller zero locations that will attract the closedloop poles into
the lefthalf plane. Draw the path that you expect each closedloop
pole to follow, as the control gain increases from zero to very large
values.
You do not need to calculate the PID gains that go with your pro
posed zero locations.
16. Fullstate feedback of a submarine. The coeﬃcients governing the
pitch/heave dynamics of a Deep Submergence Rescue Vehicle (DSRV)
are given below. They are nondimensionalized as in the lecture notes,
using the factors ρ/2, U, and L.
I
xx
= 0.000118 M
q
= −0.0113 Z
q
= −0.0175
I
zz
= 0.00193 M
˙ q
= −0.00157 Z
˙ q
= −0.000130
x
G
= 0 M
w
= 0.0112 Z
w
= −0.0439
m
= 0.0364 M
˙ w
= −0.000146 Z
˙ w
= −0.0315
U = 2.0m/s M
δ
= −0.0128 Z
δ
= −0.0277
L = 15.0m M
θ
= −0.156/U
2
(a) Write the linearized (nondimensional) dynamics in the matrix form
˙ x = Ax + Bu, where the input u = δ, and the state vector is
x = [w
, q
, θ, Z
]. Z
here is the elevation of the vehicle in inertial
coordinates; your approximation for
˙
Z
should include both w
and
θ.
(b) Is the DSRV vehicle openloop stable, that is, without control action
δ? You can assess this either by ﬁnding the eigenvalues of your A
matrix above, or by computing the stability parameter C
.
(c) In preparation for controller design, create a simulation using Mat
lab. Make a graph of the openloop step response, showing all the
dimensional state variables versus dimensional time.
(d) Assuming that all the states can be measured accurately, a full
state controller δ = −Kx can be used, where K is a 1 4 gain
matrix. Under this control, the system dynamics are governed by
175
˙ x = (A − BK)x, so that if A − BK has all negative eigenvalues,
the system is stable.
Use the Matlab function place to put the four closedloop poles at
g(−0.95 ± 0.31i) and g(−0.59 ± 0.81i), g = 1. Demonstrate that
your closedloop design is stable against nonzero initial conditions.
What are the eﬀects of increasing or decreasing g?
Note that the pole locations suggested above are with respect to
the dimensional system, i.e., using dimensional time.
17. Nyquist stability. The inverted pendulum shown below is often used
as a simple model for rocket ﬂight, and can also illustrate the dynamic
behavior of an unstable ocean vessel which is propelled from the stern,
e.g., a barge being pushed by a tugboat.
For this problem, we assume that all of the mass (m) is concentrated
at the distal end of an arm of length l; a highperformance positioning
system sets the horizontal position of the cart u
o
.
u
o
m
φ
g
l
A Lagrangian derivation of the dynamics gives:
¨
φ =
g
l
sin(φ) −
1
l
cos(φ)¨ u
o
.
Note that the acceleration of the cart is now considered to be the input
to the plant, e.g., u = ¨ u
o
.
(a) If the observation is taken to be the angle of the bar φ, i.e., y =
φ, write the statespace representation of the dynamics linearized
about the point φ = 0.
176 24 PROBLEMS
(b) Show that this plant is fully state observable and controllable.
(c) Now make the assumption that l = 1, and g = 1 (it can be achieved
through a proper nondimensionalization) What is the transfer func
tion of the plant, P(s)? Use this plant model for the rest of this
question.
(d) Where are the plant poles and zeros, if any?
(e) A stabilizing controller is not hard to ﬁnd, and a suggested one is:
C(s) =
−4(2s + 1)
s + 2
.
Create a rough Nyquist plot of the loci of P(s)C(s), for the frequen
cies ω = [0, 2.2, ∞]rad/s. These three frequencies are all that you
need to sketch the overall shape; there are no hidden loops or other
features. In the case of ω = 2.2, for which you will need to make
explicit calculations, you may ﬁnd the following identity useful:
a + jb
c + jd
=
(a + jb)
(c + jd)
(c −jd)
(c −jd)
=
(ac + bd) + j(bc −ad)
c
2
+ d
2
.
Be sure to include the complexconjugate points for −ω on your
plot.
(f) Invoke the Nyquist stability criterion to conﬁrm that the closed
loop system is stable.
(g) About how many degrees of phase margin does this design provide?
What reduction in lowfrequency gain can be tolerated? Is it a
reasonable design?
18. Root locus and loopshaping. The parameters governing the surface
maneuvering of a highspeed container ship are given below for reference:
m
0.00792 I
zz
0.000456 x
G
−0.05
L 175.0m U 8m/s Y
˙ v
−0.00705
Y
˙ r
0.0000 Y
v
−0.0116 Y
r
0.00242
Y
δ
−0.00258 N
˙ v
0.0000 N
˙ r
−0.000419
N
v
−0.00385 N
r
−0.00222 N
δ
0.00126
177
Note that the center of vessel mass is located aft of the origin; for this
model, the origin coincides with the center of added mass, so that Y
˙ r
=
N
˙ v
= 0. The nondimensional system with states x
= [v
, r
] evolves
according to dx
/dt
= Ax
+ Bδ, where
A =
¸
−0.90 −0.42
−4.8 −2.3
¸
, B =
−0.13
1.4
¸
.
The relevant output is yaw rate: C = [0 1] and D = 0. For the purposes
of autopilot design, however, the transfer function φ(s)/δ(s) is needed.
The following steps create two heading autopilots, using the rootlocus
and loopshaping techniques. In addition to the Matlab commands listed
below, you will ﬁnd very useful the convolution function conv() which
can be used to combine systems, e.g., for numerators,
numPC = conv(numP,numC);. Also, be sure that you equalize axis scal
ing for your plots in the complexplane, by using axis(’equal’);.
(a) Use the Matlab command tf(), or ss(), to create a system model
of the openloop transfer function P(s)C(s), using the plant above
and a PIDtype controller:
C(s) = k
p
1 + τ
d
s +
1
τ
i
s
.
The actual numerical values for k
p
, τ
d
, and τ
i
are to be found in the
next step.
(b) Using τ
d
= 2 and τ
i
= 6 as suggested values, use the Matlab com
mand rlocus() and then rlocfind() to select a controller gain
k
p
, that puts the three slow poles in the following sector: 1) min
imum undamped frequency (nondimensional) of 0.3, 2) maximum
frequency of 0.5, and 3) minimum damping ratio 0.7. Give a root
locus plot, with your pole locations clearly marked on top of the
trajectories taken as k
p
varies. You don’t need to show the fourth,
fast pole, which will be quite far to the left.
(c) Apply the k
p
you selected to P(s)C(s), and then use the Matlab
functions feedback() to create the resulting feedback system, and
178 24 PROBLEMS
step() to plot the closedloop system response to a step input in
desired heading.
(d) Use the Matlab command nyquist() to make a Nyquist plot of
P(s)C(s) for your design. Make a visual estimate of the gain and
phase margins.
(e) An alternate approach for controller design of this stable plant
is loopshaping: For the openloop function L(s) = ω
c
/s, where
ω
c
= 2.0, invert the plant to come up with a compensator: C(s) =
L(s)/P(s). This design has inﬁnite gain margin and 90 degrees
phase margin.
(f) As above, create the feedback system, and plot the closedloop step
response.
(g) The loopshaping control is not quite a PIDcontroller; how does it
diﬀer, and what would L(s) have to contain to make it a PID?
(h) The controllers you just designed are in nondimensional time coor
dinates; give the P,I, and D gains for use on a real time scale, for
the rootlocus design.
19. LQR. Consider the statespace system and LQR design:
A =
¸
0 −1
1 0
¸
,
B = [1 0]
T
C = [0 1]
D = 0
Q = C
T
C
R = ρ.
The plant is an undamped oscillator with undamped poles at ±j. Note
that the plant output is position for this problem.
(a) What is the control gain K in terms of ρ? Hints: There are two
solutions for p
12
; choose the positive one. Also, the expression for
p
22
is messy; luckily, you won’t need to use it.
179
(b) Determine the limiting approximations for K with ρ very small
and very large – these are the cheap control and expensive control
problems.
(c) Derive the limiting closedloop pole locations for ρ −→0, giving the
frequency and damping ratio of the Butterworth pattern in terms of
ρ. You can get the characteristic equation for the poles as det(sI −
(A−BK)) = 0, and then make it ﬁt the form s
2
+ 2ζω
n
+ω
2
n
= 0.
20. LQR. Consider the ﬁrstorder, unstable system governed by ˙ x = x +u,
for which the output is y = x. The statespace matrices are A = B =
C = 1.
(a) Solve the LQR Riccati equation for P, for the case Q = C
T
C and
R = ρ. Your answer should give a positive P as a function of ρ
alone.
(b) Consider the cheapcontrol problem, with ρ →0. Write the leading
term approximations for P and then K, and compute the eigenvalue
(there is only one) of the closedloop system. Note that eig(M) =
M, if M is a scalar instead of a matrix.
(c) Now look at the expensivecontrol problem, with ρ → ∞. Again,
write the approximations for P and K, and ﬁnd the closedloop
system eigenvalue.
21. LQG/LTR. For the inverted pendulum plant (see previous question)
with statespace matrices
A =
¸
0 1
1 0
¸
, C = [0 1] ,
the KF Riccati equation yields H = [100 14]
T
, for the choices V
1
=
[100 0 ; 0 0], and V
2
= 0.01.
(a) Compute the openloop transfer function for the Kalman ﬁlter loop:
L(s) = C(sI −A)
−1
H.
180 24 PROBLEMS
(b) Even though this L(s) is unstable, its magnitude plot conﬁrms that
it is a reasonable loopshaping design. What is the lowfrequency
gain for this L(s), and what kind of tracking performance should
we expect from the associated LTR design?
(c) Now consider the closedloop transfer function, which you can write
as S(s) = L(s)/(1 +L(s)) (the sensitivity). What is the character
istic equation for the closedloop system, and about what damping
ratio has the Kalman ﬁlter provided?
22. LQG/LTR design. The parameters for the linearized sway/yaw mo
tions of a swimmer delivery vehicle are given below.
% Parameters, all nondimensional except [U,L]
U = 4.0 ; % m/s
L = 5.3 ; % m
Izz = 0.006326 ;
m = 0.1415 ;
xg = 0. ;
Yv = 0.1 ;
Yr = 0.03 ;
Nv = 0.0074 ;
Nr = 0.016 ;
Ydelta = 0.027 ;
Yvdot = 0.055 ;
Yrdot = 0. ;
Nvdot = 0. ;
Nrdot = 0.0034 ;
Ndelta = 0.013 ;
You are asked to develop an LQG/LTR controller for this plant, and it is
suggested that you compose a single Matlab script to perform the steps
in sequence. Please make sure you answer all the questions, and include
a listing of your code. This entire design is made in nondimensional
coordinates.
(a) Plant Modeling and Characteristics
181
i. Construct a statespace plant model, to take rudder angle δ as
an input and give heading angle φ as an output. Please provide
the numerical values for the A, B, C matrices. There should be
three states in your model, with one input channel and one
output channel.
ii. Compute and list the controllability and observability matrices;
is the plant statecontrollable and stateobservable?
iii. Where are the poles of your plant model? Is this model stable?
iv. Show a plot of your plant’s step response.
(b) LQR and KF Designs
i. Using the Matlab command lqr(), you can compute the LQR
feedback gain K, for given A, B, Q, and R matrices. With the
choices Q = C
T
C, and R = ρ, list K and plot the closedloop
step responses for the choices ρ = [0.1, 0.001, 0.00001]. How do
the gains and step responses change as you make ρ smaller and
smaller?
Note that the fundamental closedloop LQR system is
˙
x = (A −BK)x + BKx
desired
y = x,
i.e., the input to the closedloop system is x
desired
and the out
put is x. Your plot should show speciﬁcally the output φ, for an
input of x
desired
= [v
desired
= 0, r
desired
= 0, φ
desired
= 1]. This
compression can be achieved in one step by premultiplying the
system by C
T
, and postmultiplying it by C:
˙
x = (A −BK)x + BKC
T
y
desired
y = Cx,
ii. The Matlab command lqe() can be used to generate the Kal
man ﬁlter gain H, given design matrices A, C, V
1
, and V
2
. For
the choices V
1
= I
3×3
and V
2
= 0.01, compute H, and make
a plot of the closedloop step response. Be sure to give the
numerical values of H.
182 24 PROBLEMS
Note that the lqe() command asks for a disturbance gain ma
trix G; you should set this to I
3×3
. The closedloop KF system
is as follows:
˙
ˆ x = (A −HC)ˆ x + Hy
ˆ y = Cˆ x,
i.e., the input is the measurement y and the output is an esti
mated version of it, ˆ y.
(c) Loop Transfer Recovery
The LQG compensator is a combination of the KF and LQR designs
above. With normal negative feedback, the compensator C(s) has
the following state space representation:
˙
z = (A −BK −HC)z + He
u = Kz,
so that the input to the compensator is the tracking error e =
r − y, and its output u is the control action to be applied as
input to the plant. The total openloop transfer function is the
P(s)C(s); in Matlab, you may simply multiply the systems, e.g.,
sysPC = sysP * sysC ;.
i. Make a log(magnitude) plot of the KF openloop transfer func
tion L(s) = C(sI − A)
−1
H, versus log(frequency). You may
ﬁnd the Matlab command freqresp() helpful. [L(s)[ should
be large at low frequencies, and small at high frequencies, con
sistent with the rules of loopshaping.
ii. As ρ →0, the product P(s)C(s) →L(s). Demonstrate this by
computing P(s)C(s) for the three diﬀerent values of ρ above,
and overlaying the respective [P(s)C(s)[ over the plot of part
3a).
iii. Make a closedloop step response plot for the smallest value of
ρ. How does it compare with the KF step response of part 2b)?
In real LTR applications, the particular values of V
1
and V
2
can be picked
to control the lowfrequency gain, and crossover frequency of the open
loop KF system L(s) = C(sI −A)
−1
H.
183
23. Inertial navigation. A triaxial accelerometer package is aligned with
the [x,y,z] (fwd, port, up) axes of an underwater vehicle. At a particular
instant in time, the three raw accelerations from the strain guages are
[¨ x = −0.96, ¨ y = 0.80, ¨ z = 9.73] meters/second. Note that under the
inﬂuence of gravity alone, this sensor will always report that the vehicle
is accelerating upwards.
(a) Ignoring the coupling of Euler small angles, what estimates can
you give for the roll and pitch angles? I suggest that you draw the
measurement vector in the coordinates of the vehicle frame, and
then consider the orientation of the body which would lead to it.
(b) Given that the acceleration due to gravity is 9.81m/s, how do we
know that these are reasonable (but not foolproof!) angle calcula
tions to make?
184 24 PROBLEMS
24. LQG/LTR design: term project. You are asked to carry out mod
eling and control system design for the surge dynamics of an oceanic
cablelaying vessel. This vessel needs to have very good speed control
so that the cable, which is being paid out constantly and sinks slowly
under its own weight, lies properly on the seaﬂoor. The bathymetry
for the path is wellknown but variable, so the vessel will need to slow
down and speed up frequently. Unexpected deviations in speed from the
calculated trajectory are likely to create loops or kinks in the cable, or
induce large tensions.
vessel length L 145m
draft h 9m
beam b 22m
displacement ∇ 17750m
3
surge added mass m
a
0.07ρ∇
thrust reduction factor t 0.19
wetted surface area A
w
5585m
3
towed resistance coeﬃcient C
r
0.0025
wake fraction at propeller w 0.22
propeller diameter D 7.21m
pitch/diameter ratio P/D 1.1
rotative eﬃciency η
r
1.025
zerospeed thrust coeﬃcient k
t
(J = 0) 0.6
slope of k
t
with J 0.522
zerospeed torque coeﬃcient k
q
(J = 0) 0.1
slope of k
q
with J 0.0833
rotational moment of inertia I
p
2.03 10
5
kgm
2
engine parameter a 0.876
b 0.208
c 2.173
d .0939
maximum fuel rate f
m
1.5kg/s
maximum torque Q
m
70000Nm
maximum speed n
m
60Hz
gearbox reduction ratio λ 32
eﬃciency η
g
0.97
185
It is desired to create a linear controller for the nominal surge motion,
and then demonstrate that it will work with a simulation of the real
nonlinear system, within a neighborhood of the nominal speed.
The vessel is powered by a single gas turbine engine, driving one pro
peller. The parameters for the vessel, propeller, and engine system are
as given in the table.
The gas turbine torquespeed characteristic ﬁts the relation:
Q
e
Q
m
= −
a
f
f
m
+ b
n
e
n
m
+
c
f
f
m
+ d
,
where Q
e
represents the engine torque in Nm, f the fuel rate in kg/s,
and n
e
the rotation rate of the engine in Hz.
(a) Make a map of the gas turbine characteristics. For instance, make
a contour plot which gives Q
e
/Q
m
(the fraction of maximum torque
developed) as a function of f/f
m
(the fraction of the maximum fuel
rate) and n
e
/n
m
(the fraction of the maximum rotation rate of the
engine).
Be sure to include clipping on your contours, where the calculated
torque exceeds Q
m
. Similarly, the engine cannot develop negative
torque.
(b) Make a table of some steady operating conditions for f
o
/f
m
in the
range of 0.05–0.95, in which you show: advance ratio seen by the
propeller, engine torque, propeller rotational speed, engine rota
tional speed, and vessel linear speed.
Plot the pairs [f
o
/f
m
, n
eo
/n
m
] on the characteristic plot from Part
1, and assess whether the vessel, engine, and propeller are well
matched to operate over a range of f
o
.
(c) Construct a linear approximation for the plant dynamics around
the operating condition f
o
/f
m
= 0.8. If the plant input is δf, and
the output is δu, list the A, B, C matrices for your plant, and the
eigenvalues. δ here indicates the perturbed value from the steady
state, e.g., f = f
o
+ δf.
186 24 PROBLEMS
(d) Create an LTRtype controller for the plant. First, choose the ﬁlter
gains V
1
and V
2
to make the crossover frequency (where the magni
tude of L(s) = C(sI−A)
−1
H passes through unity) about 0.6rad/s.
This choice corresponds with the closedloop system being able to
overcome waves with period 10seconds or longer. Note that for this
scalar design, you will be able to move the curve [L(s)[ up or down
on the Bode plot, but you cannot easily change its shape, or move
it sidetoside.
Secondly, recover this loop shape with an LQGtype controller, us
ing a small control penalty.
You should prepare for this part: a plot of log([L(s)[) versus log(ω),
a listing of your V
1
and V
2
choices, and the A, B, C matrices of your
LQG compensator.
(e) In the event of a current or wind disturbance, the actual vessel
speed will vary from the nominal value. This is not acceptable in
the long term, given that the vessel is laying out a cable.
An integral action can be added quite easily in the LTR design
technique. First, add an integrator to the input of the plant, so
that an “augmented plant” is created. This augmented plant has
one more state than the original plant:
A
aug
=
¸
A B
0 0
¸
B
aug
= [0 1]
T
C
aug
= [C 0]
The idea is to then carry out the KF design and LTR as before, with
the augmented plant model. When this is done, move the integrator
from the augmented plant into the compensator. The addition of an
integrator channel to the plant or compensator can be accomplished
with the command sysPaug = sysP * tf([0 1],[1 0]) ;, for ex
ample.
Present a ﬁgure of [L(s)[ and the other quantities as you gave them
in Part 4. This is to be kept as a separate controller design.
(f) For the purpose of demonstrating your controllers, construct a non
linear simulation of the plant system, with input f. You don’t
187
need to show anything for this task; it is mostly to verify that
the steady conditions you calculated above, with f = 0.8f
m
, are
actually steady conditions from the point of view of the whole sim
ulation. You may wish to use global variables in Matlab so that
you don’t have to type everything in more than once.
(g) Augment your simulation above with the compensator dynamics. If
the compensator has states z, and your simulation has states [u, n
p
],
make a new state vector [u, n
p
, z]. You will propagate the ﬁrst two
states as usual with the nonlinear equations and the fuel rate f as
input. The compensator states z are propagated by a set of system
matrices [A
c
, B
c
, C
c
], which you just designed. Note that the input
to the compensator is an error signal: e = r − u, where u is the
speed of the vessel and r is a reference speed. The output of the
compensator is then fuel rate f.
We are interested mainly in disturbance rejection, for which you
will just set r = u
o
, the steady condition speed. Recall that you are
implementing a controller for operation about a nominal condition;
whenever your compensator generates f from the error signal, be
sure to then add on 0.8f
m
, the part needed to keep us near the
steady condition.
Demonstrate four properties, showing timedomain simulations for
each:
• The ﬁrst controller, without integral action, does not com
pletely reject steady disturbances. I suggest you implement
a disturbance as an acceleration:
uprime = uprime + k. In the absence of a controller, this
line would drive the system with an additional acceleration of
km/s
2
, thus modelling the eﬀects of steady wind or current.
This line is to be added after computing the usual parts of
uprime.
• The ﬁrst controller does not reject high frequency disturbances.
A suggestion is to let
uprime = uprime + k*sin(10*omega_c*t), where omega_c is
the crossover frequency. This adds an oscillating acceleration
at ten times the crossover frequency.
• The ﬁrst controller does, however, reject some lower frequency
disturbances. For example, try
188 24 PROBLEMS
uprime = uprime + k*sin(omega_c/10*t)/100. Note the ex
tra factor of 100 is included so that the result can be compared
directly with the highfrequency disturbance case above. The
eﬀect of an acceleration disturbance onto the observed output,
velocity u, scales with the inverse of the frequency.
• The second controller, with integral action, rejects steady dis
turbances. Use the same approach as above for steady distur
bances
(h) What is the propeller response during the highfrequency distur
bances? Is it reasonable or do we need to slow down the controller
bandwidth?
These notes were developed in the instruction of the MIT graduate subject 13.49: Maneuvering and Control of Surface and Underwater Vehicles. We plan many enhancements; your comments are welcome!
Maneuvering and Control of Marine Vehicles Latest Revision: January 11, 2002 c Michael S. Triantafyllou and Franz S. Hover
. . . . . . . . 1. . . .2 Linear Momentum in a Moving Frame . . . .6 Determinant . . . .3 Laplace Transform . . . . . . . 2. . . . . . . . . . . . . . . . . . . .2 Convergence . . . . . . .1 Deﬁnition . . . . . . . . . . . . . . . . 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Contents 1 MATH FACTS 1. . 3. . . . . . . . . . . . .3. . . . . . . . . . . . . . . . . . . . . . . . .11 Singular Value . . . .2. . . . .2 Vector Magnitude . . .2. . . . . . . . . . . . . 1. . . . . 1.4 Common Matrices . .8 Trace . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1. .4 Dead Reckoning .2 Rotating Frame Attached to Pivot Point i . . . . . . . .2.2. 1. . . .3 Rate of Change of Euler Angles .2 Matrices . . . . . . . . 1. . . . . . .1 Deﬁnition . . . 2. . . .2. . . . . . . 3. . . 1. . . .1. . . . . . . . . . . . . . . . . . . . . . .2. . . .3. . . .2 Multiplying a Vector by a Matrix . . . . . . . . . . 1.3. . . . .3 Multiplying a Matrix by a Matrix . . 1. . . . . . . . 2. . . . . . . . . . 3. 1. . . . . . . . . . . . . . . . .1.2. . . . . . .1. . . . . . . . . . . .2. . . . . . . . . . . . . . 1.2 Diﬀerential Rotations . . . . . . . . . . .4 Vector Cross Product . . . . . . . . . . . .5 Transpose . . . . . . . . . . . Transform 1 1 1 2 2 2 3 3 3 4 4 5 5 6 7 7 9 9 11 11 11 11 13 14 14 16 18 19 20 20 21 22 23 23 . . . . . . . . . . . . .4 Solution of Diﬀerential Equations by Laplace 2 KINEMATICS OF MOVING FRAMES 2. . . . . . . . . . . .2. . . . . . . . . . . . . . .3.2.9 Eigenvalues and Eigenvectors . . . . . . . . . . .1. . . . . . . . . . . . . 1.1 Rotation of Reference Frames . . . . 1. . . .1 Momentum of a Particle . . . .3 Vector Dot Product .7 Inverse . . . .3 Convolution Theorem . 3 VESSEL INERTIAL DYNAMICS 3. . . . . . . . . . . . . . . . 1. . . . . . . .1 Deﬁnition . . . .1 Moving Frame Aﬃxed to Mass . . . . . . . . . . . . . . . .10 Modal Decomposition .2. . . . . . . . 1. . . . . . . . . 1.3 Example: Mass on a String . . . . . . . 1. 1. . . . . . . . . . . . . . . . . . . . . . 3. . . . . . . . . . . . . . . . . 1. .1 Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1. . . . . .3. . . . . . . . . . . . . . . . . . . . .3. . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . .6 Summary of the Linear Maneuvering Model . . . .1 Dieudonn´ Spiral .1 Towtank . . . . . 3. . . . . . . . . .7 3. 24 24 26 26 27 27 28 28 30 30 31 32 34 36 36 37 37 37 38 38 40 40 42 44 46 48 48 48 49 52 52 52 53 53 4 HYDRODYNAMICS: INTRODUCTION 4. . . . 5 SIMILITUDE 5. . .7 Stability in the Vertical Plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4. . . . .5. . . . . .1 Taylor Series and Hydrodynamic Coeﬃcients . . . .3 PlanarMotion Mechanism . . . .4.2 Rotating Arm Device . . . . . . . . .5. . . . . . .1 Adding Yaw Damping through Feedback 4.1 Drift Angle . . . . . . . . .5. . .4 Basic Rudder Action in the Sway/Yaw Model . . . . . . Basis for Simulation . . . . . . . . . . . . . . . . . Example: Spinning Book 3. . . . . . . . . . . . . . . . . . . . .4 Roll Equation Similitude . . . . . . . .2 ZigZag Maneuver . . . . 4. . . . . . . . . Parallel Axis Theorem .1 Phase 1: Accelerations Dominate . . . . . . . . . . . . . . . . . . . . 4. . . . . . . . . . . . . . .2 Surface Vessel Linear Model . . 5. . . . . . . . . . . . . . . . . . . . 6. . . . . . . . . . . . . . . . . . . . . . . . .1 xaxis . . . . . . . . . . . . 4. . . . . . .6 3. . . .3. . . . . . e 7. .5 Response of the Vessel to Step Rudder Input . . . . . . . .5. . . .3 Stationary Frame Angular Momentum . . . . . . . .3 zaxis . . . . . . . . . .3 Similitude in Maneuvering . . . . . . . . . . . . . .5 3. . . . . . . . . . ii . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4 3. . . . .3 Circle Maneuver . . . . . . . . . . . . . 7 STANDARD MANEUVERING TESTS 7. 4. . . . . . . . . . . . 4. . . . . . . . . 6 CAPTIVE MEASUREMENTS 6. . . . 4. . . . . . . . . . . . . . . .3. . . . . . . . 3. . . .2 Common Groups in Marine Engineering 5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7. . . . . . . . . . . . . . . . . . . . . . . . .2 Phase 3: Steady State . . .3. . . . . . . . . . . . . . . .1 Use of Nondimensional Groups . . . . . . . . . . .5. . . . . 6. . . . .2 yaxis . . . . . . . 4. .2 Heading Control in the Sway/Yaw Model 4. 7. . . . . . . .3 Stability of the Sway/Yaw System . . . 5. . . . .4. . . . . . .
. . . . . . . . . . . . . .2 Kinematics Following the Fluid . . . . . . . . . . . . . .4 SlenderBody Theory vs. . . . 11 FINS AND LIFTING SURFACES 11. . . . . . . . . . . . . 8. . . . . . . . . . . . .2 Steady Propulsion of Vessels . . . . . . . .6 Aggregate Eﬀects of Body and Fins . . . . . . .6 Total Moment on a Vessel . . . . .5 SlenderBody Approximation for Fin Lift . . . . iii . . . . . . 9. . . . . . . . . . . . . . 10. . . . . . . . . . 11. . . . Mw . . . . .7 Relation to Wing Lift . .1 Characteristics of LiftProducing Mechanisms 10. . . . . . . . . . . . . . . . . . .2 Jorgensen’s Formulas . . .3. . . . . . . . . . . . . . . . . . . . . . .5 Role of Fins in Moving the Aerodynamic Center . . . . . . . . . . . . . . . . . 12 PROPELLERS AND PROPULSION 12. . . . . . . . . . .2. . . . . . . . 9. . . 11. . . . . . 9. . . . . . . . . . . . . . . . .7 Coeﬃcients Zw .7. . . . . . . . 8. . . . . .3. . . . . . . . 9. . . . . . .1 Basic Characteristics . Zq . . . .1 Introduction . . . . . . . .3 7. . . . .3 Separation Moment . . . . 53 53 54 56 56 56 57 57 58 60 60 62 62 62 63 64 64 65 67 67 68 68 68 70 71 73 74 74 74 75 77 77 77 77 8 STREAMLINED BODIES 8. . . . . . . . . .3 Ring Fins . . .8 Convention: Hydrodynamic Mass Matrix A . . . . 8. . . . .1 Origin of Lift . . . . . . . . . . . . . 10. 9. . . 9. . . . . . . . . . . . . . .3. . . . . . . . . Heeling in Submarines with Sails . . . . . . . . . . . . . . . . . . . . . . . Heel Angle . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4 Diﬀerential Force on the Body . . . . . . . . . . . . . . . . . 8. . . . and Mq for a Slender Body 9 SLENDERBODY THEORY 9. . 12. . . 10. .2 ThreeDimensional Eﬀects: Finite Length . . . . . . . . . .4 Speed Loss . . . . . . . . . . . 8. . . . . . . . . . . 12. . . . . .2 Munk Moment . . . . . . . .2 7. . . . . . . . . . . . . . . . . . . .4 Net Eﬀects: Aerodynamic Center . . . 9. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Experiment . . . . . . . .3 Derivative Following the Fluid . . . . . . . . . . . . . . . . . . . . . . . . .1 Introduction . . . .1 Nominal Drag Force . .3 Hoerner’s Data: Notation . . 8. . .5 Total Force on a Vessel . . . . 10 PRACTICAL LIFT CALCULATIONS 10. . . . . . . . . . . .
.2 The Need for Modeling . . . . . . . . . . . . . 14. .2. . . . . . . . . 101 15. . . . . . . . . . . . . . . . . .1 Introduction . . . .12. . . . . . . . . . . . . . . . . . . .5 Example: Partial Fractions with Complex Poles . . . 101 15. . .2 Damped Axial Motion 13.1 Proportional Only . . .4 Example: PID Control . . Inputs. . . . . . . . . . .2. .1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1. . . . . . . . . . . . . .3 Example: Partial Fractions with Unique Real Poles 14. . . . . . . . . . . . . . . . . . . . . . . . . . 100 15. . . . . 14. . . . . . 100 15. . . . . . . . . . . . . . 100 15. . . . 12. . . 13. .2.2 Solution for Steady Conditions 12. . . . . . . 13. . . . . . . . . . . . . . . . . . . . . . 12. 81 81 83 83 84 85 85 85 87 89 89 91 94 94 96 96 96 97 98 98 98 99 99 . . . . . . . . . . . . . . . . . . . . . . . . . . . .6 Stability in Linear Systems . . .1 Standard StateSpace Form . 13. .3 Cable Strumming . . . . . . . . . . . . . . .1 Partial Fractions . . . . . . .1. . . . . .3. . . . . 12.2 Partial Fractions: Unique Poles . . .2.2 TwoState Model: Healey et al. . . . . . . . . . . 14. . .1 OneState Model: Yoerger et al. . . . 13. . . . . . . . . .3 PID Controllers . . . . . . . . . . . . . . . . . . .2. . . . .3 Nonlinear Control . .2. . . . . . 13.1. . . . . . . . . .1 Plants. . .1 Derivation . . . . . . . . . . . . .1. . . . . . . . . .3. 103 15. . . 13 TOWING OF VEHICLES 13. . . . .1 Force Balance . . . . . . . . . . . . . . . . . . . . . . .2 Linearized Dynamics . . . . 103 15. .2 Critical Angle . . . . .3 Converting a Transfer Function into a StateSpace Model 102 15. . . . .3 Engine/Motor Models . . . . . . 14. . . . . . . . . . . . . . .8 General Stability . . 14 TRANSFER FUNCTIONS & STABILITY 14. . . . . 101 15. . . . . . . . . . . . . 104 iv . . . . . . . . . . . . . . .4. . . . . . . 13. . . . . . . . . . . 14. . . . . . . . . . . . . . .2. . .4 Partial Fractions: ComplexConjugate Poles . 15 CONTROL FUNDAMENTALS 100 15. . . . . . . . . . . . . . . . . . . . . . . . . .2 Representing Linear Systems . . . . . . . . .1 Statics . 14.2 Converting a StateSpace Model into a Transfer Function 102 15. . . . . . . . . . . and Outputs . . . . . . . . . . .4 Vehicle Design . . . . . . . . . . . . . .3 Unsteady Propulsion Models . . . . . .7 Stability ⇐⇒ Poles in LHP . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16. . . . . . .4. . . 104 105 105 106 106 106 107 109 109 109 109 109 110 111 112 113 113 113 114 114 115 116 116 118 118 119 121 121 121 121 123 125 16 MODAL ANALYSIS 16. . . . . . . . . . . . . . .7 The Recipe for Loopshaping . .5 Robust Performance . 17. . . . . . .3 Primary Transfer Functions . . . . 15. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16. . . 17. . . . 17. . . . . . . . . . . . . . . . . . .15. . . . . . . .1 Introduction . . . . . . . . . . . . . . . . . 17. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 Robustness on the Nyquist Plot . . 16. . . . . . . . . . . . . . 16. . . . . . .2. . . . . . . . . . . .5 Heuristic Tuning . . . . . . . . . 17. . 17 CONTROL SYSTEMS – LOOPSHAPING 17. . . . . .1 Mapping Theorem . . . . . . . . . . . . 18. . . . . . . . . . .5 Properties and Use of the LQR . . . . . . . . 17. . . . . . . . . . . . . . . . .6. . .2. . . . . . . .2 Block Diagrams: General Case . . . . . . . . . . . . .1 Introduction . . .2.4 Plant Output and Observability . . . . .2 ProportionalDerivative Only .2. 15. 17. . . . .6. . . . . . . . . . . . . . . . . . . . . . .4. . . . . .2. . . . . . .3 Design for Nominal Performance . . . . . . . . . . .3 Forced Response and Controllability .2 Roots of Stability – Nyquist Criterion . .6 Implications of Bode’s Integral . . . . . . . . . .2 FullState Feedback . . . .3 Dynamic Programming . . . . . . . . . . . . . . . . . . 17. . . . . . . 15. . . . . . . 18. . . . . . . . 18. . .2 Nyquist Criterion . . . . 17. . . . . . . . . . . . . . . v . . . . .6 Block Diagrams of Systems . . .3 ProportionalIntegralDerivative 15. . . . . . . 18 LINEAR QUADRATIC REGULATOR 18. . .4 Design for Robustness . 15. . . . . . . . . . . . . . . . . . . . . . . . . .4 Dynamic Programming and FullState Feedback 18. . . .2 Modal Canonical Form . . . 15. . .6. . . . . .1 Deﬁnition . . . . . . . .3 Modal Decomposition of Response 16. . . . . . . . . .2. . . . . .2 Matrix Exponential . . . . . . . . . . . .1 Introduction . . . . . . . .1 Fundamental Feedback Loop . 16. . . . . . .
. . .7.2 Problem Statement . . . 21. .5 TimeDomain Simulation . . . . . . . . . . . . . . . . . . . 21. . . . . . . . . . . . 20. . . . . . .1 Introduction . . . . . . . . . . . . 19. . . .6 Combination of LQR and KF . . . . . . . . . . . 153 22. . . . . . . . . . .4. . . . .5 Properties of the Solution . 22 CARTESIAN NAVIGATION 153 22.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . 19. . . . . . . . . . . . . . . . . . . . 19.4 Step 2: H as a Function of Σ . . . . . . . . . . . . .19 KALMAN FILTER 19. . . . . . . . . 128 128 128 129 131 132 133 134 134 135 135 136 137 137 138 139 141 141 143 143 143 145 146 147 148 150 21 SYSTEM IDENTIFICATION 21. . . . . . . . . . . . 155 23 REFERENCES 24 PROBLEMS 159 163 vi . . . . . . .7. . . .7. ∂ 19. . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Visual Output from a Simple Input . . . . . . . . . . . .7 Proofs of the Intermediate Results . . . . . . . . . . . . . . . . . . . . . . 20 LOOP TRANSFER RECOVERY 20. 19. . . . . . . .5 Three Lemmas . . . . 19. ∂ 19. . . . . .2 Global Positioning System (GPS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Proof that E(eT W e) = trace(ΣW ) . . . . . . . . . .1 Introduction . . . . . .4. . . . . . . 20. . . . . . . . . .4 Transfer Function Estimation – Broadband Input 21.4 Usage of the Loop Transfer Recovery . . . . . . . . . . .3 Step 1: An Equation for Σ . . . 21. . .7. . . . .1 Fourier Transform of Sampled Data . . . . 21.1 Acoustic Navigation . . . . . .2 A Special Property of the LQR Solution 20. . . . . . . . . . . .2 Estimating the Transfer Function . . . . . . . . . . . . . . . . . . . . . .3 Proof that ∂H trace(−ΛΣC T H T ) = −ΛΣC T 19. . 21. . 20. . ˙ 19. . . . .3 The Loop Transfer Recovery Result . . . . . . . .3 Transfer Function Estimation – Sinusoidal Input . . . . . . .2 Proof that ∂H trace(−ΛHCΣ) = −ΛT ΣC T . . . .4 Proof of the Separation Principle . . 19. . . . .
.
.
1 1. Vector addition is pointwise.1. which is particularly easy to see in two or three dimensions. addition is stringing the vectors together head to tail. z 7 a 1 2 x y 1. = 7 2 9 3 a+b = c 5 Graphically. in threespace. we write 2 a = 1 . a number with a direction.1 MATH FACTS Vectors Deﬁnition We use the overhead arrow to denote a column vector.e. . For example. 2 1 + 3 4 .1 1 1.. 7 The elements of a vector have a graphical interpretation. i.
1. is given by m x = i=1 x2 . 2.1. The dot product also satisﬁes x · y = xy cos θ.1. 1. 7 −14 1.3 Vector Dot Product The dot product of two vectors is the sum of the products of the elements: m x · y = xT y = i=1 xi yi .2 2. where θ is the angle between the vectors. x×y = z. .2 Vector Magnitude 2 The total length of a vector of dimension m. its Euclidean norm. i this scalar is commonly used to normalize a vector to length one. whose 1. direction is given by the righthand rule. direction is normal to the plane formed by the two vectors.4 Vector Cross Product The cross product of two threedimensional vectors is another vector.1. Scalar multiplication is pointwise. rotating from x to y. 1 MATH FACTS −4 −2 × 1 = −2 .
c will have as many columns as A has rows (m). The schoolbook formula is x2 y3 − x3 y2 x×y = x3 y1 − x1 y3 . is equivalent to a set of row vectors.2. say 2 3 A = [a b] = 1 3 . then the multiplication c = Av is constructed as: n ci = Ai1 v1 + Ai2 v2 + · · · + Ain vn = j=1 Aij vj . 1.1 Matrices Deﬁnition A matrix. measured from x to y. x1 y2 − x2 y1 1. We usually use an uppercase symbol to denote a matrix.2.2 Multiplying a Vector by a Matrix If Aij denotes the element of matrix A in the i’th row and the j’th column. (signed) magnitude is equal to xy sin θ.2 Matrices 3 3. or array. where θ is the angle between the two vectors. a vector is a special case of a matrix with one column. arranged side by side. permit pointwise addition and scalar multiplication.2 1. magnitude is the area of the parallelogram formed by the two vectors – the cross product of two parallel vectors is zero – and 4. Note that this multiplication is welldeﬁned only if v has . like vectors. Matrices.1. where n is the number of columns in A. 7 2 This matrix has three rows (m = 3) and two columns (n = 2).
. 0 0 1 I3×3 The identity always satisﬁes AIn×n = Im×m A = A. 1. where k is the number of columns in matrix B. Associative. The same inner dimension condition applies as noted above: the number of columns in A must equal the number of rows in B. The product vA would be wellposed only if A had one row.3 Multiplying a Matrix by a Matrix The multiplication C = AB is equivalent to a sidebyside arrangement of column vectors C:j = AB:j . so that each A:j is the j’th column vector.2.g. they have consistent inner dimension n.. 1 0 0 = 0 1 0 . and comprises a square matrix with ones on the diagonal.4 Common Matrices Identity . A(B + C) = AB + AC. NOT Commutative. There is another important interpretation of this vector multiplication: Let the subscript : indicate all rows. AB = BA. and the proper number of columns.4 1 MATH FACTS as many rows as A has columns. Then c = Av = A:1 v1 + A:2 v2 + · · · + A:n vn . e. Distributive. except in special cases. Matrix multiplication is: 1. (B + C)A = BA + CA. 3. (AB)C = A(BC). so that C = AB = [AB:1 AB:2 · · · AB:k ]. 2. and zeros elsewhere. 1. We are multiplying column vectors of A by the scalar elements of v.2. The identity matrix is usually denoted I.
and in this case the operation is commutative.2. 0 0 3 The product of a diagonal matrix with another diagonal matrix is diagonal. and illustrates the principle of volume: det(A) = A11 A22 − A21 A12 1 −1 = 1 + 1 = 2. it is equivalent to “ﬂipping” the vector or matrix around the diagonal line. the following is a diagonal matrix: 4 0 0 A = 0 −2 0 . For instance. and has all zeros oﬀ the diagonal.5 Transpose The transpose of a vector or matrix.6 Determinant The determinant of a square matrix A is a scalar equal to the volume of the parallelepiped enclosed by the constituent vectors. 1. The twodimensional case is particularly easy to remember. A very useful property of the transpose is (AB)T = B T AT .2.1. 1 1 det . For example. A diagonal matrix is square.2 Matrices 5 Diagonal Matrices . 1 a= 2 −→ a T = {1 2 3} 1 4 8 2 5 9 A = 4 5 −→ AT = 3 1 2 8 9 . indicated by a T superscript results from simply swapping the rowcolumn indices of each entry. 1.
T . and this results from the fact that the constituent vectors do not span the matrix dimension. The general formula allows one to pick a row k. Its computation requires the determinant above. and apply j=n det(A) = j=1 Akj (−1)k+j ∆kj . in the sense that one could also target the k’th column: j=n det(A) = j=1 Ajk (−1)k+j ∆jk . in three dimensions. a singular matrix has all its vectors lying in a (twodimensional) plane. in two dimensions. For instance. If the determinant of a matrix is zero. Note also that det(A) = det(AT ). where ∆kj is the determinant of the submatrix formed by neglecting the k’th row and the j’th column. denoted A−1 . the determinant is more complicated to compute. The formula is symmetric.7 Inverse The inverse of a square matrix A. a singular matrix has the vectors colinear.2. and the following deﬁnition of the n × n adjoint matrix: (−1)1+1 ∆11 · · · (−1)1+n ∆1n ··· ··· ··· adj(A) = .6 y Area = det(A) = 2 A:2 1 A:1 x 1 1 MATH FACTS In higher dimensions. then the matrix is said to be nonsingular. satisﬁes AA−1 = A−1 A = I. 1. If det(A) = 0. then the matrix is said to be singular – there is no volume. perhaps the one containing the most zeros. n+1 n+n (−1) ∆n1 · · · (−1) ∆nn .
2. det(A) = 0. .2 Matrices Once this computation is made. then the inverse does not exist.e. the inverse follows from A−1 = adj(A) . Since the above is equivalent to (A − λI)x = 0. 1. We ask for what nonzero vectors x (right eigenvectors). The inverse ﬁnds common application in solving systems of linear equations such as Ax = b −→ x = A−1 b.9 Eigenvalues and Eigenvectors A typical eigenvalue problem is stated as Ax = λx. det(A) 7 If A is singular. it is clear that det(A − λI) = 0. and λ is a scalar.1.2. x is a column vector with n elements. i.. 1. where A is an n × n matrix. here is an example for the twodimensional case: 4 −5 2 −3 A = A − λI = −→ 4−λ −5 −→ 2 −3 − λ det(A − λI) = (4 − λ)(−3 − λ) + 10 = λ2 − λ − 2 = (λ + 1)(λ − 2). This observation leads to the solutions for λ.8 Trace The trace of a matrix is simply the sum of the diagonals: n tr(A) = i=1 Aii . and scalars λ (eigenvalues) will the equation be satisﬁed.
(A − λ1 I)x1 = 0 −→ 5 −5 x1 = 0 −→ 2 −2 √ √ x1 = 2/2. 2 29/29 T . 2/2 T . . Each is associated with a right eigenvector x. generated from the equation Ax = λx. A and AT share the same eigenvalues λ. Left eigenvectors. In this example. (A − λ2 I)x2 = 0 −→ 2 −5 x2 = 0 −→ 2 −5 √ √ x2 = 5 29/29. λ1 = −1 and λ2 = 2. A has two eigenvalues. −5 29/29 (AT − λ2 I)y2 = 0 −→ 2 2 y2 = 0 −→ −5 −5 √ √ y2 = 2/2. and positive ﬁrst element (as above).8 1 MATH FACTS Thus. Eigenvectors have arbitrary magnitude and sign. they are often normalized to have unity magnitude. The condition that rank(A − λi I) = rank(A)−1 indicates that there is only one eigenvector for the eigenvalue λi . − 2/2 T . also useful for many problems. The above discussion relates only the right eigenvectors. T . pertain to the transpose of A: AT y = λy. then there are multiple unique eigenvectors that go with λi . Example: (AT − λ1 I)y1 = 0 −→ 5 2 y1 = 0 −→ −5 −2 √ √ y1 = 2 29/29. A set of n eigenvectors is always linearly independent. If the lefthand side is less than this. since they share the same determinant.
1.2 Matrices 1.2.10 Modal Decomposition
9
The right and left eigenvectors of a particular eigenvalue have unity dot product, that is xT yi = 1, with the normalization noted above. The dot product i of a left eigenvector with the right eigenvector of a diﬀerent eigenvalue is zero. Thus, if X = [x1 · · · xn ] , and Y = [y1 · · · yn ] , then we have Y T X = I, or Y T = X −1 . Next, construct a diagonal matrix of eigenvalues:
λ1 · 0
0 λn
Λ= it follows that
;
AV = V Λ −→ A = V ΛW T
n
=
i=1
T λi vi wi .
Hence A can be written as a sum of modal components.1 1.2.11 Singular Value
Let G(s) be an m × n, possibly complex matrix. The singular value decomposition (SVD) computes three matrices satisfying
By carrying out successive multiplications, it can be shown that Ak has its eigenvalues at λk , and keeps the same eigenvectors as A. i
1
10
1 MATH FACTS
G = U ΣV ∗ , where U is m × m, Σ is m × n, and V is n × n. The star notation indicates a complexconjugate transpose. The matrix Σ is diagonal, with the form
Σ=
σ1 0 0 0
0 0 · 0 0 σp 0 0
0 0 0 0
,
where p = min(m, n). Each nonzero entry on the diagonal is a real, positive singular value, ordered such that σ1 > σ2 > · · · σp . The notation is common that σ1 = σ, the maximum singular value, and σp = σ, the minimum singular value. The auxiliary matrices U and V are unitary, i.e., they satisfy X ∗ = X −1 . Like eigenvalues, the singular values of G are related to projections. σi represents the Euclidean size of the matrix G along the i’th singular vector:
σ = maxx=1 Gx σ = minx=1 Gx. Other properties of the singular value include: • σ(AB) ≤ σ(A)σ(B). • σ(A) = • σ(A) = λmax (A∗ A). λmin (A∗ A).
• σ(A) = 1/σ(A−1 ). • σ(A) = 1/σ(A−1 ).
1.3 Laplace Transform
11
1.3
1.3.1
Laplace Transform
Deﬁnition
The Laplace transform converts timedomain signals into a frequencydomain equivalent. The signal y(t) has transform Y (s) deﬁned as follows:
∞ 0
Y (s) = L(y(t)) =
y(τ )e−sτ dτ,
where s is an unspeciﬁed complex number; Y (s) is considered to be complex as a result. Note that the Laplace transform is linear, and so it is is distributive: L(x(t) + y(t)) = L(x(t)) + L(y(t)) will hold throughout. The following table gives a list of some useful transform pairs and other properties, for reference. The last two properties are of special importance: for control system design, the diﬀerentiation of a signal is equivalent to multiplication of its Laplace transform by s; integration of a signal is equivalent to division by s. The other terms that arise will cancel if y(0) = 0, or if y(0) is ﬁnite, so they are usually ignored. 1.3.2 Convergence
We note ﬁrst that the value of s aﬀects the convergence of the integral. For instance, if y(t) = et , then the integral converges only for Re(s) > 1, since the integrand is e1−s in this case. Convergence issues are not a problem for evaluation of the Laplace transform, however, because of analytic continuation. This result from complex analysis holds that if two complex functions are equal on some arc (or line) in the complex plane, then they are equivalent everywhere. This fact allows us to always pick a value of s for which the integral above converges, and then by extension infer the existence of the general transform. 1.3.3 Convolution Theorem
One of the main points of the Laplace transform is the ease of dealing with dynamic systems. As with the Fourier transform, the convolution of two signals in the time domain corresponds with the multiplication of signals in the frequency domain. Consider a system whose impulse response is g(t), being
12
1 MATH FACTS
y(t) ←→ Y (s) (Impulse) δ(t) ←→ 1 1 (Unit Step) 1(t) ←→ s 1 (Unit Ramp) t ←→ 2 s 1 −αt e ←→ s+α ω sin ωt ←→ 2 s + ω2 s cos ωt ←→ 2 s + ω2 ω e−αt sin ωt ←→ (s + α)2 + ω 2 s+α e−αt cos ωt ←→ (s + α)2 + ω 2 1 1 (e−at − e−bt ) ←→ b−a (s + a)(s + b) 1 1 1+ (be−at − ae−bt ) ab a−b √ 1− √ ←→ 1 s(s + a)(s + b)
2 ωn ωn e−ζωn t sin ωn 1 − ζ 2 t ←→ 2 2 s + 2ζωn s + ωn 1 − ζ2
1 e−ζωn t sin ωn 1 − ζ 2 t + φ 2 1−ζ √ 1 − ζ2 −1 φ = tan ζ (Pure Delay)
←→
2 ωn 2 s(s2 + 2ζωn s + ωn )
y(t − τ )1(t − τ ) ←→ Y (s)e−sτ dy(t) (Time Derivative) ←→ sY (s) − y(0) dt 0+ t Y (s) 0− y(t)dt y(τ )dτ ←→ + (Time Integral) s s 0
1. Generate the timedomain response from the simple transform pairs.4 Solution of Diﬀerential Equations by Laplace Transform The Convolution Theorem allows one to solve (linear timeinvariant) diﬀerential equations in the following way: 1. using the transform pairs.3. . Apply time delay as necessary. Perform the multiplication in the Laplace domain to ﬁnd Y (s). Ignoring the eﬀects of pure time delays. 2. break Y (s) into partial fractions with no powers of s greater than 2 in the denominator. then y(t) represents the output of this system when it is driven by the external signal x(t). The Convolution Theorem is t y(t) = 0 g(t − τ )x(τ )dτ ⇐⇒ Y (s) = G(s)X(s). Here’s the proof given by Siebert: Y (s) = 0 ∞ ∞ 0 0 ∞ 0 0 ∞ 0 ∞ 0 y(t)e−st dt t = = = = g(t − τ ) x(τ ) dτ e−st dt g(t − τ ) 1(t − τ ) x(τ ) dτ e−st dt ∞ 0 ∞ x(τ ) g(t − τ ) 1(t − τ ) e−st dt dτ x(τ ) G(s)e−sτ dτ = G(s)X(s) When g(t) is the impulse response of a dynamic system. the output is y(t) = g(t)∗x(t). 4.3 Laplace Transform 13 driven by an input signal x(t). 1. 3. Speciﬁc examples of this procedure are given in a later section on transfer functions. Transform the system impulse response g(t) into G(s). and the input signal x(t) into X(s).
which we now derive through successive rotations of the three . We will develop the equations of motion using Euler angles. and yaw. For example. some of them might use the same axis twice. and has a solid link with the intuitive notions of roll. but that the body frame has a diﬀerent angular orientation. then about the new pitch axis. rotate the movable system about its yaw axis. pitch. it is easy to see that any possible orientation can be achieved. consider the sequence of [yaw.14 2 KINEMATICS OF MOVING FRAMES 2 2. roll]: starting from an orientation identical to some inertial frame. Successively rotating the system about three of its own principle axes.1 KINEMATICS OF MOVING FRAMES Rotation of Reference Frames We say that a vector expressed in the inertial frame has coordinates x. we assume that the origins of these frames are coincident. then about the newer still roll axis. The former method involves successive rotations about the principle axes. there are many valid Euler angle rotation sets possible to reach a given orientation. z=z’ z’’ z’’’ x’ y’’’ y y’=y’’ x x’’=x’’’ Figure 1: Successive application of three Euler angles transforms the original coordinate frame into an arbitrary orientation. and in a bodyreference frame xb . pitch. Needless to say. including the Euler angles and the Euler parameters (quaternions). Tape three pencils together to form a righthanded threedimensional coordinate system. A ﬁrst question is: what is the coordinate of a point ﬁxed in inertial space. For the moment. The angular orientation has several wellknown descriptions. but with more abstraction. Quaternions present a more elegant and robust method. referenced to a rotated body frame? The transformation takes the form of a 3×3 matrix.
1 Rotation of Reference Frames 15 Euler angles.. and to the order in which we performed the rotations. xb = 0 0 − sin ψ cos ψ (2) (3) This represents the location of the original point. as written.2. xb = sin θ 0 cos θ Finally. Now rotate the movable frame yaw axis (z) through an angle φ. ψ). (4) All of the transformation matrices. i. rotate the body system an angle ψ about its newest xaxis: 1 0 0 3 cos ψ sin ψ0 xb2 = R(ψ)xb2 . we should note that the rotation matrix R is universal to all representations of orientation. Additionally.e. pitch about the new yaxis by the angle θ: cos θ 0 − sin θ 2 0 1 0 xb1 = R(θ)xb1 . We will use the notation xb instead of xb3 from here on. xb3 . we have . Now apply the second rotation. We have cos φ sin φ 0 xb1 = − sin φ cos φ 0 xb0 = R(φ)xb0 . 0 0 1 (1) Rotation about the zaxis does not change the zcoordinate of the point. in the fullytransformed bodyreference frame. θ. the bodyreferenced coordinate matches that of the inertial frame: xb0 = x. Before the ﬁrst rotation. are orthonormal: their inverse is equivalent to their transpose. The three independent rotations can be cascaded through matrix multiplication (order matters!): xb = R(ψ)R(θ)R(φ)x cθcφ cθsφ −sθ cψcφ + sψsθsφ sψcθ x = −cψsψ + sψsθcφ sψsφ + cψsθcφ −sψcφ + cψsθsψ cψcθ = R(φ. The roles of the trigonometric functions. including quaternions. ψ)x. In the case that the movable (body) reference frame has a diﬀerent origin than the inertial frame. θ. the other axes are modiﬁed according to basic trigonometry. are speciﬁc to Euler angles. including R(φ.
Small rotations are completely decoupled. = −δφ δθ −δψ 0 (6) R comprises the identity plus a part equal to the (negative) crossproduct operator [−δ E×]. (5) where x0 is the location of the moving origin. instead of in inertial space. xb = x − δ E × x x = xb + δ E × xb . z]. we have also R−1 = I3×3 + δ E×. calling its location in the body frame r (radius).16 2 KINEMATICS OF MOVING FRAMES x = x0 + R T xb . with respect to the ﬁrst frame as follows: x(t) = r x(t + δt) = R r = r + δ E × r. the vector of Euler angles ordered with the axes [x. Dividing by the diﬀerential time step gives T (9) .2 Diﬀerential Rotations Now consider small rotations from one frame to another. where δ E = [δψ. 2. (7) (8) We now ﬁx the point of interest on the body. the order of the small rotations does not matter. so that we can write the location of the point before and after the rotation. The diﬀerential rotations occur over a time step δt. using the small angle assumption to ignore higherorder terms gives R 1 δφ −δθ 1 δψ −δφ δθ −δψ 1 0 δφ −δθ 0 δψ + I3×3 . Since R−1 = RT . δθ. y. expressed in inertial coordinates. δφ].
(10) where the rotation rate vector ω dE/dt because the Euler angles for this inﬁnitesimal rotation are small and decoupled. plus a component . The resultant derivative is in the moving body frame. In the case that the radius vector changes with respect to the body frame. dt ∂t Finally. (15) ∂t where vo is the bodyreferenced velocity of the origin. we need an additional term: dxb ∂r =ω×r+ .2 Diﬀerential Rotations 17 δx δE = ×r δt δt = ω × r.2. such that δxb δE = ×r δt δt = ω × r. the time rate of change of a constant radius vector is the crossproduct of the rotation rate vector ω and the radius vector itself. (11) (12) On a rotating body whose origin point is ﬁxed. The total velocity of the particle is equal to the velocity of the reference frame origin. allowing the origin to move as well gives ∂r dxo dxb =ω×r+ + . dt ∂t dt This result is often written in terms of bodyreferenced velocity v: v =ω×r+ (13) (14) ∂r + vo . This same crossproduct relationship can be derived in the second frame as well: xb (t) = Rr = r − δ E × r xb (t + δt) = r.
and the ﬁnal Euler angle no additional rotations: 0 0 dθ dt ω = R(ψ)R(θ) 0 + R(ψ) + 1 0 cos ψ = 0 0 − sin ψ cos ψ cos θ Taking the inverse gives dφ dt − sin θ sin ψ cos θ 0 dψ dt dθ dt dφ dt dψ dt 0 0 (17) . the pitch and roll rotations moved the axis. because of the division by cos θ. and determine the eﬀects on the rotation vector. the Euler yaw angle (applied ﬁrst) is deﬁnitely not about the ﬁnal body yaw axis. 3π/2}.roll]. and hence this otherwise useful equation for propagating the angular orientation of a body fails when the vehicle rotates about the intermediate yaxis by ninety . we seek a mapping ω → dE/dt. The idea is to consider small changes in each Euler angle. The ﬁrst Euler angle undergoes two additional rotations. dt ∂t (16) 2. (18) Singularities exist in Γ at θ = {π/2. An important part of any simulation is the evolution of the Euler angles.18 2 KINEMATICS OF MOVING FRAMES due to rotation of this frame. with the sequence [yaw. Since the physics determine rotation rate ω. the second angle one rotation. For example. The velocity equation can be generalized to any bodyreferenced vector f : df ∂f = + ω × f. 1 sin ψ tan θ cos ψ tan θ dE cos ψ − sin ψ ω = 0 dt 0 sin ψ/ cos θ cos ψ/ cos θ = Γ(E)ω.3 Rate of Change of Euler Angles Only for the case of inﬁnitesimal Euler angles is it true that the time rate of change of the Euler angles equals the bodyreferenced rotation rate.pitch.
2.4 Dead Reckoning
19
degrees. In applications where this is a real possibility, for example in orbiting satellites and robotic arms, quaternions provide a seamless mapping. For most ocean vessels, the singularity is acceptable, as long as it is not on the yaw axis!
2.4
Dead Reckoning
The measurement of heading and longitudinal speed gives rise to one of the oldest methods of navigation: dead reckoning. Quite simply, if the estimated longitudinal speed over ground is U , and the estimated heading is φ, ignoring the lateral velocity leads to the evolution of Cartesian coordinates: x = U cos φ ˙ y = U sin φ. ˙ Needless to say, currents and vehicle sideslip will cause this to be in error. Nonetheless, some of the most remarkable feats of navigation in history have depended on dead reckoning.
20
3 VESSEL INERTIAL DYNAMICS
3
VESSEL INERTIAL DYNAMICS
We consider the rigid body dynamics with a coordinate system aﬃxed on the body. A common frame for ships, submarines, and other marine vehicles has the bodyreferenced xaxis forward, yaxis to port (left), and zaxis up. This will be the sense of our bodyreferenced coordinate system here.
3.1
Momentum of a Particle
Since the body moves with respect to an inertial frame, dynamics expressed in the bodyreferenced frame need extra attention. First, linear momentum for a particle obeys the equality d (mv) (19) dt A rigid body consists of a large number of these small particles, which can be indexed. The summations we use below can be generalized to integrals quite easily. We have F = Fi + Ri = d (mi vi ) , dt (20)
where Fi is the external force acting on the particle and Ri is the net force exerted by all the other surrounding particles (internal forces). Since the collection of particles is not driven apart by the internal forces, we must have equal and opposite internal forces such that
N
Ri = 0.
i=1
(21)
Then summing up all the particle momentum equations gives
N
Fi =
i=1
d (mi vi ) . i=1 dt
N
(22)
Note that the particle velocities are not independent, because the particles are rigidly attached. Now consider a body reference frame, with origin 0, in which the particle i resides at bodyreferenced radius vector r; the body translates and rotates, and we now consider how the momentum equation depends on this motion.
3.2 Linear Momentum in a Moving Frame
21
z, w, φ y, v, θ x, u, ψ
Figure 2: Convention for the bodyreferenced coordinate system on a vessel: x is forward, y is sway to the left, and z is heave upwards. Looking forward from the vessel bridge, roll about the x axis is positive counterclockwise, pitch about the yaxis is positive bowdown, and yaw about the zaxis is positive turning left.
3.2
Linear Momentum in a Moving Frame
The expression for total velocity may be inserted into the summed linear momentum equation to give
N
Fi =
i=1
d (mi (vo + ω × ri )) i=1 dt
N ∂vo d + ω× m i ri , ∂t dt i=1
N
(23)
= m
where m = N mi , and vi = vo +ω ×ri . Further deﬁning the center of gravity i=1 vector rG such that
N
mrG =
i=1
mi ri ,
(24)
we have
N
Fi = m
i=1
∂vo d + m (ω × rG ). ∂t dt
(25)
Using the expansion for total derivative again, the complete vector equation in body coordinates is
22
3 VESSEL INERTIAL DYNAMICS
F =
i=1
N =m
dω ∂vo + ω × vo + × rG + ω × (ω × rG ) . ∂t dt
(26)
Now we list some conventions that will be used from here on: vo = {u, v, w} (bodyreferenced velocity) rG = {xG , yG , zg } (bodyreferenced location of center of mass) ω = {p, q, r} (rotation vector, in body coordinates) F = {X, Y, Z} (external force, body coordinates). The last term in the previous equation simpliﬁes using the vector triple product identity ω × (ω × rG ) = (ω · rG )ω − (ω · ω)rG , and the resulting three linear momentum equations are ∂u dq dr + qw − rv + zG − yG + (qyG + rzG )p − (q 2 + r2 )xG (27) ∂t dt dt ∂v dr dp Y = m + ru − pw + xG − zG + (rzG + pxG )q − (r2 + p2 )yG ∂t dt dt ∂w dp dq + pv − qu + yG − xG + (pxG + qyG )r − (p2 + q 2 )zG . Z = m ∂t dt dt
X = m
Note that about half of the terms here are due to the mass center being in a diﬀerent location than the reference frame origin, i.e., rG = 0.
3.3
Example: Mass on a String
Consider a mass on a string, being swung around around in a circle at speed U , with radius r. The centrifugal force can be computed in at least three diﬀerent ways. The vector equation at the start is ∂vo dω + ω × vo + × rG + ω × (ω × rG ) . ∂t dt
F =m
0}T = {0.3 Example: Mass on a String 3. 0. 0}T . 0. 0}T = {0. with the same orientation as above but allowing it to rotate with the string. 0. r. 0}T . gives = {U. 0}T = {0. U/r}T = {0. U 2 /r.2 Rotating Frame Attached to Pivot Point Aﬃxing the moving reference frame to the pivot point of the string.3.3.3. vo ω rG ∂vo ∂t ∂ω ∂t giving the same result: F = mω × (ω × rG ) = m{0. 0. 0}T = {0. U 2 /r. The force of the string pulls in on the mass to create the circular motion. 0}T . 0. 0. . we have = {0. 0. 0}T = {0. vo ω rG ∂vo ∂t ∂ω ∂t such that F = mω × vo = m{0. 0. 0.1 Moving Frame Aﬃxed to Mass 23 Aﬃxing a reference frame on the mass. with the local x oriented forward and y inward towards the circle center. 0}T = {0. 3. 0}T . U/r}T = {0.
24 3. dt (28) where Mi is an external moment on the particle i. ∂t The second term expands as (using the triple product) mrG × ∂ω mi ri × × ri ∂t i=1 N N (29) = i=1 mi (ri · ri ) ∂ω ∂ω − · ri r i ∂t ∂t + zi2 )p − (yi q + zi r)xi ) ˙ ˙ ˙ 2 + zi )q − (xi p + zi r)yi ) ˙ ˙ ˙ 2 + yi )r − (xi p + yi q)zi ) ˙ ˙ ˙ (30) = N 2 i=1 mi ((yi N 2 i=1 mi ((xi N 2 i=1 mi ((xi . In this case. can also be used for the calculation. the summed particle equation is (Mi + ri × Fi ) = i=1 i=1 ri × d (mi vi ). U 2 /r. U sin δψ. . summed internal moments cancel.1). and the ﬁrst term becomes ∂vo + ω × vo . vector subtraction gives δv = {0. 0}T . it follows easily that in the ﬁxed frame dv/dt = {0.3. ˙ Since ψ = U/r. 3. Similar to the case for linear momentum. 0}T {0. U δψ. 0}T . The summation in the ﬁrst term of the righthand side is recognized simply as mrG . and momentarily coincident with the frame on the mass (3. as before.4 Angular Momentum N N For angular momentum. We have N N (Mi + ri × Fi ) = i=1 i=1 N i=1 mi ri × N ∂ω ∂vo + ω × vo + m i ri × × ri + ∂t ∂t i=1 mi ri × (ω × (ω × ri )).3.3 Stationary Frame 3 VESSEL INERTIAL DYNAMICS A frame ﬁxed in inertial space. as the string travels through a small arc δψ.
but oﬀers no similar condensation: N N mi ri × ((ω · ri )ω − (ω · ω)ri ) = i=1 i=1 mi ri × ω(ω · ri ) (31) = = (Izz − Iyy )rq N i=1 mi (yi r − zi q)(xi p + yi q + zi r) N i=1 mi (zi p − xi r)(xi p + yi q + zi r) N i=1 mi (xi q − yi p)(xi p + yi q + zi r) Iyz (q 2 − r2 ) + Ixz pq − Ixy pr Ixz (r2 − p2 ) + Ixy rq − Iyz pq + Ixy (p2 − q 2 ) + Iyz pr − Ixz qr (Ixx − Izz )rp . The third term can be worked out along the same lines. the second term of the angular momentum righthand side collapses neatly into I∂ω/∂t. Ixx Ixy Ixz I = Iyx Iyy Iyz Izx Izy Izz N 25 (inertia matrix) Ixx = i=1 N 2 mi (yi + zi2 ) Iyy = i=1 N mi (x2 + zi2 ) i 2 mi (x2 + yi ) i i=1 N Izz = Ixy = Iyx = − i=1 N mi xi yi mi xi zi i=1 N (crossinertia) Ixz = Izx = − Iyz = Izy = − i=1 mi yi zi .4 Angular Momentum Employing the deﬁnitions of moments of inertia. (Iyy − Ixx )qp .3.
3. N } be the total moment acting on the body.5 Example: Spinning Book Consider a homogeneous rectangular block with Ixx < Iyy < Izz and all oﬀdiagonal moments of inertia are zero. are dp + (Izz − Iyy )rq = 0 dt dq Iyy + (Ixx − Izz )pr = 0 dt dr Izz + (Iyy − Ixx )qp = 0. The interesting result is that rotations about the x and z axes are unstable. p = Ω + δp.26 3 VESSEL INERTIAL DYNAMICS Letting M = {K.. with constant angular rate Ω. while rotation about the y axis is not. ˙ ˙ (32) 3. the complete moment equations are K = Ixx p + Ixy q + Ixz r + ˙ ˙ ˙ (Izz − Iyy )rq + Iyz (q 2 − r2 ) + Ixz pq − Ixy pr + m [yG (w + pv − qu) − zG (v + ru − pw)] ˙ ˙ M = Iyx p + Iyy q + Iyz r + ˙ ˙ ˙ (Ixx − Izz )pr + Ixz (r2 − p2 ) + Ixy qr − Iyz qp + m [zG (u + qw − rv) − xG (w + pv − qu)] ˙ ˙ N = Izx p + Izy q + Izz r + ˙ ˙ ˙ (Iyy − Ixx )pq + Ixy (p2 − q 2 ) + Iyz pr − Ixz qr + m [xG (v + ru − pw) − yG (u + qw − rv)] . i. q = δq. The ﬁrst equation above is uncoupled . where the δ preﬁx indicates a small value compared to Ω.1 xaxis In the case of the xaxis. dt Ixx We consider in turn the stability of rotations about each of the main axes. M.e. and r = δr. the left side of Equation 28. with no external forces or moments.5. The linearized angular momentum equations.
and the solution grows without bound. ¨ when δ q(0) = 0. ∂t2 Iyy Ixx The coeﬃcients are positive.3. both coeﬃcients of the diﬀerential ˙ equation are positive. following δp(t) = √ δp(0)e −αt . ∂t2 Izz Ixx Here the second coeﬃcient has negative sign.5. that is. and hence α > 0.3 zaxis Finally. it too oscillates.2 yaxis Now suppose q = Ω + δq: diﬀerentiate the ﬁrst equation and substitute into the third equation to obtain ∂ 2 δp + (Iyy − Ixx )(Iyy − Izz )Ω2 δp = 0. The exponent is real now. let r = Ω + δr: diﬀerentiate the ﬁrst equation and substitute into the second equation to obtain ∂ 2 δp + (Ixx − Izz )(Iyy − Izz )Ω2 δp = 0. 3. so bounded oscillations occur. The imaginary exponent indicates √ that the solution is of the form δq(t) = δq(0)cos αt.5 Example: Spinning Book 27 from the others. which has response δq(t) = δq(0)e −αt . Diﬀerentiate the second equation to obtain ∂ 2 δq ∂δr + (Ixx − Izz )Ω =0 2 ∂t ∂t Iyy Substitution of this result into the third equation yields ∂ 2 δq + (Ixx − Izz )(Ixx − Iyy )Ω2 δq = 0. it oscillates but does not grow. and indicates no change in δp.5. . and therefore α < 0. For spin about the xaxis. 3. ∂t2 √ Iyy Izz A simpler expression is δ q + αδq = 0. since the small term δqδr can be ignored. Since the perturbation δr is coupled.
Note that translation of MMOI using the parallel axis theorem must be either to or from a frame resting exactly at the center of gravity. • x. The formulas are: ¯ Ixx + m(δy 2 + δz 2 ) ¯ Iyy + m(δx2 + δz 2 ) ¯ Izz + m(δx2 + δy 2 ) ¯ Iyz − mδyδz ¯ Ixz − mδxδz ¯ Ixy − mδxδy. is the translation of the xaxis to the new frame. The parallel axis theorem allows one to translate the mass moments of inertia referenced to the mass center into another frame with parallel orientation. . the vector of bodyreferenced velocities. • E. in body coordinates. and vice versa. Ixx Iyy Izz Iyz Ixz Ixy = = = = = = (33) ¯ where I represents an MMOI in the axes of the mass center. we now have the necessary terms for writing a full nonlinear simulation of a rigid body. in this case rG = 0 also. comprising the following components: • vo . Ixz small enough that they can be ignored.28 3 VESSEL INERTIAL DYNAMICS 3. the mass center of an body is at a diﬀerent location than a more convenient measurement point. and δx. 3. Sometimes a translation of coordinates to the mass center will make the crossinertial terms Ixy . for example. as in the spinning book example. in inertial space. body rotation rate vector. the geometric center of a vessel for example. location of the body origin. • ω. so that the equations of motion are signiﬁcantly reduced. Euler angle vector. Iyz .7 Basis for Simulation Except for external forces and moments F and M . There are twelve states.6 Parallel Axis Theorem Often.
The Cartesian position propagates according to ˙ x = RT (E)vo . with some coupling which generally requires a 6 × 6 matrix inverse.7 Basis for Simulation 29 The derivatives of bodyreferenced velocity and rotation rate come from Equations 27 and 32.3. while the Euler angles follow: ˙ E = Γ(E)ω. (35) (34) .
so that a twovariable Taylor expansion would have the form (36) f (x. This section pertains to the use of hydrodynamic coeﬃcients for predicting hydrodynamic response. surface tension.1 Taylor Series and Hydrodynamic Coeﬃcients Recall the Taylor expansion of a function: ∂f (xo ) 1 ∂ 2 f (xo ) (x − xo ) + (x − xo )2 + · · · . and motions of the body. 4. vapor pressure. viscosity. This notation covers some instances where the formal Taylor series is meaningless. 2 . The most important factors for large ocean vehicles are density and motion. y) = f (xo . 2! ∂x2 fx = and so on. but the notation is still clear. including water density. ﬂuid drag is often written as 1 F = ρCd Auu = Fuu uu. yo ) + fx (x − xo ) + fy (y − yo ) + fxx (x − xo )2 + fyy (y − yo )2 + fxy (x − xo )(y − yo ) + fxxx (x − xo )3 + · · · . As one example. (37) (38) Note that all of the factorials are included in the coeﬃcients. and we can make simpliﬁcations to parameterize the most prominent relationships.30 4 HYDRODYNAMICS: INTRODUCTION 4 HYDRODYNAMICS: INTRODUCTION The forces and moments on a vessel are complicated functions of many factors. pressure. ∂x 2! ∂x2 We introduce the notation f (x) = f (xo ) + ∂f (xo ) ∂x 1 ∂ 2 f (xo ) fx x = .
Nx . yG = 0. zG is inconsequential. We assume that no hydrodynamic forces depend on the position of the vessel. in a homogeneous sea. this topic will be discussed later. {Xx . so positive r has the boat turning left. consider Xv : since this longitudinal force would have the same sign regardless of the sign of v (because of sidetoside hull symmetry). Xr . Xφ .2 Second. since ﬂuid particle acceleration ˙ ˙ ˙ ˙ relates linearly with pressure or force. Nu . where U >> u. The body xaxis is forward and the yaxis is to port. wave. with no current. Yu . Nu } = 0. We will consider motions only in the horizontal plane.2 Surface Vessel Linear Model We now discuss some of the hydrodynamic parameters which govern a ship maneuvering in the horizontal plane. Xv . Yu .4. The same argument shows that {Xr . Yy . and eliminating higherorder terms. we do not consider nonlinear acceleration Note that the linearized heave/pitch dynamics of a submarine do depend on the pitch angle. Xy . We then have at the outset ∂u − rv − xG r2 ∂t ∂v ∂r Y = m + ru + xG ∂t ∂t ∂r ∂v N = Izz + mxG + ru .2 Surface Vessel Linear Model 31 4. N = Izz ∂t ∂t X = m (40) A number of coeﬃcients can be discounted. this set is ∂u ∂t ∂v ∂r Y = m + rU + xG ∂t ∂t ∂r ∂v + mxG + rU . Ny . Yφ . it must have zero slope with v at the origin. Finally. or wind eﬀects. which means θ = ψ = p = q = w = 0. Nφ } are all zero. Thus Xv = 0. ∂t ∂t X = m (39) Letting u = U + u. Since the vessel is symmetric about the x − z plane. Yx . 2 . First.
r} and external force/moment vector F = {Y . and B is a gain matrix for the control and disturbance inputs. but in general Xvv = 0 only if the vessel is bowstern symmetric. ˙ ˙ ˙ ˙ (41) (42) (43) The right side here carries also the imposed forces from a thruster(s) and rudder(s) {X . or (44) = Ps + F = M −1 P s + M −1 F = As + B F . Yuu = 0 because of hull symmetry. For instance. It should be noted that some nonlinear terms related to those we have eliminated above are not zero. With the state vector s = {v. considering only the linear hydrodynamic terms. and therefore are of immediate interest. or higher time derivatives. 4. (45) The matrix M is a mass or inertia matrix. a statespace representation of the sway/yaw system is m − Yv mxG − Yr ˙ ˙ mxG − Nv Izz − Nr ˙ ˙ ds dt ˙ Ms ˙ s ˙ s Yv Yr − mU Nv Nr − mxG U = s + F . N }. N }. which is always invertible. ˙ .3 Stability of the Sway/Yaw System ˙ Consider the homogeneous system s = As: s1 = A11 s1 + A12 s2 ˙ s2 = A21 s1 + A22 s2 .32 4 HYDRODYNAMICS: INTRODUCTION terms. The last form of the equation is a standard one wherein A represents the internal dynamics of the system. Note that the surge equation is decoupled from the sway and yaw. (m − Xu )u = Xu u + X ˙ ˙ (m − Yv )v + (mxG − Yr )r = Yv v + (Yr − mU )r + Y ˙ ˙ ˙ ˙ (mxG − Nv )v + (Izz − Nr )r = Nv v − (Nr − mxG U )r + N . We have so far. Y . but that sway and yaw themselves are coupled.
Yr } take very small values in comparison with the others. xG = 0. (m − Yv )(Izz − Nr ) − (mxG − Yr )(mxG − Nv ) ˙ ˙ ˙ ˙ (48) A11 = A12 = A21 = A22 = (49) The denominators are identical. First.4. and can be simpliﬁed. valid for many vessels with the origin is at the geometric center. A necessary and suﬃcient condition for stability of this ODE system is that each coeﬃcient must be greater than zero: −A11 − A22 > 0 A11 A22 − −A12 A21 > 0 The components of A for the sway/yaw problem are (Izz − Nr )Yv + (Yr − mxG )Nv ˙ ˙ (m − Yv )(Izz − Nr ) − (mxG − Yr )(mxG − Nv ) ˙ ˙ ˙ ˙ −(Izz − Nr )(mU − Yr ) − (Yr − mxG )(mxG U − Nr ) ˙ ˙ (m − Yv )(Izz − Nr ) − (mxG − Yr )(mxG − Nv ) ˙ ˙ ˙ ˙ (Nv − mxG )Yv + (m − Yv )Nv ˙ ˙ (m − Yv )(Izz − Nr ) − (mxG − Yr )(mxG − Nv ) ˙ ˙ ˙ ˙ −(Nv − mxG )(mU − Yr ) − (m − Yv )(mxG U − Nr ) ˙ ˙ . these are the terms Yv and Nr . we would simply use s. Both Yv and Nr take large negative values. the terms {Nv . if the vessel is reasonably balanced with regard to forward and aft areas with respect to the origin. let xG 0. Next. ¨ ˙ (47) Note that these operations are allowed because the derivative operator is linear. in the language of the Laplace transform. Yr .3 Stability of the Sway/Yaw System We can rewrite the second equation as s2 = d(·) − A22 dt −1 33 A21 s1 (46) and substitute into the ﬁrst equation to give s1 + (−A11 − A22 )s1 + (A11 A22 − A12 A21 )s1 = 0. Linear ˙ ˙ ˙ drag and rotational drag are signiﬁcant also. If the origin is at the center of mass. ˙ and similarly Nr −Izz . . Nv . To wit. ˙ ˙ the added mass term −Yv is of the order of the vessel’s material mass m.
34 4 HYDRODYNAMICS: INTRODUCTION both large and negative. we have only to look at the numerators. since the denominators of the Aij are identical. As in our discussion of lift on the body. which has a large positive value. The second part of the second term contains mU . Adding more surface area aft drives Nv more positive. For the second condition.4 Basic Rudder Action in the Sway/Yaw Model Rudders are devices which develop large lift forces due to an angle of attack with respect to the oncoming ﬂuid. generally making stability critical on the (usually negative) Nv . The 2 . (52) Since Nr and Yv are both negative. The denominator for A’s components reduces to (m − Yv )(Izz − Nr ). and yaw/sway stability depends closely on the magnitude and sign of Nv . we require (Izz − Nr )Yv (m − Yv )Nr ˙ ˙ − [Nv Yv + (m − Yv )Nv ] [−(Izz − Nr )(mU − Yr ) + Yr Nr ] > 0. increasing stability as expected. ˙ ˙ ˙ ˙ (50) The ﬁrst term is the product of two large negative and two large positive numbers. where α is the angle of attack. positive xG increases the (positive) inﬂuence of C’s ﬁrst term. For stability. the form is as follows: L = 1 ρAU 2 Cl (α). The terms of C compete. and ˙ ˙ A11 = Yv m−Yv ˙ Nr Izz −Nr ˙ < 0 A22 = < 0. Nonzero xG shows up as follows: C = Yv (Nr − mxG U ) + Nv (mU − Yr ) > 0. Hence the ﬁrst condition for stability is met: −A11 − A22 > 0. (51) C is called the vessels stability parameter. 4. a simpler form is common: C = Yv Nr + Nv (mU − Yr ) > 0. When only the largest terms are considered for a vessel. Stability can also be improved by moving the center of gravity forward.
in tank testing with v = 0. δ. the hydrodynamic characteristics of the vessel depend strongly on the rudder.4. . We will assume that α is small enough that the linear relationship applies: Cl (α) = ∂Cl ∂α α. α=0 Note the diﬀerence between the rudder angle expressed in the body frame. even when δ = 0. A positive rudder deﬂection (deﬁned to have the same sense as the yaw angle) causes a negative yaw perturbation. other hydrodynamic derivatives are augmented to capture the necessary eﬀects. and the total angle of attack α. Thus. α=0 (53) Since the rudder develops force (and a small moment) far away from the body origin. the moment equation is quite simple. as well as v/U and lr.4 Basic Rudder Action in the Sway/Yaw Model y δ r x 35 Figure 3: Convention for positive rudder angle in the vessel reference system.. In this case the rudder still opposes yaw and sway perturbations and acts to stabilize the vessel. but the rudder stalls when the angle of attack reaches a critical value. We have 1 ∂Cl Yα = ρA 2 ∂α 1 ∂Cl Nα = − ρA 2 ∂α U2 α=0 (54) (55) lU 2 . but in real conditions. and thereafter develops much less lift. lift coeﬃcient Cl is normally linear with α near α = 0. Generally speaking. Angle of attack is inﬂuenced by δ. say a distance l aft. δ = α and Nδ = Nα . etc. and a very small positive sway perturbation. for example Nv and Nr .
we have ¨ ˙ (Izz − Nr )φ + (mxG U − Nr )φ = Nδ δ. but Nδ is large and negative. i. The control law δ = kφ (φ − φdesired ) + kr r is the basis for heading autopilots. so that a reasonable amount of control will provide good maneuvering capabilities. ˙ (58) Since all coeﬃcients are positive (recall Nδ < 0). settling under secondorder dynamics to θ(∞) = 0. which are used to track φdesired . where kr > 0 is a feedback gain. we require sensors to obtain r and φ. Hence C becomes more positive. secondorder ODE: ¨ ˙ (Izz − Nr )φ + (mxG U − Nr )φ − Nδ kφ φ = 0. the system equation becomes a homogeneous. since Yv is negative. is C = Yv (Nr − mxG U + kr Nδ ) + Nv (mU − Yr − kr Yδ ).2 Heading Control in the Sway/Yaw Model Considering just the yaw equation of motion.. a controller to calculate δ. (56) Yδ is small positive.e. On the other hand. In this case. 4. This use of an error signal to drive an actuator is in fact the essence of feedback control. ˙ (57) Employing the control law δ = kφ φ. v = 0. Control system limitations and the stalling of rudders make obvious the fact that even a very large control gain kr cannot completely solve stability problems of a poorlydesigned vessel with an inadequate rudder. the equation gives a stable θ response.36 4.1 4 HYDRODYNAMICS: INTRODUCTION Adding Yaw Damping through Feedback The stability coeﬃcient C resulting from the addition of a control law δ = kr r.4. and an actuator to implement the corrective action.4. a vessel which is overly stable (C >> 0 with no rudder action) is unmaneuverable. with a rudder. . A properlybalanced vessel just achieves stability with zero rudder action.
in which many terms are competing and contributing to the transient response.5 Response of the Vessel to Step Rudder Input 37 4.2 Phase 3: Steady State When the transients have decayed. we have ˙ Yδ δ . The equation looks like this: m − Yv mxG − Yr ˙ ˙ mxG − Nv Izz − Nr ˙ ˙ v ˙ r ˙ = Yδ Nδ δ. since the velocities are zero. The initial yaw is in the negative rdirection.5.4.5 4. If the vessel is unstable (C < 0). 4. This turning rate equation can also be used to estimate turning radius R: r=− R= U UC = . the steadystate yaw rate is negative. r −Yv Nδ δ (64) . (59) Since Yr and mxG are comparatively small in the ﬁrst row. the positive vdirection. the vessel is in a steady turning condition. The steady turning rate is thus approximated by Yv N δ δ. (60) m − Yv ˙ and the vessel moves to the left. acceleration terms dominate. since Nδ < 0: v(0) = ˙ r(0) = ˙ Nδ δ .5. it turns in the opposite direction than expected. (63) C With C > 0.1 Response of the Vessel to Step Rudder Input Phase 1: Accelerations Dominate When the rudder ﬁrst moves. The system equations reduce to v r = δ C (mxG U − Nr )Yδ + (Yr − mU )Nδ N v Y δ − Y v Nδ . Izz − Nr ˙ (61) The ﬁrst phase is followed by a period (Phase 2). (62) Note that the denominator is the stability parameter. and the accelerations are zero.
This fact is not true in the vertical plane. The linearized sway/yaw dynamics of a surface vessel are strongly coupled. and “ handsoﬀ” stability. for which the dimensional weight/buoyancy forces and moments are invariant with speed. ˙ ˙ ˙ dt m (65) (66) The last term in each equation is a hydrostatic eﬀect induced by opposing net buoyancy B and weight W . 4. Because buoyancy eﬀects do not .7 Stability in the Vertical Plane Stability in the horizontal plane changes very little as a function of speed. 3. the nonlinear equations couple surge to the other motions. Among other eﬀects. because drag and lift eﬀects generally scale with U 2 . Very tight maneuvering requires the nonlinear inertial components and hydrodynamic terms.6 Summary of the Linear Maneuvering Model We conclude our discussion of the yaw/sway model by noting that 1. 2. 4. For example. The case C < 0 should only be considered under active feedback control. The analysis is valid only up to small angles of attack and turning rates. We see also that increasing the rudder area increases Nδ . lb denotes the vertical separation of the center of gravity and the center of buoyancy. Increasing the deﬂection δ to reduce R works only to the point of stalling. The design parameter C should be slightly greater than zero for easy turning. decreasing R as desired. creating the socalled righting moment which nearly all underwater vehicles possess.38 4 HYDRODYNAMICS: INTRODUCTION The radius goes up directly with C. and the actual vessel loses forward speed during maneuvering. with xG = 0 and no actuation: ∂w − Uq = Zw w + Zw w + Zq˙ q + Zq q + (B − W ) ˙ ˙ ˙ ∂t dq Iyy = Mw w + Mw w + Mq˙ q + Mq q − Blb sin θ. indicating that too stable a ship has poor turning performance. consider the case of heave and pitch. and they are independent of the longitudinal dynamics.
7 Stability in the Vertical Plane 39 change with speed. the dynamic properties and hence stability of the vehicle may change with speed.4. .
On the basis of physical laws. · · · . The (MKS) units are [X(·)] [m] [F ] [T ] = = = = m kg kgm/s2 s. Consider the ﬂow rate Q of water from an open bucket of height h. we require that all terms in an equation must have the same units. through a drain nozzle of diameter d. π1 assumes only a constant (but unknown) value. At time zero. The theorem asserts that there are n − k dimensionless groups πi that can be formed. F. The dimensional function is X(T ) = fQ (m. πn−k−1 ). We form nondimensional groups out of the dimensional ones in this section. some quantities are dependent on other. There is just one nondimensional group in this relationship. not far at all from the known result that X(T ) = F T 2 /2m! Example.1 SIMILITUDE Use of Nondimensional Groups For a consistent description of physical processes. Suppose we have a block of mass m resting on a frictionless horizontal surface. and apply the technique to maneuvering. Qn−1 ). (67) There are n variables here. The Buckingham πtheorem provides a basis for all nondimensionalization. so n = 4. Simple termcancellation gives π1 = X(T )m/F T 2 . We have . but suppose also that there are only k independent ones. the distance that the block has moved as of time T .40 5 SIMILITUDE 5 5. independent quantities. We want to know X(T ). Let a quantity Qn be given as a function of a set of n − 1 other quantities: Qn = fQ (Q1 . Q2 . and therefore k = 3. T ). and the functional equivalence is reduced to πn−k = fπ (π1 . π2 . (68) Example. k is equivalent to the number of physical unit types encountered. · · · . a steady force of magnitude F is applied.
making sure that each of the original (dimensional) quantities is represented. . π2 ) −→ √ ρd gh h Q √ = fπ . ρ. Hence. µ. We ﬁnally have the useful result π4 = fπ (π5 . only three nondimensional groups exist. To arrive at a set of valid groups. and thus k = 3. and the (MKS) units of these quantities are: [Q] [h] [d] [ρ] [µ] [g] = = = = = = m3 /s m m kg/m3 kg/ms m/s2 There are only three units that appear: [length.1 Use of Nondimensional Groups 41 Q = fQ (h. they could easily change places in the ﬁrst two groups. π2 is more awkward. √ and we may construct π4 = π1 /π2 = Q/d2 gh √ nondimensionalize Q with to a pressure velocity. π1 is recognized as a Reynolds number pertaining to the oriﬁce ﬂow. and one is a unique function of the other two. g is the acceleration due to gravity. µ d d2 gh The uncertainty about where to use h and d. Since h and d have the same units. and then π5 = π1 /π4 = ρd gh/µ to establish an oriﬁce Reynolds number independent of Q. we must create three nondimensional quantities. mass]. time. We have n = 6. but products and fractions of groups are themselves valid groups. π2 = dρ gh/µ. and π3 = h/d. However. Intuition and additional manipulations come in handy. as we now show. √ Three plausible ﬁrst groups are: π1 = ρQ/dµ. Intuition is that h/d is im .5. d. and the questionable importance of h/d as a group are remnants of the theorem. Note that all six quantities appear at least once. and its absolute viscosity µ. No other parameters aﬀect the ﬂow rate. where the water density is ρ. g).
g is the acceleration due to gravity. In the ﬂow example. Cavitation number: δ= P∞ − P v . Froude number: U Fr = √ . degradation of performance.. gL (69) where U is the speed of the vessel. 5. 1 ρU 2 2 (70) where P∞ represents the ambient total pressure.2 Common Groups in Marine Engineering One frequently encounters the following groups in ﬂuid mechanics and marine engineering: 1. with no constraints about the actual physics taking place. The Froude number appears in problems involving pressure boundary conditions. and possible deterioration of the material. where λ is the wavelength. such as in waves on the ocean surface. π5 is a Reynolds number. 2. A low cavitation number means that the Bernoulli pressure loss across the lifting surface will cause the ﬂuid to vaporize.g. and U the propeller inlet velocity. Reynolds number: Re = Ul .42 5 SIMILITUDE material. µ/ρ (71) . the theorem reduced the number of independent parameters from ﬁve to two. 3. Pv the vapor pressure of the ﬂuid. and L is the waterline length of the vessel. e. it relates the vessel speed U to water wave speeds of wavelength L. Roughly speaking. the phase speed of a surface wave is V = λg/2π. causing bubbles. and the other two terms have a nice physical meaning. The power of the πtheorem is primarily in reducing the number of parameters which must be considered independently to characterize a process.
Various forces on the box scale as (inertia) Fi = ρ (gravity) Fg = (pressure) Fp = (shear) Fs = (surface tension) Fσ = ∂v ∂v dxdydz + ρv dxdydz ∂t ∂x 3 ρgdxdydz ρgl P dxdy P l2 ∂v µ dxdy µU l ∂z σdx σl. ρU 2 l2 Thus the groups listed above can be written as Fi Fr = Fg Fp δ= Fi Fi Re = Fs Fi W = Fσ U2 gl P ρU 2 ρU l µ ρU 2 l σ . and l normalizes length. σ (72) where σ is the surface tension of a ﬂuid. consider a box having side lengths [dx. viscous eﬀects are negligible. dz]. Since Re appears in many applications. and ﬂow across a plate.5. and ρ is density. dy. Weber number: W = ρU 2 l . 4. l represents one of many length scales. To appreciate the origins of these terms from a ﬂuid particle’s point of view. µ is absolute viscosity. Reynolds number is a ratio of ﬂuid inertial pressures to viscous pressures: When Re is high. ρU 2 normalizes pressure. Re can be used to characterize pipe ﬂow. bluﬀ body wakes. The Weber number indicates the importance of surface tension. among others.2 Common Groups in Marine Engineering 43 where U is velocity. Given that [σ] = N/m (MKS).
we set gmodel = gship . Suppose that we conduct a model test in which Re is abnormally large. the model and the true vessel operate in the same gravity ﬁeld. ˙ ˙ ˙ ˙ These equations can be nondimensionalized in a standard way. With this achieved. l) = fπ (Re. As a result. however. First. This rate cannot be achieved for lowerRe conditions because of ﬂuid drag in the oriﬁce. Then Reynolds number scaling implies directly νmodel = νship /λ3/2 . In a vessel. by using the quantities [U. For surface vessels. the viscous eﬀects are negligible. the ﬂow rate is Q = πd2 2gh/4. Unfortunately. Suppose that lship = λlmodel . and submarines near the surface.e.e. L. few ﬂuids with this property are workable in a large testing tank. W is very large.44 5 SIMILITUDE When testing models. W ). accurate scaling of Re for large vessels to model scale is quite diﬃcult. Above a critical value Re 500. U. and the all of the relevant groups governing ﬂuid dynamics. i. σ. and hence surface tension plays no role. we write the functional relationship for drag as a starting point: Cr = D 1 ρAU 2 2 = fQ (ρ. √ Froude number similitude requires Umodel = Uship / λ. ρ]: these three values provide the necessary units of length. one then tries to match F r closely... the surface roughness. Under inviscid √ conditions. since l is large. g. i. Consider the example of nozzle ﬂow from a bucket.3 Similitude in Maneuvering The linear equations of motion for the horizontal yaw/sway problem are: (m − Yv )v − Yv v + (mU − Yr )r + (mxG − Yr )r = Y ˙ ˙ ˙ ˙ (Izz − Nr )r + (mxG U − N r)r + (mxG − Nv )v − Nv v = N. it is imperative to maintain as many of the nondimensional groups as possible of the fullscale system. it is a routine procedure to employ turbulence stimulators to achieve ﬂow that would normally occur with shipscale Re. . 5. we look at Re = U l/ν and F r = U/ gl. This holds for the geometry of the body and the kinematics of the ﬂow. √ Next. µ. F r. both of which are important for surface vessels. Cf is not sensitive to Re. 000. so that usually λ >> 1. additionally.
for consistency with our previous expressions: v ˙ = m = m 1 ρL3 2 (74) Izz = xG = U = Izz 1 ρL5 2 xG L U =1 U Yv ˙ 1 ρL3 2 Yv 1 ρU L2 2 Yr ˙ 1 ρL4 2 Yr 1 ρU L3 2 Y 1 ρU 2 L2 2 Nv ˙ 1 ρL4 2 Nv 1 ρU L3 2 Yv = ˙ Yv = Yr = ˙ Yr = Y = Nv = ˙ Nv = . and mass. we create nondimensional states. First. denoted with a prime symbol: L v ˙ (73) U2 1 v = v U L2 r = ˙ r ˙ U2 L r = r. U We follow a similar procedure for the constant terms as follows. and furthermore are readily accessible to the user.5.3 Similitude in Maneuvering 45 time. including a factor of 1/2 with ρ.
46 Nr ˙ 1 ρL5 2 Nr 1 ρU L4 2 .4 Roll Equation Similitude Certain nondimensional coeﬃcients may arise which depend explicitly on U . The nondimensional terms are Ixx 1 ρL5 2 Kp ˙ 1 ρL5 2 Kp 1 ρU L4 2 Kψ 1 ρU 2 L3 2 Ixx = Kp = ˙ Kp = Kψ = (77) . and every moment 2 1 with 2 ρU 2 L3 . ˙ ˙ (76) For a surface vessel. (m − Yv )v − Yv v + (m U − Yr )r + (m xG − Yr )r ˙ ˙ ˙ ˙ (Izz − Nr )r + (m xG U − N r )r + (m xG − Nv )v − Nv v ˙ ˙ ˙ ˙ = Y (75) = N. Thus we arrive at a completely equivalent set of nondimensional system equations. and has the form Kψ = −ρg (GM ). and therefore require special attention. 5. Since ﬂuid forces and moments generally scale with U 2 . the roll moment Kψ is based on metacentric stability. 1 ρU 2 L3 2 N 5 SIMILITUDE Nr = ˙ Nr = N = Note that every force has been normalized with 1 ρU 2 L2 . and GM is the metacentric height. where is the displaced ﬂuid volume of the vessel. Let us carry out a similar normalization of the simpliﬁed roll equation (Ixx − Kp )p − Kp p − Kψ ψ = K. time has been also nondimensionalized with L/U . the nondimensionalized description holds for a range of velocities.
(78) U2 Note that the roll angle φ was not nondimensionalized. The nondimensional coeﬃcient becomes (Ixx − Kp )p − Kp p − ˙ ˙ Bh . this Kψ needs to be maintained in model tests. . and h is the righting arm.4 Roll Equation Similitude L2 = p ˙ U2 L p. since it appears explicitly in the nondimensional righting moment term.5. the righting moment has the form Kψ = −Bh. where B is the buoyant force. = U 47 p ˙ p leading to the equivalent system 2gL (GM )φ = K . and also has a strong inﬂuence on Kp . ρU 2 L3 2 Kψ = − 1 Kψ again depends strongly on U . since B and h are ﬁxed. In the case of a submarine. The Froude number has a very strong inﬂuence on roll stability.
which scales the inﬂuence gL of surface waves. need not be matched as long as the scale model attains turbulent ﬂow (supercritical Re). The idea is to measure the crossbody force and yaw moment as a function of r. 6. Since the control surface(s) and propeller(s) aﬀect the coeﬃcients. For example. tow the vehicle at diﬀerent angles of attack. Finally. higherorder terms can be generated if the points deviate from a straight line. 6. In model tests. The slope of the curve at zero angle determines Yv and Nv respectively. Finally. which scales the eﬀect of viscosity. the vessel is ﬁxed on an arm of length R. modern sensors and computer control systems make possible the estimation of certain coeﬃcients based on openwater tests of a model or a fullscale design. The available computational approaches should be considered as well. a preliminary search of the literature may turn up useful estimates. giving the coeﬃcients Yr and Nr . the measurement is made over one rotation only. test results for many hullforms have already been published. One can use turbulence stimulators near the bow if necessary. must be maintained between model and fullscale surface vessels. the Froude number F r = √U .2 Rotating Arm Device On a rotating arm device. Note that the lateral force also contains the component (m − Yv )r2 R.1 Towtank In a towtank. ˙ The coeﬃcients Yv and Nv can also be obtained by running with a ﬁxed angle of attack. they should both be implemented in model testing. Reynolds number Re = UνL . these are very good for predicting added mass in particular. rotating at constant rate r: the vessel forward speed is U = rR. so that the vessel does not reenter its own wake. and the basic lifting surface models are not diﬃcult. Rudder derivatives can be computed also by towing with various rudder angles. .48 6 CAPTIVE MEASUREMENTS 6 CAPTIVE MEASUREMENTS Before making the decision to measure hydrodynamic derivatives. measuring sway force and yaw moment.
the vessel is towed at constant forward speed U . therefore producing variable yaw. If linearity holds. 2l v = − (82) (80) Equating the sine terms and then the cosine terms. (79) and the transverse forces on the posts are measured and approximated as Ya (t) = Fa cos(ωt + θa ) Yb (t) = Fb cos(ωt + θb ). we obtain four independent equations: . When a = b. The model moves in pure sway if ya (t) = yb (t). then (m − Yv )v − Yv v + (mU − Yr )r + (mxG − Yr )r = Ya + Yb (81) ˙ ˙ ˙ ˙ (Izz − Nr )r + (mxG U − Nr )r + (mxG − Nv )v − Nv v = (Yb − Ya )l. but is held by two posts. these become ˙ ˙ ˙ ˙ aω (sin ωt(1 + cos ψ) + cos ωt sin ψ) 2 aω 2 v = − ˙ (cos ωt(1 + cos ψ) − sinωt sin ψ) 2 aω r = − (sin ωt(cos ψ − 1) + cos ωt sin ψ) 2l aω 2 r = − ˙ (cos ωt(cos ψ − 1) − sin ωt sin ψ) .3 PlanarMotion Mechanism 49 6. Usually a sinusoidal motion is imposed: ya (t) = a cos ωt yb (t) = b cos(ωt + ψ). in a pure yaw motion about the midlength point if ya (t) = −yb (t). one forward and one aft. or in a combination sway and yaw motion.3 PlanarMotion Mechanism With a planar motion mechanism. ˙ ˙ ˙ ˙ We have v = (ya + ya )/2 and r = (yb − ya )/2l. which can each impose independent sway motions.6. The connection points are a distance l forward and aft from the vessel origin.
50 6 CAPTIVE MEASUREMENTS (m − Yv ) − ˙ aω 2 (1 + cos ψ) − 2 aω Yv − sin ψ + 2 aω (mU − Yr ) − sin ψ + 2l aω 2 (mxG − Yr ) − (cos ψ − 1) = Fa cos θa + Fb cos θb ˙ 2l (m − Yv ) − ˙ (83) aω 2 (− sin ψ) − 2 aω (1 + cos ψ) + Yv − 2 aω (mU − Yr ) − (cos ψ − 1) + 2l aω 2 (mxG − Yr ) − (− sin ψ) = −Fa sin θa − Fb sin θb ˙ 2l aω 2 (cos ψ − 1) + 2l aω sin ψ + (mxG U − Nr ) − 2l aω 2 (mxG − Nv ) − (1 + cos ψ) − ˙ 2 aω Nv − sin ψ = l(Fb cos θb − Fa cos θa ) 2 (Izz − Nr ) − ˙ (Izz − Nr ) − ˙ (84) (85) aω 2 (− sin ψ) + (86) 2l aω (mxG U − Nr ) − (cos ψ − 1) + 2l aω 2 (mxG − Nv ) − (− sin ψ) − ˙ 2 aω Nv − (1 + cos ψ) = l(−Fb sin θb + Fa sin θa ) 2 .
6. (87) to be solved respectively for [Yv . Nv . . we can write the eight solutions ˙ ˙ ˙ ˙ directly: For ψ = 0. aω 2 2 aω −Yv − 2 aω 2 (mxG − Nv ) − ˙ 2 aω −Nv − 2 (m − Yv ) − ˙ (2) = Fa cos θa + Fb cos θb (2) = −Fa sin θa − Fb sin θb (2) = l(Fb cos θb − Fa cos θa ) (2) = l(−Fb sin θb + Fa sin θa ). ψ. a. Izz . Fb . Yr . Yv .3 PlanarMotion Mechanism 51 In this set of four equations. (88) to be solved for [Yr . Yv . It turns out that the two cases of ψ = 0 (pure sway motion) and ψ = 180o (pure yaw motion) yield a total of eight independent equations. Nv ]. The PMM can be driven with more complex trajectories which will target speciﬁc nonlinear terms. From the experiment. for a given speed. the eight linear coeﬃcients for a surface ˙ ˙ vessel maneuvering. we have ˙ ˙ aω 2 2l aω (mU − Yr ) − 2l aω 2 (Izz − Nr ) − ˙ 2l aω (mxG U − Nr ) − 2l (mxG − Yr ) − ˙ (−2) = Fa cos θa + Fb cos θb (−2) = −Fa sin θa − Fb sin θb (−2) = l(Fb cos θb − Fa cos θa ) (−2) = l(−Fb sin θb + Fa sin θa ). For ψ = 180o . exactly what is required to ﬁnd the eight coeﬃcients [Yv . Nv . Yr . ω]. we obtain [Fa . We note that the nonlinear terms will play a signiﬁcant role if the motions are too large. Nr ]. Remarkably. Nr ]. θb ]. and from the rigidbody model we have [m. Nr . θa . Yr . Nr . Thus. Nv . xG ]. we know from the imposed motion the values [U. can be deduced from two tests with a planar motion mechanism. and that some curve ﬁtting will be needed in any event.
Turn rudder quickly by 15◦ . achieve steady speed for one minute. manifested as a hysteresis in yaw rate r. 2. For example. Proceed back up to 15◦ . Repeat in decrements of 5◦ . and e then a contracting spiral in the opposite direction. suppose that the ﬁrst 15◦ rudder deﬂection causes the vessel to turn right. 3. Reduce rudder angle by 5◦ . .52 7 STANDARD MANEUVERING TESTS 7 STANDARD MANEUVERING TESTS This section describes some of the typical maneuvering tests which are performed on fullscale vessels. It is easy to see that the rudder in this case has to be used excessively driving the vessel back and forth. and hold until the vessel turns to 20◦ with respect to the starting heading. 3. to 15◦ . is still to the right. on the way negative. 4. and keep it there until steady yaw rate is maintained for one minute. We say that the vessel is unstable. Deﬂect the rudder to 20◦ . 5.1 Dieudonn´ Spiral e 1. and clearly a poor design. The test reveals if the vessel has a memory eﬀect. Achieve steady speed and direction for one minute. 4. to assess stability and performance. 7. With zero rudder. Repeat. 2. but that the yaw rate at zero rudder. No changes in speed setting are made after this point. the same memory eﬀect may occur.2 ZigZag Maneuver 1. 7. But if the corrective action causes the vessel to turn left at all. and will require a negative rudder action to pull out of the turn. and keep it there until steady yaw rate is maintained for one minute. Deﬂect the rudder to 20◦ . and hold until the vessel turns 20◦ . The Dieudonn´ maneuver has the vessel path following a growing spiral. The vessel has gotten “stuck” here.
At this location. the drift angle β. A line along the vessel centerline reaches closest to the true center of the turning circle at a point termed the turning center. parameters of interest are the turning diameter. In a surface vessel. similar tests can be made with diﬀerent rudder angles and diﬀerent threshold vessel headings. 7. After steady state is reached again. and the total period for the 20◦ oscillations. When the rudder is ﬁrst deﬂected. These are: the response time (time to reach a given heading). 7.3. The vessel responds by turning in a circle.7. Of course.3. the speed loss.3 Heel Angle Heel during turning occurs as a result of the intrinsic coupling of sway. The vesselreference velocity components are thus u = U cos β and v = −U sin β.2 Speed Loss Speed loss occurs primarily because of drag induced by the drift angle.3 Circle Maneuver 53 This maneuver establishes several important characteristics of the yaw response.3. the rudder is moved to a new setting. inertial terms dominate (Phase 1) and the sway equation is . but the center of gravity is near the waterline or above. the ﬂuid forces act below the waterline.3 Circle Maneuver From a steady speed.1 Drift Angle The drift angle is the equivalent to angle of attack for lifting surfaces. A vessel which drifts very little may have very little speed loss. yaw. and describes how the vessel “skids” during a turn. and it appears to an observer there that the vessel turns about this point. 7. If the turning circle has radius R (measured from the vessel origin). and the angle of heel ψ. then the speed tangential to the circle is U = rR. the yaw overshoot (amount the vessel exceeds ±20◦ when the rudder has turned the other way). zero yaw rate condition. and roll caused by the center of gravity. which may or may not exist on the physical vessel. 7. there is no apparent lateral velocity.
In ˙ the steady state. From the sway equation. we see that if the rudder is straightened out at this point. −Yv v + (mU − Yr )r = Yδ δ. −Yv v (acting high). The transition between the inertiallydominated and steadyturning regimes includes an overshoot: in the above formulas. hydrodynamic forces equalize the centrifugal force mU r and the rudder force Yδ δ. which have the rigid mass center above the center of ﬂuid forcing. the roll angle of a submarine with a sail is always into . submarines have the mass center below the rudder action point. ˙ ˙ ˙ ˙ (89) The coeﬃcients for r are quite small. and Yv v > Yδ δ.54 7 STANDARD MANEUVERING TESTS (m − Yv )v − (Yr − mxG )r = Yδ δ. the roll will momentarily be even worse! In summary. the keel out of the turn. toward the center of the turn. 7. (92) The drift angle β keeps the Yv force. (90) with Yr small. the vessel ultimately rolls to port (negative) for a positive rudder action. Unlike surface vessels. Hence. and again centrifugal force mU r causes the bottom of the submarine to move out of the turn. the vessel rolls into the turn initially. the vessel rolls under the sail. The sway equation is −Yv v + (mU − Yr )r = Yδ δ. The inertial equation (m − Yv )v − (Yr − mxG )r = Yδ δ ˙ ˙ ˙ ˙ (91) is dominated by mv (acting low). and additionally feel the eﬀects of a large sail above both. but then out of the turn in the steady state.4 Heeling in Submarines with Sails Submarines typically roll into a turn during all phases. the vessel overshoots the ﬁnal port roll angle as the vessel slows. and thus the vessel ﬁrst rolls to starboard ˙ (positive) for a positive rudder action. Because the centrifugal force acts above the waterline. v < 0 when r > 0 for most vessels. acting high. and Yδ δ (intermediate).3. ˙ ˙˙ Because Yv  > m. When steady turning conditions are reached (Phase 3).
.7.3 Circle Maneuver 55 the turn. The heel angle declines as the speed of the submarine drops. both initially and in the steady state.
and hence area Ao is usually that of the wetted surface. which has the form 1 FA = − ρCA Ao U 2 . and the negative sign indicates a negative pitch with respect to the vehicle’s pitch axis. 8. We will limit the present discussion to the vertical plane. Note that the Asubscript will be used to denote zero angle of attack conditions. (94) Azz > Axx for a slender body. Consider a symmetric body with added mass components Axx along the vehicle (slender) xaxis (forward). taken to be positive with the nose up – this equates to a negative pitch angle φ in vehicle coordinates. where pressure is highest on the front of the body (decelerating ﬂow) and lowest on the back (accelerating ﬂow). This Munk moment arises for a simple reason. (93) 2 The drag coeﬃcient CA has both pressure and skin friction components. but similar arguments can be used to describe the horizontal plane. because it opposes the vehicle’s xaxis. and Azz along the vehicle’s zaxis z (up). if it is moving horizontally. The Munk moment is always destabilizing. Let α represent the angle of attack. also. or from slender body approximation (to follow).2 Munk Moment Any shape other than a sphere generates a moment when inclined in an inviscid ﬂow. in the sense that it acts to turn the vehicle perpendicular to the ﬂow. The Munk moment is: 1 Mm = − (Azz − Axx )U 2 sin 2α 2 −(Azz − Axx )U 2 α.56 8 STREAMLINED BODIES 8 8. d’Alembert’s paradox predicts zero net force. the asymmetric location of the stagnation points. the sign of FA is negative. from numerical calculation. but not necessarily a zero moment. . The added mass terms Azz and Axx can be estimated from analytical expressions (available only for regular shapes such as ellipsoids).1 STREAMLINED BODIES Nominal Drag Force A symmetric streamlined body at zero angle of attack experiences only a drag force.
Each constant γ is taken as positive. 2 ∂α Fn = (95) With a positive angle of attack.8. and a small region near the trailing end. and it may cover a larger area (increasing the stabilizing moment and drag) for a larger angle of attack.3 Separation Moment 57 8. The formation of the vortex depends on the angle of attack. i. First we simplify: Fa = −γa Fn = γn α Mm = −γm α. In this latter area. 8. the transverse force Fn can be written in the same form as for control surfaces: 1 ρCn Ao U 2 2 1 ∂Cn ρ αAo U 2 . and their magnitudes determine the stability of a hullform. with a noseup angle of attack. For a small angle of attack.. Since vortices correlate with low pressure. but it also induces drag. and the signs reﬂect orientation in the vehicle reference frame.e. with the exceptions of the boundary layer. The Munk moment is a . (96) The last relation is based on writing CA (U cos φ)2 as (CA cos2 φ)U 2 . where 2 Ca = CA cos2 φ. ﬂow over a streamlined body is similar to that of potential ﬂow. a decomposition using apparent velocity. a helical vortex may form and convect downstream. the eﬀect of such a vortex is stabilizing.4 Net Eﬀects: Aerodynamic Center The Munk moment and the moment induced by separation are competing. The zeroα drag force FA is modiﬁed by the vortex shedding: 1 Fa = − ρCa Ao U 2 .3 Separation Moment In a viscous ﬂuid. this force is in the positive zdirection.
The summed forces on the body are thus: . since the Munk moment persists even in inviscid ﬂow. For selfpropelled vehicles. The net moment about O is zero if we select ln = γm . and are signed again to match the vehicle frame. AC moves inﬁnitely far forward as viscosity eﬀects diminish. and write the total pitch moment about O as: M = Mm + Fn ln = (−γm + γn ln )α. The point AC has an intuitive explanation: it is the location on the hull where Fn would act to create the total moment. 8. the net moment is stabilizing. We pick a temporary origin O for Fn however. Similarly.58 8 STREAMLINED BODIES pure couple which does not depend on a reference point.5 Role of Fins in Moving the Aerodynamic Center Control surfaces or ﬁxed ﬁns are often attached to the stern of a slender vehicle to enhance directional stability. and in this case. the moment is destabilizing. If the origin lies behind AC. the mass center must be forward of AC for stability. if the vehicle’s origin lies in front of AC. (97) where ln denotes the (positive) distance between O and the application point of Fn . γn (98) and the location of O is then called the aerodynamic center or AC. the towpoint must be located forward of AC. with γ > 0 and α > 0. In many cases with very streamlined bodies. for towed vehicles. a rigid sting would have to extend at least to AC in order for stable towing. As a ﬁnal note. Fixed surfaces induce lift and drag on the body: 1 L = ρAf U 2 Cl (α) 2 1 D = − ρAf U 2 Cd 2 γL α −γD (constant) (99) These forces act somewhere on the ﬁn. the aerodynamic center is signiﬁcantly ahead of the nose. Hence.
γn Hence. the ﬁns act directly in the denominator to shorten lf . lf n < 0. so that lf might be very small. If we say that the ﬁns act a distance lf behind the temporary origin O. To solve it explicitly. let lf n denote the (positive) distance that the ﬁns are located behind Fn . Equation 102 contains two length measurements. since both eﬀects usually act near the stern. Without ﬁns. The moment about O vanishes if γm = γn ln + lf (γL + γD ). We solve for lf : γm + γn lf n . this is likely a small number. and thus AC can be referenced to any other ﬁxed point easily. but since AC is referenced to the ﬁns.5 Role of Fins in Moving the Aerodynamic Center 59 X = Fa − D cos α + L sin α −γa − γD + γL α2 Z = Fn + L cos α + D sin α γn α + γL α + γD α. A vehicle with excessively large ﬁns will be diﬃcult to turn and maneuver. Note that if the ﬁns are located forward of the vortex shedding force Fn . lf is reduced. it can be recalled that lf = ln = γm . (103) γn + γL + γD This is the distance that AC is located forward of the ﬁns. i. referenced to an arbitrary body point O.e. With very large ﬁns. and that the moment carried by the ﬁns themselves is very small (compared to the moment induced by Llf ) the total moment is as follows: M = (−γm + γn ln )α + (γL + γD )lf α. this is the case of AC moving aft toward the ﬁns. (102) (101) The Munk moment γm opposes the aggregate eﬀects of vorticity lift γn and the ﬁns’ lift and drag γL + γD .8. (100) All of the forces are pushing the vehicle up. there is no net gain in stability.. this latter term is large. .
60 8 STREAMLINED BODIES 8. which depends on both w and q. we consider the contribution of the ﬁns only – slender body theory. Setting Fa and Fn to account for the fuselage and ﬁns. Mw . Zq . we have 1 ˆ ˆ − ρCa Ao U 2 2 1 ˆ ˆ 2 ˆ Z = Fn ρ C Ao U α 2 n 1ˆ ˆ ˆ M = −Fn xAC − Cn Ao U 2 xAC α.7 Coeﬃcients Zw . The ﬁn center of pressure is located a distance lf aft of the body origin. provides good results for the hull. as well as their derivatives. and we assume that the vehicle is moving horizontally.6 Aggregate Eﬀects of Body and Fins Since all of the terms discussed so far have the same dependence on α. The angle of attack seen by the ﬁn is a combination of a part due to θ and a part linear with q: αf −θ + lf q U (107) . introduced shortly. U We can then write several linear hydrodynamic coeﬃcients easily: 1 ˆ ˆ Zw = − ρCn Ao U 2 1 ˆ ˆ Mw = ρC Ao U xAC . it is ˆ ˆ possible to group them into a condensed form. and Mq for a Slender Body The angle of attack α is related to the crossbody velocity w as follows: α = − tan−1 − w u (105) w for U >> w. 2 n (106) The rotation of the vessel involves complex ﬂow. To start. with an instantaneous pitch angle of θ. 2 ˆ X = Fa (104) 8.
8.7 Coeﬃcients Zw , Mw , Zq , and Mq for a Slender Body and so lateral force and moment derivatives (for the ﬁn alone) emerge as 1 Zq = − ρCl Af U lf 2 1 2 Mq = − ρCl Af U lf . 2
61
(108)
62
9 SLENDERBODY THEORY
9
9.1
SLENDERBODY THEORY
Introduction
Consider a slender body with d << L, that is mostly straight. The body could be asymmetric in crosssection, or even ﬂexible, but we require that the lateral variations are small and smooth along the length. The idea of the slenderbody theory, under these assumptions, is to think of the body as a longitudinal stack of thin sections, each having an easilycomputed added mass. The eﬀects are integrated along the length to approximate lift force and moment. Slenderbody theory is accurate for small ratios d/L, except near the ends of the body. As one example, if the diameter of a body of revolution is d(s), then we can compute δma (x), where the nominal added mass value for a cylinder is π (109) δma = ρ d2 δx. 4 The added mass is equal to the mass of the water displaced by the cylinder. The equation above turns out to be a good approximation for a number of twodimensional shapes, including ﬂat plates and ellipses, if d is taken as the width dimension presented to the ﬂow. Many formulas for added mass of twodimensional sections, as well as for simple threedimensional bodies, can be found in the books by Newman and Blevins.
9.2
Kinematics Following the Fluid
The added mass forces and moments derive from accelerations that a ﬂuid particles experience when they encounter the body. We use the notion of a ﬂuid derivative for this purpose: the operator d/dt indicates a derivative taken in the frame of the passing particle, not the vehicle. Hence, this usage has an indirect connection with the derivative described in our previous discussion of rigidbody dynamics. For the purposes of explaining the theory, we will consider the twodimensional heave/surge problem only. The local geometry is described by the location of the centerline; it has vertical location (in body coordinates) of zb (x, t), and local angle α(x, t). The timedependence indicates that the conﬁguration is free to change with time, i.e., the body is ﬂexible. Note that the curvilinear coordinate s is nearly equal to the bodyreference (linear) coordinate x. The velocity of a ﬂuid particle normal to the body at x is wn (t, x):
9.3 Derivative Following the Fluid
63
wn =
∂zb cos α − U sin α. ∂t
(110)
The ﬁrst component is the time derivative in the body frame, and the second due to the deﬂection of the particle by the inclined body. If the body reference frame is rotated to the ﬂow, that is, if w = 0, then ∂zb /∂t will contain w. For small angles, sin α tan α = ∂zb /∂x, and we can write ∂zb ∂zb −U . ∂t ∂x
wn
The ﬂuid derivative operator in action is as follows:
wn =
dzb = dt
∂ ∂ −U zb . ∂t ∂x
9.3
Derivative Following the Fluid
A more formal derivation for the ﬂuid derivative operator is quite simple. Let µ(x, t) represent some property of a ﬂuid particle. d 1 [µ(x, t)] = lim [µ(x + δx, t + δt) − µ(x, t)] δt→0 δt dt ∂µ ∂µ = −U . ∂t ∂x The second equality can be veriﬁed using a Taylor series expansion of µ(x + δx, t + δt): ∂µ ∂µ δt + δx + h.o.t., ∂t ∂x
µ(x + δx, t + δt) = µ(x, t) +
and noting that δx = −U δt. The ﬂuid is convected downstream with respect to the body.
64
9 SLENDERBODY THEORY
9.4
Diﬀerential Force on the Body
If the local transverse velocity is wn (x, t), then the diﬀerential inertial force on the body here is the derivative (following the ﬂuid) of the momentum: δF = − d [ma (x, t)wn (x, t)] δx. dt (111)
Note that we could here let the added mass vary with time also – this is the case of a changing crosssection! The lateral velocity of the point zb (x) in the bodyreference frame is ∂zb = w − xq, ∂t such that wn = w − xq − U α. Taking the derivative, we have δF ∂ ∂ =− −U [ma (x, t)(w(t) − xq(t) − U α(x, t))] . δx ∂t ∂x We now restrict ourselves to a rigid body, so that neither ma nor α may change with time. δF ∂ = ma (x)(−w + xq) + U ˙ ˙ [ma (x)(w − xq − U α)] . δx ∂x (114) (113) (112)
9.5
Total Force on a Vessel
The net lift force on the body, computed with strip theory is
xN
Z =
xT
δF dx
(115)
where xT represents the coordinate of the tail, and xN is the coordinate of the nose. Expanding, we have
9.6 Total Moment on a Vessel A similar procedure can be applied to the moment predictions from slender body theory (again for small α): . ˙ ˙ x=xT ma (x) [−w + xq] dx + U ˙ ˙ xN xN We made use here of the added mass deﬁnitions xN m33 = xT ma (x)dx xN xT m35 = − xma (x)dx.9. however. with no eﬀect on Z. Additionally. α could take any value(s) along the body. In general. for vessels with pointed noses and ﬂat tails. Another noteworthy fact is that the lift force depends only on α at the tail. It is interesting to note that both Zw and Zα(xT ) depend on a nonzero base area. the added mass ma at the nose is zero. since their diﬀerence depends only on how the body coordinate system is oriented to the ﬂow.6 Total Moment on a Vessel 65 Z = ∂ [ma (x)(w − xq − U α)] dx xT xT ∂x = −m33 w − m35 q + U ma (x)(w − xq − U α)x=xN . the strip theory thus provides Zw ˙ Zq˙ Zw Zq Zα(xT ) = = = = = −m33 −m35 −U ma (xT ) U xT ma (xT ) U 2 ma (xT ). The two terms are clearly related. so this should come as no surprise. ˙ ˙ (116) In terms of the linear hydrodynamic derivatives. so that a simpler form occurs: Z = −m33 w − m35 q − U ma (xT )(w − xT q − U α(xT )). potential ﬂow estimates do not create lift (or drag) forces for a smooth body.
∂x xT xT Then we make the further deﬁnition x xN m55 = xT x2 ma (x)dx. the linear hydrodynamic moment derivatives are Mw ˙ Mq˙ Mw Mq Mα = = = = = −m35 −m55 U xT ma (xT ) + U m33 −U x2 ma (xT ) + U m35 T −U 2 xT ma (xT ) − U 2 m33 . we now assume that α is in fact constant on the length. whose linearization would provide Mw = (m33 − m11 )U . Mw and Mα are closely related. The Munk moment (an exact result) may therefore be used to make a correction to the second term in the slenderbody approximation above of Mw . (note that m35 = m53 ) and use integration by parts to obtain M = −m35 w − m55 q − U xma (x)(w − xq − U α)x=xN + ˙ ˙ x=xT xN U xT ma (x)(w − xq − U α)dx. As with the lift force. depending only on the orientation of the body frame to the ﬂow. . Finally. leading to M = −m35 w − m55 q + U xT ma (xT )(w − xT q − U α) + ˙ ˙ U m33 w + U m35 q − U 2 m33 α. The integral above contains the product ma (x)α(x).66 9 SLENDERBODY THEORY M = − = xN xδF dx xT xN ∂ ∂ −U [ma (x)(w − xq − U α)] dx ∂t ∂x xT xN xN ∂ = xma (x)(w − xq)dx − U ˙ ˙ x [ma (x)(w − xq + U α)] dx. The derivative Mw is closelyrelated to the Munk moment. For simplicity. which must be calculated if α changes along the length.
r]. 9. where M is the mass matrix of the material vessel and F is a generalized force.7 Relation to Wing Lift There is an important connection between the slender body theory terms involving added mass at the tail (ma (xT )). Where does the slenderbody predict the force will act? Recalling that Mw = U m33 + U xT ma (xT ).7 Relation to Wing Lift 67 9. This location will tend to stabilize the wing. The lift force from the latter is of the form L = − 1 ρU ACl w. in the sense that it acts to orient the wing parallel to the incoming ﬂow. and using the slenderbody estimate for Zw . we obtain π L = − ρs2 U w. and since m33 = 0 for a frontback symmetric wing. By listing the bodyreferenced velocities ˙ in the order s = [u.3 = −Mw . 2 the product of chord (long) and span (short). q.9. A5. v. p. we calculate for lift: Z = −ma (xT )U w π = − ρs2 U w. and so on. ˙ ˙ . w. (118) 4 Now we look at a slender body approximation of the same force: The added mass at the tail is ma (xT ) = ρs2 π/4. where A = cs. Inserting this approximation into the lift formula.8 Convention: Hydrodynamic Mass Matrix A Hydrodynamic derivatives that depend on accelerations are often written as components of a mass matrix A. the estimated lift force acts at the trailing edge. and low aspectratio wing theory. 4 Slenderbody theory is thus able to recover exactly the lift of a lowaspect ratio wing. Therefore A33 = −Zw . (117) 2 where (AR) is the aspect ratio. The lift coeﬃcient slope is approximated by (Hoerner) 1 Cl ≈ π(AR). we write (M + A)s = F .
The ﬁns can often be properly modeled using experimental data parametrized with aspect ratio and several other geometric quantities. The body travels at speed U . and ﬁns. and reference area Ar . and angle of attack α. and we now present one of them due to Jorgensen. The helical vortices can become bigger while remaining stable. and can be accurately modeled with slender body theory. 10. The blunt trailing end induces lift as a product of added mass eﬀects. this shedding is nonsymmetric. a slender body experiences transverse force due to: helical body vortices. and the rest peels away from the body. usually toward the stern of the vehicle. the approximations in the ﬁn and slenderbody analysis will break down. but eventually will split randomly. which itself is stabilizing. and are convected continuously into the wake. The normal force. the planform area.1 PRACTICAL LIFT CALCULATIONS Characteristics of LiftProducing Mechanisms At a small angle of attack. This area could be the frontal projected area. As the angle of attack becomes larger. The low pressure associated with the vortices provides the suction force. and moment coeﬃcients are deﬁned as follows: FN 1 ρU 2 Ar 2 FA 1 ρU 2 Ar 2 CN = CA = (119) . Let the body have length L. Some of it convects downstream. and are best suited to vehicles with a blunt trailing edge. vorticity sheds as if from a bluﬀbody. the blunt trailing end. and greatly increases drag by widening the wake. or the wetted area.68 10 PRACTICAL LIFT CALCULATIONS 10 10. A blunt trailing edge also induces some drag.2 Jorgensen’s Formulas There are some heuristic and theoretical formulas for predicting transverse force and moment on a body at various angles of attack. The formulas provide a good systematic procedure for design. We call the area of the stern the base area. axial force. In the limit of a 90◦ angle of attack. and there is little axial convection. The helical body vortices are stable and symmetric in this condition.
006 for slender streamlined bodies. 3 × 105 < Ren < 7 × 105 – Cd ≈ 0.3. Ab = 0 for a body that tapers to a point at the stern.2. The parameters used here are • Ab : stern base area. If “normal” Reynolds number Ren = U sin αD/ν. which is also not a constraint in the complete equations. • Ap : planform area. • CAo : axial drag at zero angle of attack.002 − 0. It depends on Re = U L/ν. equivalent to that of an inﬁnite circular cylinder. and appears explicitly in the formula for Cm . • xc : distance from the nose backwards to the center of the planform area. • : body volume.2 Jorgensen’s Formulas M xm . measured back from the nose.10. based on wetted surface area.(122) L We have listed only the formulas for the special case of circular crosssection. this location is arbitrary. both frictional and form. Ren < 3 × 105 – Cd ≈ 0. although the complete equations do account for more complex shapes. • Cdn : crossﬂow drag coeﬃcient. Further.6. we have assumed that L >> D. CAo 0. Jorgensen gives the coeﬃcients as follows: Ab α Ap sin 2α cos + Cd sin2 α Ar 2 Ar n = CAo cos2 α − Ab (L − xm ) α Ap sin 2α cos − Cdn = − Ar L 2 Ar CN = CA CM (120) (121) xm − xc sin2 α. 7 × 105 < Ren . . 1 ρU 2 Ar L 2 69 CM = The moment Mxm is taken about a point xm . – Cd ≈ 1.
if the moment is referenced to the center of the planform area. wings. These are the ﬁrst term in the normal force (CN ). Thus.70 10 PRACTICAL LIFT CALCULATIONS In the formula for normal force. First. Finally. through a sin2 αdependence. The aerodynamic center (again referenced toward the stern. we note that the second term in CM disappears if xm = xc . It contains a large amount of aerodynamic data from many diﬀerent types of vehicles. from the nose) can be found after the coeﬃcients are computed: xAC = xm + CM L. while CN is always positive. A few notations are used throughout the books. Hoerner. 10. D q CY = CYd The ﬁrst version uses the rectangular planform area as a reference. and the entire ﬁrst term in the moment (CM ). only drag forces act to create lift. the moment seeks to move the AC forward on the body. Similarly. while the second uses the square frontal area. the moment coeﬃcient is negative if the moment destabilizes the body. Hence. and other common engineering shapes. i. In both cases.3 Hoerner’s Data: Notation An excellent reference for experimental data is the twovolume set by S. we see that if Ab = 0. and are described here.. CN (123) As written.e. scaling of U 2 into the body principle directions is all that is required. modiﬁed by cos2 α. CY = CYd D/L. Two moment coeﬃcients are: . the axial force is simply the zeroα result. dynamic pressure is given as q = 1 ρU 2 . The idea here is that the fore and aft components of crossﬂow drag cancel out. but the eﬀect is moderated by the lift force. such that two typical body lift 2 coeﬃcients are: Y DLq Y = 2 . There are several terms that match exactly the slenderbody theory approximations for small α. whether or not Ab = 0.
2 The lift and moment coeﬃcients are strongly dependent on angle of attack. It follows from above that Cn·b = Cnb + Cydb /2. = ∂b Cnb = Cn·b Cyb and so on. usually in degrees. and gives two further examples of how well the slenderbody theory matches experiments. LD2 q where N is the moment taken about the body midpoint. Experiment In an experiment. acting to destabilize the vehicle.10.018/deg for a long ellipsoid. 10. Comparison of the measurements and the theory allows us to place the action point of the suction force. where b is the angle of attack. the slenderbody (pure added mass) estimates give Cnb −0.4 SlenderBody Theory vs.015/deg. Experiment 71 CN = CN1 N LD2 q N1 = . The value compares well with −0. This section gives the formula for this location in Hoerner’s notation.4 SlenderBody Theory vs. and N1 is taken about the nose. Note that the reference area for moment is the square frontal projection. it also . The following relation holds for these deﬁnitions 1 CN1 = CN + CYd . the net moment is measured.027/deg for a long cylinder and −0. Hoerner uses the notation ∂CN ∂b ∂CN1 = ∂b ∂Cy . comprising both the destabilizing part due to the potential ﬂow. and the stabilizing part due to vortex shedding or a blunt tail. ˜ For L/D > 6. and the reference length is body length L.
with trailing edge radius r. It comes out to be 1 Z = ρU 2 (πr2 )(2α). however.63 blunt 6. the moment slope referenced to the nose. (125) 2 such that the ﬁrst term in parentheses is an eﬀective area.003/deg. In the table following are values from Hoerner (p. It can be compared with a lowaspect ratio wing. proportional to LD. but it acts too close to the midpoint to add any stability there. The blunttailed geometry has a much larger lift force. The experimental lift force is typically given by CY b 0. The lift force dependence on the blunt tail is not diﬃcult to see. The point at which the viscous forces act can then be estimated as the following distance aft of the nose: ˜ Cn·b − Cnb x = (124) L Cydb The calculation uses experimental values of Cydb and Cn·b .021 0. this acts at a point on the latter half of the vehicle.60 In comparing the two body shapes.0027 for (AR) = 10 D/L. 132.0031 0.012 0. Figure 2) for a symmetric and a bluntended body. Note that the negative sign here is consistent with Hoerner’s convention that destabilizing moments have negative sign. stabilizing the angle.0093 (unstable) 0. and the second is a lift coeﬃcient.0012 (stable) 0. The slenderbody lift force associated with this end is simply the product of speed U and local added mass ma (xT ) (in our previous notation).7 0. With respect to the area πr2 .0094 (unstable) 0. both vehicles are equally unstable. Because this coeﬃcient scales roughly with wetted area.0031 (stable) 0. the lift curve slope is therefore . which achieves an equivalent lift of π(AR)/2 = 0. Consider a body.7 0.009 for L/D = 4.012 0.72 10 PRACTICAL LIFT CALCULATIONS reduces to −0. we see that the moment at the nose is much more stable (positive) for the body with a blunt trailing edge.0037 0. L/D ˜ Cnb Cyb Cydb Cn·b Cnb x/L symmetric 6. At the body midpoint.025 0. it changes little with L/D. using slenderbody theory.
acting at the tail end of the vehicle. substitution will give a lift curve slope of Cl = π (AR). the equivalent lift coeﬃcient is Cydb = 0. . due solely to the blunt end condition.0044/deg. This lift diﬀerence. and worsening further as (AR) grows.5 SlenderBody Approximation for Fin Lift 73 2/rad. overestimating the lift by about 30% when (AR) = 2.4 for the data in the table. It is inadequate for higheraspect ratio wings however. is consistent with measurements. where we made the assumption here that 2r/D = 0. The slenderbody approximation of lift as a result of bluntend conditions is Z = πs2 ρU 2 α. Expressed in terms of the Hoerner reference area D2 . where Af is the total area of the ﬁn pair. 10.10. (126) Letting the aspect ratio be (AR) = (2s)2 /Af . This is the case if the vehicle body tapers to a point where the ﬁns have their trailing edge. and is quite accurate for (AR) 1. 2 This is known as Jones’ formula.5 SlenderBody Approximation for Fin Lift Let us now consider two ﬁns of span s each.
Prandtl’s inviscid theory provides some insight. Their utility arises mainly from the high lift forces they can develop. which requires that ﬂuid particle streamlines do not wrap around the trailing edge of the surface. and a lifting surface requires circulation in order to meet the Kutta condition. 11. Lift is always deﬁned to act in a direction perpendicular to the ﬂow.2 ThreeDimensional Eﬀects: Finite Length Since all practical lifting surfaces have ﬁnite length. but instead rejoin with streamlines from the other side of the wing at the trailing edge. Circulation is the integral of velocity around the crosssection. elevators.1 Origin of Lift A lifting surface is nominally an extrusion of a streamlined crosssection: the crosssection has a rounded leading edge. the ﬂow near the ends may be threedimensional. . The theory of lifting surfaces centers on the Kutta condition. Since the separation point on the front of the section rotates with the angle of attack. More formally. leading to socalled wingtip vortices. (127) and then L = −ρU Γ. 11. Another description for the wingtip vortices is that the pressure diﬀerence across the surface simply causes ﬂow around the end. and a smooth surface. and drag in the same direction as the ﬂow. it is clear that the ﬂuid must travel faster over one side of the surface than the other. with little drag penalty. Since bound circulation cannot end abruptly at the wing end. sharp trailing edge.74 11 FINS AND LIFTING SURFACES 11 FINS AND LIFTING SURFACES Vessels traveling at signiﬁcant speed typically use rudders. This continuation causes induced velocities at the tips. This fact is true for a nonstalled surface at any angle of attack. lift arises from circulation Γ: Γ= V · ds. and other streamlined control surfaces to maneuver. and some induced drag. The reduced Bernoulli pressure this induces can then be thought of as the liftproducing mechanism. it continues on in the ﬂuid.
and has one empirical description Cl = 1 1 2π α ¯ + 1 π(AR) + 1 2π(AR)2 . measured back from the front of the wing. The eﬀective span is taken to be the length between the free ends of a symmetric wing. (130) where α is in radians. and in this case the eﬀective aspect ratio is therefore twice the physical value. The aspect ratio is a strong determinant of wing performance: for a given angle of attack. a larger aspect ratio achieves a higher lift value. which depends mainly on the aspect ratio. this fact creates both a net force and a net moment. (129) 2 where A is the singleside area of the surface. πc (131) . When ¯ AR → ∞. For high AR. α 0. xA c/4. since they are omnidirectional. (128) chord area The second representation is useful for nonrectangular control surfaces. and AR is the eﬀective aspect ratio. For angles of attack α below stall.3 Ring Fins 75 A critical parameter which governs the extent of threedimensional eﬀects is the aspect ratio: span span2 = . the eﬀective span is twice the physical value. For low AR.90. The lift generated on a surface is the result of a distributed pressure ﬁeld. A single equivalent force acts at a socalled center of action xA . but also stalls earlier.3 Ring Fins Ring ﬁns are useful when space allows. where Cl is called the lift coeﬃcient slope. The eﬀective aspect ratio for a ring of diameter d is given as AR = 4d .11. the theoretical and maximum value for Cl is 2π. If the wing is attached to a wall. Lift is written as AR = 1 L = ρU 2 ACl . xA c/2. the lift coeﬃcient Cl is nearly linear with α: Cl = Cl α. by reﬂection. and structurally more robust than cantilevered plane surfaces. 11.
76 11 FINS AND LIFTING SURFACES The eﬀective area of the ring is taken as π dc. 1 0. 2 1 and we thus have L = 2 ρU 2 Ae Cl α. where one formula for Cl is Ae = Cl = 1 .63 + π(AR) (132) (133) .
Selfpropelled conditions refer to the propeller being installed and its propelling the vessel. 12. In the second section. we do not distinguish between the magnitude of forces in model and fullscale vessels. Furthermore. an engine (diesel or gas turbine..1 PROPELLERS AND PROPULSION Introduction We discuss in this section the nature of steady and unsteady propulsion. when the hull is towed with no propeller we use a t subscript. i.2. commercial surface vessels spend the vast majority of their time operating in openwater and at constant speed. and there are two diﬀerent ﬂow conditions to consider. say) or an electric motor drives the propeller through a linkage of shafts. and bearings. we list several loworder models of thrusters. Large. not behind a hull. which have recently been used to model and simulate truly unsteady conditions. Conversely. when the propeller is run in open water.1 Basic Characteristics In the steady state. force balance in selfpropulsion requires that Rsp = Tsp . and the eﬀects of each part are important in the response of the net system. such as would be caused by a towing bar or hawser. generalization to either condition is implied.e. We use an sp subscript to indicate speciﬁcally selfpropulsion conditions.2 Steady Propulsion of Vessels The notation we will use is as given in Table 1. reducers.77 12 12. 12. In this case. there are no additional forces or moments on the vessel. we use an o subscript. (134) The gear ratio λ is usually large. When subscripts are not used. An approximation of the transient behavior of a system can be made using the quasistatic assumption. Finally. In many marine vessels and vehicles. because of similitude (using diameter D in place of L when the propeller is involved). the ﬂow around the hull interacts with the ﬂow through the propeller. steady propulsion conditions are generally optimized for fuel eﬃciency. The following relations deﬁne the gearbox: . indicating that the propeller turns much more slowly than the driving engine or motor.
ρnp D KT = (136) (137) . note that in the wake of the vessel. We call J = Up /np D the advance ratio of the prop when it is exposed to a water speed Up . Up may not be the same as the speed of the vessel U . A propeller operating in open water can be characterized by two nondimensional parameters which are both functions of J: To (thrust coeﬃcient) ρn2 D4 p Qp KQ = 2 o 5 (torque coeﬃcient). for any ﬂow condition. (135) and power follows as Pp = ηg Pe .78 Rsp Rt T ne nm np λ Qe Qp ηg Pe Pp D U Up Qm f fm N N N Hz Hz Hz 12 PROPELLERS AND PROPULSION hull resistance under selfpropulsion towed hull resistance (no propeller attached) thrust of the propeller rotational speed of the engine maximum value of ne rotational speed of the propeller gear ratio Nm engine torque Nm propeller torque gearbox eﬃciency W engine power W propeller shaft power m propeller diameter m/s vessel speed m/s water speed seen at the propeller Nm maximum engine torque kg/s fuel rate (or energy rate in electric motor) kg/s maximum value of f Table 1: Nomenclature ne = λnp Qp = ηg λQe .
0 0. A typical wake fraction of 0. for example.0 0. 2πnp Qpo 2πKQ (138) This eﬃciency divides the useful thrust power by the shaft power. γ2 ] are usually positive. (139) As written. β2 . and therefore ﬁt the approximate form: KT (J) = β1 − β2 J KQ (J) = γ1 − γ2 J. the four coeﬃcients [β1 .06 K Q 0.6 J 0.00 0. In practical terms. The propeller is operating in a wake.08 0. We next introduce three factors useful for scaling and parameterizing our mathematical models: • Up = U (1 − w).1.6 0.8 1.0 0.2 0. w is referred to as the wake fraction.02 0. γ1 . Thrust and torque coeﬃcients are typically nearly linear over a range of J.4 0. Behind .12. indicates that the incoming velocity seen by the propeller is only 90% of the vessel’s speed.0 K Q Figure 4: Typical thrust and torque coeﬃcients.04 0.10 0. the wake fraction comes about this way: Suppose the open water thrust of a propeller is known at a given U and np . as shown in the ﬁgure.8 K η T 0 η0 0.2 KT 0. The open propeller eﬃciency can be written then as ηo = To U J(U )KT = .2 Steady Propulsion of Vessels 79 1.4 0.
translates selfpropelled torque to open water torque. The ratio (1 − t)/(1 − w) is often called the hull eﬃciency. The wake fraction can also be estimated by making direct velocity measurements behind the hull. and thus that ηo is the open propeller eﬃciency at this speed. The thrust deduction is particularly useful. • Rt = Rsp (1 − t). it is obviously speciﬁc to both the hull and the propeller(s). The rotative eﬃciency ηR . A common measure of eﬃciency. thrust T . as well as the turbulence induced by the hull. and we see that a . which may be greater than one. the prop creates some extra thrust. It follows that To (Up ) = Tsp . is based on the towed resistance. t is a small positive number. and the selfpropelled torque.80 12 PROPELLERS AND PROPULSION a vessel moving at speed U . The quasipropulsive eﬃciency can be greater than one. which was used to complete the above equation. with no propeller. with 0. w scales U at the prop and thus J. for the same incident velocity Up . • Qpo = ηR Qpsp . and rotation rate np . and with the propeller spinning at the same np . Often. (1 − w) ηQP = (140) To and Qpo are values for the inﬂow speed Up . the quasipropulsive eﬃciency. Note that in comparison with the wake fraction. a propeller will increase the resistance of the vessel by creating lowpressure on its intake side (near the hull). In this case. w is then chosen so that the open water thrust coeﬃcient matches what is observed. t is called the thrust deduction even though it is used to model resistance of the hull. which makes Rsp > Rt . Rt U 2πnp Qpsp To (1 − t)Up ηR = 2πnp (1 − w)Qpo (1 − t) = ηo ηR .2 as a typical value. if only the towed resistance of a hull is known. rotative eﬃciency equalizes torque instead of thrust. ηR is meant to account for spatial variations in the wake of the vessel which are not captured by the wake fraction. and can be estimated from published values. and how they interact. since it relies on the towed resistance and in general Rt > Rsp .
12. The vessel speed can be computed by recalling that J(U ) = U/np D and Up = U (1 − w).2 Steady Propulsion of Vessels 81 small thrust deduction t and a large wake fraction w are beneﬁcial eﬀects. depending only on the propeller open characteristics. Equating the selfpropelled thrust and resistance then gives Tsp = Rsp To = Rt /(1 − t) 1 ρCr Aw U 2 KT (J(Up ))ρn2 D4 (1 − t) = p 2 2 Up 1 (β1 − β2 J(Up ))ρn2 D4 (1 − t) = ρCr Aw p 2 (1 − w)2 Cr Aw β1 − β2 J(Up ) = J(Up )2 2 (1 − t)(1 − w)2 2D δ J(Up ) = −β2 + 2 β2 + 4β1 δ .2. 12.2 Solution for Steady Conditions The linear form of KT and KQ (Equation 140) allows a closedform solution for the steadyoperating conditions.12. (141) 2 where Cr is the resistance coeﬃcient (which will generally depend on Re and F r).2. Suppose that the towed resistance is of the form 1 Rt = ρCr Aw U 2 . which necessitates a model of the drive engine or motor. This requires a torque equation. and Aw is the wetted area. and on the hull.3 Engine/Motor Models The torquespeed maps of many engines and motors ﬁt the form . but it is clear that we need now to ﬁnd np . but which are in competition. 2δ (142) The last equation predicts the steadystate advance ratio of the vessel. A high rotative eﬃciency and open water propeller eﬃciency (at Up ) obviously contribute to an eﬃcient overall system.
(145) . For example. f2 . More speciﬁcally. gas turbines roughly ﬁt the curves shown in the ﬁgure (Rubis). nm /λ F (f /ff m. (143) where F () is the characteristic function. f3 .82 12 PROPELLERS AND PROPULSION Q e /Q m f3 f2 f1 n e /n m Figure 5: Typical gas turbine engine torquespeed characteristic for increasing fuel rates f1 . The manipulations begin by equating the engine and propeller torque: Qpo (J(Up )) 2 5 ρnp D KQ (J(Up )) ρn2 D5 (γ1 − γ2 J(Up )) p n2 p (nm /λ)2 = ηR Qps p (J(Up )) = ηR ηg λQe = ηR ηg λQm F (f /fm . if F () has the form f ne f +b + c +d fm nm fm np = −α1 + α2 . ne /nm ) = − a (144) then a closedform solution for ne (and thus np ) can be found. Qe = Qm F (f /fm . ne /nm ) = ρD5 (γ1 ηR ηg λQm np + α2 −α1 2 − γ2 J(Up ))(nm /λ) (nm /λ) 2 α2 1 − α1 + np = (nm /λ) + 4 α2 2 . ne /nm ).
the transient behavior of the propulsion system needs to be considered. Instead. (147) (148) where Ip is the total (material plus ﬂuid) inertia reﬂected to the prop.the propeller spins at ωp radians per second. is given by (m + ma )u = Tsp − Rsp ˙ 2πIp np = ηg λQe − Qps p . since the positioning problem has been central to bluﬀ vehicles with multiple electric thrusters. This model requires the identiﬁcation of three parameters: Ip . this is left as a problem for the reader. The diﬀerential equation in ωp pits the torque delivered by the motor against a quadraticdrag type loss which depends on rotation speed. We use the subscript m to denote a quantity in the motor. 12. whose bandwidth depends directly on Qm . although there have been some very successful models in recent years. The torque equation at the propeller and the thrust relation are Ip ωp = λQm − Kω ωp ωp  ˙ T = Ct ωp ωp . The thrust is then given as a static map directly from the rotation speed. It should be pointed out that the models described below all pertain to openwater conditions and electric motors. .1 OneState Model: Yoerger et al. The dynamic response of the coupled propulsion and ship systems. It is a ﬁrstorder.3 Unsteady Propulsion Models When accurate positioning of the vehicle is critical.12. under the assumption of quasistatic propeller conditions. and Ct . lowpass ﬁlter from Qm to T .3. the quasistatic assumption used above does not suﬃce. nonlinear. 12. and p for the propeller. The problem of unsteady propulsion is still in development.3 Unsteady Propulsion Models 83 Note that the fuel rate enters through both α1 and α2 . Kω . ˙ (146) Making the necessary substitutions creates a nonlinear model with f as the input.
(150) Here A is the disc area of the tunnel. Kω represents losses in the motor due to spinning (friction and resistive). This model has the advantage that it creates a thrust overshoot for a step input. which invokes lift and drag coeﬃcients. The second dynamic equation is for the ﬂuid velocity at the propeller: ˙ ρALγ Up = −ρA∆β(Up − U )Up − U  + T. As with the onestate model of Yoerger et al. or the propeller disk diameter if no tunnel exists. L denotes the length of the tunnel. and Kv is the gain on the input voltage (so that the current ampliﬁer is included in Kv ).3.2 12 PROPELLERS AND PROPULSION TwoState Model: Healey et al. these formulae are static maps. Together. ρALγ is the added mass that is accelerated by the blades. ˙ (149) Here. The thrust and torque of the propeller are approximated using wing theory. as well as an eﬀective angle of attack and the propeller pitch.84 12. this mass is always nonzero. The parameter ∆β is called the diﬀerential momentum ﬂux coeﬃcient across the propeller. even if there is no tunnel. and up to 2. and γ is the eﬀective added mass ratio. The twostate model includes the velocity of a mass of water moving in the vicinity of the blades. It can accommodate a tunnel around the propeller. it may be on the order of 0.0 for open propellers. is referenced to the motor and given as Im ωm = −Kω ωm + Kv V − Qp /λ.. The torque equation. However. and therefore introduce no new dynamics. similarly to the above. .2 for propellers with tunnels. this version requires the identiﬁcation of the various coeﬃcients from experiments. which is very common in thrusters for positioning. which is in fact observed in experiments.
The cable itself is an important factor in the behavior of the complete towed system. We will not discuss light systems speciﬁcally here. the rationale is that gravity will tend to keep the cable vertical and make the deployment robust against currents and towing speed. which can generally be handled with the same tools as other vehicles. as in a towed array. if they are to be supported by a cable.. The extreme L/D ratio for these cables obviates any bending stiﬀness eﬀects. Modern cables can easily exceed 5000m in length. Cable systems come in a variety of conﬁgurations.1 Statics Force Balance For the purposes of deriving the static conﬁguration of a cable in a ﬂow. Some basic guidelines for vehicle design are given at the end of this section. The cables are generally circular in cross section.. slenderbody theory. i. operates at low speed. For example. however.1 13. and must have large propulsors to control the tether if there are currents. On the other hand. etc. towed vehicles may be quite heavy in water.e. and one main division may be made simply of the density of the cable. but rather look at heavy systems. Some towed vehicles might have active lifting surfaces or thrusters for attitude control. wing theory. employ a heavy cable and possibly a heavy weight. Most of the analysis can be adapted to either case. we concentrate on cable mechanics more than vehicle characteristics. Tension and hydrostatic . and do not have to be selfpropelled. or a vehicle capable of maneuvering itself. The towed array is a relatively highvelocity system that nominally streams out horizontally behind the vessel. and may carry power conductors and multiple communication channels (ﬁber optic).85 13 TOWING OF VEHICLES Vehicles which are towed have some similarities to the vehicles that have been discussed so far. and usually need good directional stability. such as a remotelyoperated vehicle. on the other hand. and in this section. Heavy systems. 13. linearization. even a heavy steel cable with 2cm diameter. The heavy systems will generally transmit surface motions and tensions to the towed vehicle much more easily than lighttether systems.1. we assume for the moment that that it is inextensible. in contrast. towed vehicles are often streamlined. An ROV. Lighttether systems are characterized by neutrallybuoyant (or nearly so) cables. with either a minimal vehicle at the end.
Wn cos φ + Rn . which we take to be zero at the bottom end of the cable. • T (s): local tension. the ﬂuid velocity vector. leading to . Force balance in the tangential and normal coordinates gives two coupled equations for T and φ: dT amp. In both cases. is to be projected onto the relevant axes. and the normal drag scales with a crossﬂow drag coeﬃcient Cn . • Rt (s): external tangential force. We employ the curvilinear axial coordinate s. The freebody diagram shown has the following components: • Wn : net inwater weight of the cable per unit length. the tangential drag is controlled by a frictional drag coeﬃcient Ct . Wn sin φ − Rt ds dφ T amp. • φ(s): local inclination angle. ds (151) (152) The external forces are primarily ﬂuid drag. per unit length.86 13 TOWING OF VEHICLES y x s+ ds ds Rt φ T+dT U ds Rn ds Wn T pressure will elongate a cable. upwards along the cable is the positive direction. = amp. U horizontal toward the left. • Rn (s): external normal force. but the eﬀect is usually a small percentage of the total length. per unit length. = amp.
φ. provides both T (0) and φ(0). = amp. y) deﬁnes the cable conﬁguration. with T (0) equal to the inwater weight of the vehicle.2 Critical Angle For very deep systems. 2 (153) (154) Note that we simpliﬁed the drag laws from a usual form vv to v 2 . to meet the bottom boundary condition. and can be approximated easily. The uniform angle is called the critical angle φc . the total weight of cable will generally exceed that the vehicle. For example.1 Statics 87 1 Rt = − ρCt dU 2 cos2 φ 2 1 Rn = − ρCn dU 2 sin2 φ. cos φ ds dy amp. since as drawn.1. and the common case is that a force balance on the vehicle. With regard to Cartesian coordinates x. x.13. 13. ρCn dU 2 δ= . sin φ. In the straight part of the cable. normal weight and drag components are equalized. This gives rise to a conﬁguration in which the cable is straight for a majority of its length. but turns as necessary at the vehicle end. weight. a very heavy but lowdrag vehicle will impose φ(0) π/2. The equations for T and φ can be integrated along the cable coordinate s to ﬁnd the cable’s static conﬁguration. say U is a function of y. y. ds (155) (156) The simultaneous integration of all four equations (T. dominated by drag. = amp. Two boundary conditions are needed. Let the relative importance of weight be given as Wn . the cable conﬁguration follows dx amp. and current dependency may be included. and the cable tension. 0 ≤ φ ≤ π/2.
we solve a quadratic equation and keep only the positive solution: cos φc = √ δ 2 + 1 − 1. φc amp.1. The cables are generally circular in cross section. the same approximation gives . and usually need good directional stability. The towed array is a relatively highvelocity system that nominally streams out horizontally behind the vessel.e. wing theory. On the other hand. Lighttether systems are characterized by neutrallybuoyant (or nearly so) cables. We will not discuss light systems speciﬁcally here. employ a heavy cable and possibly a heavy weight. For example. Some basic guidelines for vehicle design are given at the end of this section. as in a towed array. in contrast. Heavy systems. δ is large. even a heavy steel cable with 2cm diameter. Modern cables can easily exceed 5000m in length. 13.. Tension and hydrostatic so that the condition dφ/ds = 0 requires from the force balance 1 2 sin φc = 0. (158) For a very light cable. 2 2δ cos φc amp. and the linear approximation of the square root 1 + ≈ 1 + /2 gives 1 −→ 2δ π 1 amp. which can generally be handled with the same tools as other vehicles. and may carry power conductors and multiple communication channels (ﬁber optic). Some towed vehicles might have active lifting surfaces or thrusters for attitude control.1 13. we assume for the moment that that it is inextensible. amp. towed vehicles are often streamlined. The cable itself is an important factor in the behavior of the complete towed system. Cable systems come in a variety of conﬁgurations. linearization. The heavy systems will generally transmit surface motions and tensions to the towed vehicle much more easily than lighttether systems. An ROV. such as a remotelyoperated vehicle. and must have large propulsors to control the tether if there are currents. we concentrate on cable mechanics more than vehicle characteristics. with either a minimal vehicle at the end. and do not have to be selfpropelled. δ is small. and in this section. if they are to be supported by a cable. the rationale is that gravity will tend to keep the cable vertical and make the deployment robust against currents and towing speed. The extreme L/D ratio for these cables obviates any bending stiﬀness eﬀects. (157) In the case of a √ very heavy cable. Substituting sin2 φc = 1−cos2 φc . but rather look at heavy systems. on the other hand. or a vehicle capable of maneuvering itself. Most of the analysis can be adapted to either case. etc. towed vehicles may be quite heavy in water.1 Statics Force Balance For the purposes of deriving the static conﬁguration of a cable in a ﬂow. however.. 2 δ cos φc − We are considering the case of 0 < φc < π/2.88 13 TOWING OF VEHICLES 85 13 TOWING OF VEHICLES Vehicles which are towed have some similarities to the vehicles that have been discussed so far. slenderbody theory. and one main division may be made simply of the density of the cable. − . operates at low speed. i.
t) = T (s) + T (s.  13.45 0. 0. 1.1 Linearized Dynamics Derivation The most direct procedure for deriving useful linear dynamic equations for a planar cable problem is to consider the total tension and angle as made up of static parts summed with dynamic parts: ¯ ˜ T (s.0 amp.57 amp.32 amp. δ << 1 0. t) ¯ ˜ φ(s. = amp.13.47 amp. 1 ∂q amp.2. exact amp.63 0.5 amp.0 amp. 0. 1. 1. ∂t ∂s ∂s 2 ∂t ¯ ˜ ∂2q ¯ ˜ ¯ ˜ ∂ φ + ∂ φ − Wn cos(φ + φ) + (m + ma ) 2 amp. We also write the axial deﬂection with respect to the static conﬁguration as ˜ p(s. and the lateral deﬂection q(s. = amp. 0.2 amp. t). 1. 1. . amp. ρCn d U sin φ − 2 ∂t 2 2 . t).2 Linearized Dynamics 89 cos φc 1 − δ −→ φc √ 2δ. It follows that φ = ∂q/∂s. t) = φ(s) + φ(s.41 2. (T + T ) ∂t ∂s ∂s amp. and the approximations.00 1. 1. δ amp. The table below gives some results of the exact solution. 5. Now augment the two static conﬁguration equations with inertial components: ¯ ∂T ˜ ∂T 1 ∂p ∂2p ¯ ˜ + − Wn sin(φ + φ) − ρCt d U cos φ + m 2 amp.1 amp.47 amp.61 amp.91 amp. δ >> 1 amp. 1.2 13. 0. t). 0.14 amp.44 amp.33 amp.0 amp. amp. 0. 1.07 amp.
To a linear approximation. (m + ma ) 2 + bn ∂t ∂t ∂s ∂s m ¯ ¯ where we made the substitution bt = ρCt dU cos φ and bn = ρCn dU sin φ. in and out . ∂s ∂s The static solution cancels out of both governing equations. An additional weight term Wn sin φ∂q/∂s also appears. T 2 + Wn sin φ . = amp. and can develop longitudinal resonant conditions (next section). and keeping only linear terms we obtain ∂2p ∂2p ¯ ∂q ¯ ∂p amp. Note that avoiding the drag law form vv again. T 2 + EA ∂t ∂s ∂s ∂s ∂s ∂t m The axial dynamics (p) couples with the lateral equation through the weight ¯˜ term −Wn cos φφ. = amp. ¯ Now we perform the trigonometry substitutions in the weight terms. = amp. = amp. and substitute the constitutive (Hooke’s) law ˜ ∂T ∂2p = EA 2 . and its transverse added mass is ma . The lateral dynamics (q) couples with the axial through the ¯ ˜ ¯ term T ∂ φ/∂s. heavy cables easily transmit motions and tensions along their length. Because of light damping in the tangential direction. EA 2 ∂t2 ∂t ∂s 2 2 ∂ q ∂q ¯ ∂q ¯∂ q amp. say U = 0. the lateral cable motions are heavily damped. let φ φ for the calculation of drag. In contrast. The nature of the lateral response. (m + ma ) 2 amp. The uncoupled dynamics are both in the form of damped wave equations ∂2p ∂p ∂2p + bt amp. then equivalent linearization can be used for the quadratic drag.90 13 TOWING OF VEHICLES Here the material mass of the cable per unit length is m. EA 2 − Wn cos φ − ρCt dU cos φ 2 ∂t ∂s ∂s ∂t 2 2 ¯ ∂ q ∂p ∂ φ ¯ ∂q ¯ ∂q ¯∂ q + Wn sin φ − ρCn dU sin φ . If it is not the case. the outofplane vibrations of a cable are also governed by the second equation above. such that disturbances only travel a few tens or hundreds of meters before they dissipate. we have implicitly assumed that U cos φ > ∂p/∂t and U sin φ > ∂q/∂t.
Highfrequency vessel motions in the horizontal plane are completely missed by the vehicle. is a very slow. lateral amp. we impose the motion p(L. Mode Shape. ¨ (161) ∂s These top and bottom behaviors comprise the boundary conditions for the wave equation. 13. amp. Ct O(0. amp. nπ L EA m amp. and primes to indicate spatial derivatives. SLOW amp. so that ˜ EA . Bessel function amp. is an undamped mass responding to the local tension variations: ∂p(0. while lowfrequency motions occur sluggishly.2 Damped Axial Motion amp.2 Linearized Dynamics 91 of the towing plane. O nπ L ¯ T (L/2) m+ma mode shape amp. t). ¯ T (s) m+ma amp.2. Consider a long cable governed by the damped wave equation m¨ + bt p = EAp p ˙ (159) We use overdots to indicate time derivatives. (160) while the towed vehicle. We let p(s. damped nonlinear ﬁlter. since it is lightly damped and forced by the heaving of vessels in seas. EA m amp. t) = P cos ωt. t) = M p(0.01) disturbances amp. YES cable? amp. axial wave speed amp. NO amp. travel down amp. At the surface. sine/cosine damping amp. t) = p(s) cos ωt. at the lower end.13. and only after a signiﬁcant delay time. Cn O(1) amp. The axial direction is of particular interest. FAST natural frequency amp.
It is possible to compute the dynamic tension via T = EA˜ . This is normally problematic at the top of the cable. and investigating the singularity of p. where the static tension is highest. c1 + δc2 . ˜ . k= P amp. The natural frequency can be found by letting bt = 0. We obtain p δ sin ks + cos ks ˜ T = −EAP k . = amp. = amp. respectively.92 13 TOWING OF VEHICLES p + ˜ mω 2 − iωbt p = 0. These can be combined to give the solution δ cos ks − sin ks . In general. for which δ cos kL = sin kL. (163) EA Note that k is complex when bt = 0. the cable will unload completely and then reload with extremely high impulsive forces. c1 cos kL + c2 sin kL 0 amp. p=P ˜ ˜ Dynamic Tension. Natural Frequency. (164) δ cos kL − sin kL In the case that M → 0. δ cos kL − sin kL There are two dangerous situations: (165) ¯ ˜ • The maximum tension is T + T  and must be less than the working load of the cable. The top and bottom boundary conditions give. where T is low. simplifying the result to p = ˜ P cos ks/ cos kL. ˜ ¯ • If T  > T . ˜ EA (162) This admits the solution p(s) = c1 cos ks + c2 sin ks. where ˜ mω 2 − iωbt . the scalar δ → ∞. it occurs ¯ primarily at the vehicle. where δ = EAk/ω 2 M . This is known as snap loading.
g. it may be undesirable to expose the vehicle to these motions. the system is dominated by the vehicle mass. The natural frequencies near wave excitation at the surface vessel must be taken into account in any design or deployment.e. a clump weight below which a light cable is employed. In very deep water. i. and the mass that is oscillating is M +mL/2. then ω= EA/L . ρc ω A few examples are given below for a steel cable with E = 200 × 109 P a. 6 Employing the deﬁnition for δ. if ρc = m/A is the density of the cable. heave compensation through an active crane.2 Linearized Dynamics 93 kL << 1.. and an Sshaped length of cable at the bottom formed with ﬂotation balls.13. and recalling that ω 2 = k 2 EA/m. SWATH). .. Some solutions in use today are: stable vessels (e. Even if a cable can withstand the eﬀects of resonance. We start with better approximations for sin() and cos(): (kL)2 δ 1− 2 (kL)2 = kL 1 − . we arrive at mL (kL)2 1− M 2 (kL)2 . 6 = (kL)2 1 − If we match up to second order in kL. Some higher order terms need to be kept. which is only a correct answer if M >> mL. but we ﬁnd that a ﬁrstorder approximation yields ω = EA/LM . we have the approximation 1 L 2E . and ρc = 7000kg/m3 . the eﬀects of mL/2 dominate. M + mL/2 This has the familiar form of the square root of a stiﬀness divided by a mass: the stiﬀness of the cable is EA/L.
so that larger lengths of cable are needed to reach a given depth. and the towed system lags further and further behind the surface vessel.20 for a large range of Re. 13.7rad/s amp. including obvious fatigue when the amplitudes and frequencies are high. 3. 7. and nearer . Strumming of amplitude d/2 or greater can occur for 0.0.4 Vehicle Design The physical layout of a towed vehicle is amenable to the analysis tools of selfpropelled vehicles.3 Cable Strumming Cable strumming causes a host of problems.0rad/s amp. Here are basic guidelines to be considered: 1. 2. The book by Blevins is a good general reference.2 for a nonoscillating cable to as high as 3.6rad/s amp. The most noteworthy issue with towing is that the vibrations may cause the normal drag coeﬃcient Cn to increase dramatically – from about 1. with the main exceptions that the towpoint presents a large mean force as well as some disturbances. This drag penalty decreases the critical angle of towing. for towing stability reasons. ωn = 15. where S is the Strouhal number. The preferred frequency of vortex formation is given by the empirical relation ωS = 2πSU/d.160. and that the vehicle can be quite heavy in water. Strumming of cables is caused by the proximity of a preferred vortex formation frequency ωS to the natural frequency of the structure ωn .5. The static tension will increase accordingly as well.6 < ωS /ωn < 2. for basic roll and pitch stability. This latter frequency can be obtained as a zero of the lightlydamped Bessel function solution of the lateral dynamics equation above. The towpoint must be located above the vehicle center of inwater weight. The combined center of mass (material and added mass) should be longitudinally between the towpoint and the aerodynamic center. 3. The towpoint should be forward of the aerodynamic center. 1.5rad/s 13.94 L = 500m 1000m 2000m 5000m 13 TOWING OF VEHICLES amp. about 0.
This will ensure that highfrequency disturbances do not induce excessive pitching. The center of buoyancy should be behind the inwater center of weight. away from the surface. 5. so that the vehicle enters the water ﬁns ﬁrst. and selfstabilizes with U > 0. The towpoint should be longitudinally forward of the center of inair weight. and hence the net lift force is downward.13. 4. For this reason. Meeting all of these criteria simultaneously is no small feat. fullscale experiments are commonly used in the design process. . so that the vehicle pitches downward at small U .4 Vehicle Design 95 the towpoint. and the performance of the device is very sensitive to small perturbations in the geometry.
The constants −zi are called the zeros of the transfer function or signal. it is a fact that Y (s) is equivalent to Y (s) = a2 an a1 + + ··· . if this is not the case.2 Partial Fractions: Unique Poles Under the condition m < n. then we have to divide the numerator by the denominator as necessary to ﬁnd a simple form. 14.1 Partial Fractions Solving linear timeinvariant systems by the Laplace Transform method will generally create a signal containing the (factored) form Y (s) = K(s + z1 )(s + z2 ) · · · (s + zm ) . Viewed in the complex plane. The coeﬃcient ai is termed the residual associated with the i’th pole. and −pi are the poles. Partial fraction expansions alter the form of Y (s) so that the simple transform pairs can be used to ﬁnd the timedomain output signals. 14. We must have m < n. it is clear that the magnitude of Y (s) will go to zero at the zeros. later we will see that dynamic systems are described in the same format: in that case we call the impulse response G(s) a transfer function. A system transfer function is identical to its impulse response. and once all these are found it is a simple matter to go back to the transform table and look up the timedomain responses. in the context of control systems. . since L(δ(t)) = 1.96 14 TRANSFER FUNCTIONS & STABILITY 14 TRANSFER FUNCTIONS & STABILITY The reader is referred to Laplace Transforms in the section MATH FACTS for preliminary material on the Laplace transform. s + p1 s + p2 s + pn (167) in the special case that all of the poles are unique and real. Partial fractions are presented here. (s + p1 )(s + p2 ) · · · (s + pn ) (166) Although for the moment we are discussing the signal Y (s). as the fundamental link between pole locations and stability. and to inﬁnity at the poles.
but this condition rarely occurs in applications. Note that in this example. 14.3 Example: Partial Fractions with Unique Real Poles G(s) = s(s + 6) e−2s . we were able to apply the derivative operator s after expanding the partial fractions.14. evaluate them at s = −pi . we can initially work with G(s)/se−2s . i. 5 5 The impulse response is needed to account for the step change at t = 2. The more traditional method. m ≥ n + 1. We have s+6 a1 a2 = + . exempliﬁed by Ogata. . For cases where a second derivative must be taken. (s + 4)(s − 1) Since we have a pure delay and m = n. giving (s + 4)(s − 1) s+4 s−1 2 5 a1 = a2 = (s+6)(s+4) (s+4)(s−1) s=−4 (s+6)(s−1) (s+4)(s−1) s=1 = − = 7 5 Thus 2 7 L−1 (G(s)/se−2s ) = − e−4t + et −→ 5 5 8 7 g(t) = δ(t − 2) + e−4(t−2) + et−2 .e. special care should be used when accounting for the signal slope discontinuity at t = 0. may prove easier to work through.3 Example: Partial Fractions with Unique Real Poles 97 How to ﬁnd ai ? A simple rule applies: multiply the righthand sides of the two equations above by (s + pi ). the only one left. The case of repeated real roots may be handled elegantly.. and solve for ai .
98
14 TRANSFER FUNCTIONS & STABILITY
14.4
Partial Fractions: ComplexConjugate Poles
A complexconjugate pair of poles should be kept together, with the following procedure: employ the form Y (s) = b1 s + b2 a3 + + ···, (s + p1 )(s + p2 ) s + p3 (168)
where p1 = p∗ (complex conjugate). As before, multiply through by (s + 2 p1 )(s + p2 ), and then evaluate at s = −p1 .
14.5
Example: Partial Fractions with Complex Poles
G(s) = s+1 b1 s + b2 a3 = + : s(s + j)(s − j) (s + j)(s − j) s s+1 s
s=−j
= [b1 s + b2 ]s=−j −→
1 + j = −b1 j + b2 −→ b1 = −1 b2 = 1; also s+1 (s + j)(s − j) = a3 = 1.
s=0
Working out the inverse transforms from the table of pairs, we have simply (noting that ζ = 0) g(t) = − cos t + sin t + 1(t).
14.6
Stability in Linear Systems
In linear systems, exponential stability occurs when all the real exponents of e are strictly negative. The signals decay within an exponential envelope. If one exponent is 0, the response never decays or grows in amplitude; this is called marginal stability. If at least one real exponent is positive, then one element of the response grows without bound, and the system is unstable.
14.7 Stability ⇐⇒ Poles in LHP
99
14.7
Stability ⇐⇒ Poles in LHP
In the context of partial fraction expansions, the relationship between stability and pole locations is especially clear. The unit step function 1(t) has a pole at zero, the exponential e−at has a pole at −a, and so on. All of the other pairs exhibit the same property: A system is stable if and only if all of the poles occur in the left half of the complex plane. Marginally stable parts correlate with a zero real part, and unstable parts to a positive real part.
14.8
General Stability
There are two deﬁnitions, which apply to systems with input u(t) and output y(t). 1. Exponential. If u(t) = 0 and y(0) = yo , then yi (t) < αe−γt , for ﬁnite α and γ > 0. The output asymptotically approaches zero, within a decaying exponential envelope. 2. BoundedInput BoundedOutput (BIBO). If y(0) = 0, and fi (t) < γ, γ > 0 and ﬁnite, then yi (t) < α, α > 0 and ﬁnite. In linear timeinvariant systems, the two deﬁnitions are identical. Exponential stability is easy to check for linear systems, but for nonlinear systems, BIBO stability is usually easier to achieve.
100
15 CONTROL FUNDAMENTALS
15
15.1
15.1.1
CONTROL FUNDAMENTALS
Introduction
Plants, Inputs, and Outputs
Controller design is about creating dynamic systems that behave in useful ways. Many target systems are physical; we employ controllers to steer ships, ﬂy jets, position electric motors and hydraulic actuators, and distill alcohol. Controllers are also applied in macroeconomics and many other important, nonphysical systems. It is the fundamental concept of controller design that a set of input variables acts through a given “plant” to create an output. Feedback control then uses sensed plant outputs to apply corrective inputs: Plant Jet aircraft Marine vessel Hydraulic robot U.S. economy Nuclear reactor 15.1.2 Inputs Outputs Sensors elevator, rudder, etc. altitude, hdg altimeter, GPS rudder angle heading gyrocompass valve position tip position joint angle fed interest rate, etc. prosperity inﬂation, M1 cooling, neutron ﬂux power level temp., pressure
The Need for Modeling
Eﬀective control system design usually beneﬁts from an accurate model of the plant, although it must be noted that many industrial controllers can be tuned up satisfactorily with no knowledge of the plant. Ziegler and Nichols, for example, developed a general recipe which we detail later. In any event, plant models simply do not match realworld systems exactly; we can only hope to capture the basic components in the form of diﬀerential or integrodiﬀerential equations. Beyond prediction of plant behavior based on physics, the process of system identiﬁcation generates a plant model from data. The process is often problematic, however, since the measured response could be corrupted by sensor noise or physical disturbances in the system which cause it to behave in unpredictable ways. At some frequency high enough, most systems exhibit eﬀects that are diﬃcult to model or reproduce, and this is a limit to controller performance.
15.2 Representing Linear Systems 15.1.3 Nonlinear Control
101
The bulk of this subject is taught using the tools of linear systems analysis. The main reason for this restriction is that nonlinear systems are diﬃcult to model, diﬃcult to design controllers for, and diﬃcult overall! Within the paradigm of linear systems, there are many sets of powerful tools available. The reader interested in nonlinear control is referred to the book by Slotine and Li.
15.2
Representing Linear Systems
Except for the most heuristic methods of tuning up simple systems, control system design depends on a model of the plant. The transfer function description of linear systems has already been described in the discussion of the Laplace transform. The statespace form is an entirely equivalent timedomain representation that makes a clean extension to systems with multiple inputs and multiple outputs, and opens the way to standard tools from linear algebra. 15.2.1 Standard StateSpace Form
We write a linear system in a statespace form as follows x = Ax + Bu + Gw ˙ y = Cx + Du + v where • x is a state vector, with as many elements as there are orders in the governing diﬀerential equations. • A is a matrix mapping x to its derivative; A captures the natural dynamics of the system without external inputs. • B is an input gain matrix for the control input u. • G is a gain matrix for unknown disturbance w; w drives the state just like the control u. • y is the observation vector, comprised mainly of a linear combination of states Cx (where C is a matrix. (169)
the transfer function (or transfer matrix) can be written as G(s) = y(s) = C(sI − A)−1 B + D.3 Converting a Transfer Function into a StateSpace Model It may be possible to write the corresponding diﬀerential equation along one row of the state vector. For example. and clearly y(s)/v(s) = I. A similar equation holds for y(s)/w(s). In the case of no disturbances or noise. consider the following system: my (t) + by (t) + ky(t) = u (t) + u(t) (massspringdashpot) s+1 G(s) = ms2 + bs + k . u(s) (170) where I is the identity matrix with the same size as A. which can create the same transfer function. w G + + + A x’ x + v + y + 1/s C u B D 15.2. 15. • v is an unknown sensor noise which corrupts the measurement.2 Converting a StateSpace Model into a Transfer Function There are a number of canonical statespace forms available.2. and then cascade derivatives.102 15 CONTROL FUNDAMENTALS • Du is a direct map from input to output (usually zero for physical systems).
ki U (s) = kp + + k d s (171) E(s) s 1 + τd s . y]T . which helps to reduce overshoot. = kp 1 + τi s where the last line is written using the conventions of one overall gain kp . and e is the error signal it receives. the derivative action scales with the derivative of the error. The integrator is accumulating the error signal over time. leading to an entry of 1 in the oﬀdiagonal of the second row in A. The common variations are: P . the system response to u is the same as the derivative of the system response to u.15. 15. we obtain the system dx −b/m −k/m = 1 0 dt y = [1 1] x x+ 1/m 0 u 103 Note speciﬁcally that dx2 /dt = x1 . but is also subject to an unknown .3 PID Controllers Setting x = [y . P D. If u is the output from the controller. P I. this control law has the form t u(t) = kp e(t) + ki C(s) = 0 e(τ )dτ + kd e (t).3 PID Controllers The most common type of industrial controller is the proportionalintegralderivative (PID) design. In words. Entries in the Cmatrix are easy to write in this case because of linearity. Finally. and so the control action from this part will continue to grow as long as an error exists. 15. the proportional part of this control law will create a control action that scales linearly with the error – we often think of this as a springlike action. The controller will retard motion toward zero error.4 Example: PID Control Consider the case of a mass (m) sliding on a frictionless table. plus a time characteristic to the integral part (τi ) and and time characteristic to the derivative part (τd ). It has a perfect thruster that generates force u(t). P ID.
with initial conditions y(0) = yo and y (0) = 0. and kd the part of the dashpot. so that the closedloop dynamics follow my (t) = −kp y(t) + d(t). 15. The result. and it is perfectly measured. and so the rule is kd < 2 mkp .2 ProportionalDerivative Only kp t. 2mωd y(t) = yo e 2m t cos ωd t + −kd . The system now resembles a secondorder massspringdashpot system where kp plays the part of the spring. we see that y(t) = yo cos response that is undesirable. we have the plant my (t) = u(t) + d(t). Suppose that the desired condition is simply y(t) = 0.1 Proportional Only A proportional controller alone invokes the control law u(t) = −kp y(t). m a marginally stable Let u(t) = −kp y(t) − kd y (t). In most applications.4. In the absence of d(t). and it follows that my (t) = −kp y(t) − kd y (t) + d(t). is kd sin ωd t .104 15 CONTROL FUNDAMENTALS disturbance d(t). easily found using the Laplace transform. the system would be overdamped and very slow to respond to any command. a small amount of overshoot is employed because the response time is shorter. 15. With an excessively large value for kd . If the linear position of the mass is y(t).4. The kd value for critical damping in this example is 2 mkp .
Now consider what happens if d(t) has a constant bias do : it balances exactly the proportional control part. If d(t) = do . This response is exponentially stable as desired. provided the roots are stable. tuning of a PID controller may proceed without any system model. a time derivative leads to my (t) + kp y (t) + ki y(t) + kd y (t) = 0. as desired. a negative kd could be used to cancel some of its eﬀect and speed up the system response. The control system has now created a thirdorder closedloop response. To achieve good rejection of do with a P D controller. 15. we would need to set kp very large.5 Heuristic Tuning For many practical systems. One useful approach is due to Ziegler and Nichols (1942).g.. eventually settling out at y(t = ∞) = do /kp . and has no poles at the origin: G(s) . which is unacceptable. which transforms the basic characteristics of a step response (e. so that y(t = ∞) = 0. However.4. very large values of kp will also drive the resonant frequency ωd up. lightlydamped complex poles. This is especially pertinent for plants which are openloop stable.15. 15.3 ProportionalIntegralDerivative t 0 Now let u(t) = −kp y(t) − ki y(τ )dτ − kd y (t): we have t my (t) = −kp y(t) − ki 0 y(τ )dτ − kd y (t) + d(t). Note that if the mass had a very large amount of natural damping.5 Heuristic Tuning 105 2 where ωd = 4mkp − kd /2m. and can be safely tested with varying controllers. The idea is to approximate the response curve by a ﬁrstorder lag (gain k and time constant τ ) and a pure delay T : ke−T s (172) τs + 1 The following rules apply only if the plant contains no dominating. the input is 1(t)) into a reasonable PID design.
the PDcontroller has transfer function C(s) = kp + kd s. and so has some physical meaning. its input is the error signal E(s) = −y(s).6.2τ /T 15 CONTROL FUNDAMENTALS ki = 0. force u(s). causing some error in the evaluation of e(s). The ﬁrst is a process disturbance. Another external input is the reference command or setpoint. the disturbance has the same units as the plant output. it is additive with the control action. For example. which relates the input.60τ /T 2 kd = 0. or at the output. In the second case.27τ /T 2 ki = 0. and so on. The feedback loop in block diagram form is shown below. in trying to force e(s) to zero. used to create a more general error signal e(s) = r(s) − y(s).9τ /T kp = 1. position y(s). Sensors with very poor . In turn. 15. In the former case. the plant used above (a simple mass) has transfer function P (s) = 1/ms2 . will necessarily make y(s) approximate r(s).106 P PI PID kp = 1.1 Block Diagrams of Systems Fundamental Feedback Loop The topology of a feedback system can be represented graphically by considering each dynamical system element to reside within a box. which usually corrupts the feedback signal y(s). Note that the feedback loop.60τ Note that if no pure time delay exists (T = 0). The ﬁnal input is sensor noise. and its output is force u(s). which can be taken to act at the input of the physical plant.2 Block Diagrams: General Case The simple feedback system above is augmented in practice by three external inputs.0τ /T kp = 0. into the output. having an input line and an output line.6 15.6. this recipe suggests the proportional gain can become arbitrarily high! Any characteristic other than a true ﬁrstorder lag would therefore be expected to cause a measurable delay. C(s) u(s) P(s) y(s) 15.
6 Block Diagrams of Systems 107 noise properties can ruin the performance of a control system. Other systems we encounter again later are the (forward) loop transfer function P C. du r + e C u + + P + + + n dy + y 15.3 Primary Transfer Functions Some algebra shows that e 1 = =S r 1 + PC y PC = =T r 1 + PC u C = = U.15. We would generally like S to be small at low frequencies. so that the tracking error there is small.6. r 1 + CP e/r = S relates the reference input and noise to the error. the loop transfer function broken between C and P : CP . Note that S + T = 1. y/r = T is called the complementary sensitivity function. no matter how perfectly understood are the other components. and is known as the sensitivity function. and e du y du u du e dy −P 1 + PC P = 1 + PC −CP = 1 + CP −1 = = −S 1 + PC = . implying that these two functions must always trade oﬀ. they cannot both be small or large at the same time.
then the three functions S. If the disturbance is taken at the plant output.108 y dy u dy e n y n u n 15 CONTROL FUNDAMENTALS 1 1 + PC −C 1 + CP −1 1 + PC −P C 1 + PC −C 1 + CP = = = = = =S = −U = −S = −T = −U. . T . and U (control action) completely describe the system. This will in fact be the procedure when we address loopshaping.
Namely. consider the system ˙ x = Ax. 16. This modal decomposition of A leads to a very useful statespace representation. Throughout.1 MODAL ANALYSIS Introduction The evolution of states in a linear system occurs through independent modes.1 Matrix Exponential Deﬁnition In the instance of an unforced response to initial conditions. a transformation of state variables can be made. x(t = 0) = χ. and matrix A: x(t) = eAt χ. 16. x = V z. giving a decaying exponential if a < 0.2. since A = V ΛV −1 . The same notation holds for the case of a vector x. which can be driven by external inputs.2 Modal Canonical Form Introductory material on the eigenvalue problem and modal decomposition can be found in the MATH FACTS section. In the scalar case.109 16 16.2.2 16. where (At)2 eAt = I + At + + ··· 2! eAt is usually called the matrix exponential. we use the statespace description of a system with D = 0: ˙ x = Ax + Bu y = Cx. leading to . This section provides the basis for modal analysis of systems. and observed through plant output. the response is x(t) = χeat .
the projection of the initial conditions onto the i’th mode. that is. T The product wi χ is a scalar. then the product is zero and the i’th mode does not respond. For stability of the system. The modal form is numerically robust for computations. λi .2.110 16 MODAL ANALYSIS ˙ z = Λz + V −1 Bu y = CV z. . since the states are simply the modal amplitudes. In terms of the response to an initial condition χ. we have n x(t) = i=1 T eλi t vi (wi χ). but may be coupled through the input (V −1 B) and output (CV ) mappings. (173) This is called the modal canonical form. we ﬁnd that (At)2 + ··· 2! (Λt)2 I + Λt + + ··· WT 2! eAt = I + At + = V n = V eΛt W T = i=1 T eλi t vi wi . must have negative real parts. Employing the above form for A. they are in fact the poles of the equivalent transfer function description.3 Modal Decomposition of Response Now we are ready to look at the matrix exponential eAt in terms of its constituent modes. If χ is perpendicular to wi . These states are uncoupled in Λ. the eigenvalues of A. 16. Otherwise. The projection of the i’th mode onto the states x is through the right eigenvector vi . the i’th mode does participate in the response.
then the system is controllable. · · · . such that B = [b1 . AB. T The product wi bk . If a given mode i has wi bk = 0 for all input channels k.. taking into account the initial condition for the derivative. Taking the Laplace transform of the system. Thus (sI − A)−1 can be recognized as the Laplace transform of the matrix exponential eAt . · · · .3 Forced Response and Controllability 111 16. a scalar. i. Then some rearrangement will give n m T (wi bk ) k=1 t 0 x(t) = i=1 vi eλi (t−τ ) uk (τ )dτ. then the mode is uncontrollable. b2 . the second term then has the form of a convolution of the matrix exponential and the net input Bu: t x(t) = 0 n eA(t−τ ) Bu(τ )dτ t T eλi (t−τ ) vi wi Bu(τ )dτ. In the time domain. where m is the number of input channels.16.e.3 Forced Response and Controllability Now consider the system with an external input u: ˙ x = Ax + Bu. In normal applications. If Mc has rank n. x(t = 0) = χ. (174) This matrix has size n × (nm). An−1 B]. represents the projection of the k’th control channel onto the i’th mode. all modes are controllable. . = i=1 0 Suppose now that there are m inputs. we have sx(s) − χ = Ax(s) + Bu(s) −→ x(s) = (sI − A)−1 χ + (sI − A)−1 Bu(s). controllability for the entire system is checked using the following test: Construct the socalled controllability matrix: Mc = [B. bm ]. We say that the i’th mode is controllable from the T k’th input if the product is nonzero.
and the known control u? The response due to the external input is easy to compute deterministically. · · · . We say that a system is observable if every mode can be seen in at least one output channel. The i’th mode is observable in the k’th output if the product ckT vi = 0. The observation is y = Cx (r channels of output). (175) This matrix has size n × (rn).4 Plant Output and Observability We now turn to a related question: can the complete state vector of the system be observed given only the output measurements y. and writing c1T C = · . crT the k’th channel of the output is n yk (t) = i=1 T (ckT vi )eλi t (wi χ). AT C T . . (AT )n−1 C T . Consider the part due to initial conditions χ. The usual test for system observability requires computation of the observability matrix: Mo = C T . through the convolution integral. the system is observable if Mo has rank n. We found above n x(t) = i=1 T eλi t vi wi χ.112 16 MODAL ANALYSIS 16.
is 1 + P (s)C(s) = 0. absolute values of transfer functions are replaced with the maximum singular values of transfer matrices. The Nyquist criterion allows us to assess the stability properties of a system based on P (s)C(s) only. equivalent to P (s)C(s) + P (s)C(s) = 0. whose roots are the poles of the closedloop system. respectively. to meet performance and robustness speciﬁcations.113 17 17. or LQG/LTR. 17. we shape the openloop transfer function (or matrix) P (s)C(s). Most of the analysis here is given for singleinput. This method for design involves plotting the complex loci of P (s)C(s) for the range ω = [−∞. r(s) 1 + P (s)C(s) where P (s) represents the plant transfer function. ∞]. and C(s) the compensator. The sensitivity function is written as S(s) = e(s) 1 = . Once this is done. to be presented in the next lectures. as given previously. and in this sense the design approach is quite diﬀerent from the rootlocus method (see Ogata). then the compensator must be computed.2 Roots of Stability – Nyquist Criterion We consider the SISO feedback system with reference trajectory r(s) and plant output y(s). The loopshaping approach is inherently twofold. and the nominal plant P (s). The closedloop characteristic equation. First. . The tracking error signal is deﬁned as e(s) = r(s) − y(s). where the underline and overline denote the denominator and numerator. from from knowing the nominal product P (s)C(s). Design based on singular values is the idea of L2 control. thus forming the negative feedback loop. There is no explicit calculation of the closedloop poles. In particular.1 CONTROL SYSTEMS – LOOPSHAPING Introduction This section formalizes the notion of loopshaping for linear control system design. singleoutput systems. but the link to multivariable control is not too diﬃcult.
The heart of the Nyquist analysis is the mapping theorem.e. the equivalent of lowpass ﬁltering the error signal. Conversely.1 17 CONTROL SYSTEMS – LOOPSHAPING Mapping Theorem We impose a reasonable assumption from the outset: The number of poles in P (s)C(s) exceeds the number of zeros. 17. or P (s)C(s). It is a reasonable constraint because otherwise the loop transfer function could pass signals with inﬁnitely high frequency. This leads to a slightly shorter form of the above relation: P = CCW. Namely.. and then proceeding to the right at an arbitrarily large radius. back to the negative imaginary axis. Since the zeros of F (s) are in fact the poles of the closedloop transfer function. clockwise(CW) paths in the splane.114 17.2 Nyquist Criterion The Nyquist criterion now follows from one translation.. stability requires that the number of unstable poles in F (s) is equal to the number of CCW encirclements of the origin. Since CW and CCW encirclements of the origin may cancel. S(s). this constraint can be easily met by adding a highfrequency rolloﬀ to the compensator. i. as s sweeps around the entire righthalf splane. e. every pole of F (s) enclosed in the splane generates exactly one CCW encirclement of the origin in the F (s)plane. In the case of a PID controller (two zeros) and a secondorder zeroless plant. The trick now is to make the trajectory in the splane enclose all unstable poles.2. stability requires that there are no zeros of F (s) in the righthalf splane. (176) In words. which answers the following question: How do paths in the splane map into paths in the F plane? We limit ourselves to closed.g. moving up the imaginary axis. and the remarkable result of the mapping theorem is Every zero of F (s) enclosed in the splane generates exactly one CW encirclement of the origin in the F (s)plane.2. the relation is often written Z − P = CW . instead of those of F (s): P = CCW ←→ closedloop stability (177) . Then the stability criterion can be cast in terms of the unstable poles of P (s)C(s). encirclements of the origin by F (s) are equivalent to encirclements of the point (−1 + 0j) by F (s) − 1. the path encloses the entire righthalf plane. Let F (s) = 1 + P (s)C(s).
kg ≤ 0. then the loci of P (s)C(s) must not encircle the critical point at all. . each having a very clear graphical representation. • The requirement that the number of poles in P (s)C(s) exceeds the number of zeros means that at high frequencies. There are several details to keep in mind when making Nyquist plots: • If neither the plant nor the controller have unstable modes. the deﬁnition of kg is diﬀerent for stable and unstable P (s)C(s). For a stable plant. P (s)C(s) always decays such that the loci go to the origin.3 Robustness on the Nyquist Plot The question of robustness in the presence of modelling errors is central to control system design.2 Roots of Stability – Nyquist Criterion 115 This is in fact the complete Nyquist criterion for stability. the loci of P (s)C(s) occur as complex conjugates – the plot is symmetric about the real axis. 17. • When the magnitude P (s)C(s) = 1. kg ≥ 2 and γ ≥ 30◦ .2. the procedure of looking at individual Nyquist plots for each element of a transfer matrix is unreliable and outdated. As the ﬁgure shows.5 and γ ≥ 30◦ . There are two natural measures of robustness for the Nyquist plot. we should count the encirclements for the function [det(I + P (s)C(s)) − 1] instead of P (s)C(s). its angle should not be −180◦ . These notions lead to deﬁnition of the gain margin kg and phase margin γ for a design.17. for an unstable plant. how close the loci come to it can be expressed in terms of magnitude and angle. the magnitude P (s)C(s) should not be near one.e. It is a necessary and suﬃcient condition that the number of unstable poles in the loop transfer function P (s)C(s) must be matched by an equal number of CCW encirclements of the critical point (−1 + 0j). the negative imaginary axis). The loci need to stay away from the critical point. Rules of thumb are as follows. • When the angle of P (s)C(s) is −180◦ . • For the multivariable (MIMO) case. Referring to the multivariable deﬁnition of S(s). The use of gain and phase margin in design is similar to the SISO case.. • Because the path taken in the splane includes negative frequencies (i.
(178) where W1 (s) is a stable weighting function of frequency. W1 (s) should be large in the same range. if onepercent tracking is to be maintained for all frequencies below ω = λ. ∀ω < λ. can be cast in terms of the Nyquist plot. at the lower frequencies. The disk is to be quite large. we know that S(s) should be small at low frequencies. then S(s) < 0. Since the sensitivity function maps reference input r(s) to tracking error e(s). the DC gain and slow.01. using the nominal plant model.3 Design for Nominal Performance Performance requirements of a feedback controller. and this latter condition can be interpreted as: The loci of P (s)C(s) must stay outside the disk of radius W1 (s).4 Design for Robustness It is a ubiquitous observation that models of plants degrade with increasing frequency. To force S(s) to be small at low ω. which is to be centered on the critical point (−1 + 0j). but higherfrequency components in the response may . 17. lightlydamped modes or zeros are easy to observe. For example. For example. possibly inﬁnitely large. This can be formalized by writing W1 (s)S(s) < 1. The requirement W1 (s)S(s) < 1 is equivalent to W1 (s) < 1 + P (s)C(s). We restrict the discussion to the scalar case in the following sections.116 1/kg Im(s) 17 CONTROL SYSTEMS – LOOPSHAPING P(s)C(s) Im(s) Re(s) Re(s) γ γ 1/kg P(s)C(s) Stable P(s)C(s) Unstable P(s)C(s) 17.
it will suﬃce to let P ˜ Example: Let P = k/(s − 1). which will not vary with frequency in this case. and P (s) is the actual. ˜ In the scalar case. for the two extreme values of k. One of the most useful descriptions of model uncertainty is the multiplicative uncertainty: ˜ P (s) = (1 + ∆(s)W2 (s))P (s). P (s) represents the nominal plant model used in the design of the control ˜ loop. The path of the perturbed plant could be anywhere on a disk of radius W2 (s)P (s)C(s). ∆(s)W2 (s)P (s). For constructing the Nyquist plot. The eﬀects of modeling uncertainty can be considered to enter the nominal feedback system as a disturbance at the plant output. dy . The last inequality is thus a condition for robust stability in the presence of multiplicative uncertainty parametrized with W2 . (180) where T is the complementary sensitivity function. Higherfrequency behavior may have more nonlinear properties as well. W2 (s) should be growing with increasing frequency. (179) Here. This can be written as 1 + P C > W2 P C ←→ W2 P C 1 > ←→ 1 + P C 1 > W2 T . ∆W2 . The weighting function W2 (s) scales ∆(s) with frequency. with the smallest possible value of W2 . W2 (s) should not grow any faster than necessary. We need to create a nominal model P = k0 /(s − 1). where ∆(s) is an unknown but stable function of frequency for which ∆(s) ≤ 1. where k is in the range 2–5. However. perturbed plant. The perturbation is of the multiplicative type. and W2 = 3/7. since the uncertainty grows. since it will turn out to be at the cost of nominal performance.17. we observe that ˜ P (s)C(s) = (1 + ∆(s)W2 (s))P (s)C(s).4 Design for Robustness 117 be hard to capture or even excite repeatably. . the weight can be estimated as follows: since P /P − 1 = ˜ /P − 1 < W2 . The robustness condition is that this disk should not intersect the critical point. yielding k0 = 7/2. Two equations can be written using the above estimate. centered on the nominal loci P (s)C(s).
We must therefore have P C < 1/W2 . n Example: Suppose H(s) = ω0 /sn .e. i. since P C is large. Thus. as follows: 1.6 Implications of Bode’s Integral The loop transfer function P C cannot roll oﬀ too rapidly in the crossover region. This leads directly to the performance condition P C > W1  in this range. since P C is small. Then we have just angle(H) = − n π ∞ −∞ ln(coth(ν/2))dν = − nπ . so that d(lnH)/dν = −n. should not intersect the disk of radius W2 P C. dν (182) where ν = ln(ω/ω0 ). This is met if W1 S + W2 T  < 1. It is not hard to see how the integral controls the angle: the function ln(coth(ν/2)) is nonzero only near ν = 0. It follows that H = ω0 /ω n . To see this requires a short foray into Bode’s integral. ω0 is a ﬁxed constant. At high frequency. with radius W1 . it is a simple function with n poles at n the origin. centered on the nominal locus P C. W1 S W1 /P C.5 Robust Performance The condition for good performance with plant uncertainty is a combination of the above two conditions. the angle is large. which in turn degrades the phase margin. 2 (183) . Graphically. the disk at the critical point. 17. implying that the angle depends only on the local slope d(lnH)/dν.118 17 CONTROL SYSTEMS – LOOPSHAPING 17. (181) The robust performance requirement is related to the magnitude P C at different frequencies. For a transfer function H(s). if the slope is large. At low frequency. the crucial relation is angle(H(jω0 )) = 1 π ∞ −∞ d (lnH(jω)) · ln(coth(ν/2))dν. for robustness.. The integral is hence taken over the log of a frequency normalized with ω0 . and no zeros. 2. and lnH = −nln(ω/ω0 ). W2 T  W2 P C. The simple reason is that a steep slope induces a large phase loss.
• One caveat for the wellbehaved plant is that lightlydamped poles or zeros should not be cancelled verbatim by the compensator. and the phase margin can be easily lost if the slope of P C at crossover (where the magnitude is unity) is too steep. • Nonminimum phase or unstable behavior in the plant can usually be handled by performing the loopshaping for the closest stable model. the phase may vary signiﬁcantly with frequency. Each pole at the origin clearly induces 90◦ of phase loss. We will use concepts from optimal linear control for the MIMO case. because the closedloop response will be sensitive to any slight change in the resonant frequency. we have extensively described what the open loop transfer function P C should look like. • When the plant is stable and has stable zeros (minimumphase). L = P C. and then divide through by the plant. while zeros not at the origin at 90◦ of phase lead for frequencies above the zero. The usual procedure is to widen the notch or pole in the compensator. The Nyquist loci are clearly susceptible to these variations is phase. Each zero at the origin adds 90◦ phase lead. to meet robustness and performance speciﬁcations. (184) This extraordinarily simple step involves a plant inversion. In the general case. and then explicitly considering the eﬀects of adding the unstable parts. In the immediate neighborhood of these poles and zeros. the division can be made directly. In . it suﬃces to just pick C = L/P.17. We have said very little about how to get the compensator C. through a higher damping ratio. the critical component.7 The Recipe for Loopshaping 119 This integral is trivial to look up or compute. 17. equivalent to a single pole). and may be secondorder (−40dB/decade) if an adequate phase angle can be maintained near crossover. The overall idea is to ﬁrst shape L as a stable transfer function meeting the requirements of stability and robustness.7 The Recipe for Loopshaping In the above analysis. The slope can safely be ﬁrstorder (−20dB/decade. each pole not at the origin induces 90◦ of phase loss for frequencies above the pole. let the designed loop transfer function be renamed. but in the scalar case. For clarity.
Useful properties and relations for the singular value are found in the section MATH FACTS. which is analogous to an eigenvector. the converse is true: The feedback system must be faster than the corresponding frequency of the unstable mode. We list below some basic properties of the singular value decomposition. analysis. or modal. (185) The openloop transfer matrix L should be shaped accordingly. the ideas from scalar loopshaping can be adapted using the singular value. we ﬁnd that they impose an unavoidable frequency limit for the crossover. The condition for MIMO robust performance can be written in many ways. When a control system involves multiple inputs and outputs. including a direct extension of our scalar condition σ(W1 S) + σ(W2 T ) < 1. In general. . In the case of unstable poles. the troublesome zeros must be faster than the closedloop frequency response. we use the properties of optimal state estimation and control to perform the plant inversion for MIMO systems.120 17 CONTROL SYSTEMS – LOOPSHAPING the case of unstable zeros. In the following sections.
and must match the row dimension of K. The notion of optimality is closely tied to MIMO control system design. The linear quadratic regulator (LQR) is a wellknown design technique that provides practical feedback gains. controllers that are the best possible. i. The feedback gain is a matrix K. and that all of the n states x are ˙ available for the controller. and we ask what is the shortest path in miles.3 Dynamic Programming There are at least two conventional derivations for the LQR.1 LINEAR QUADRATIC REGULATOR Introduction The simple form of loopshaping in scalar systems does not extend directly to multivariable (MIMO) plants. 18. ˙ (186) xdesired represents the vector of desired states. Poleplacement is the process of placing the poles of (A−BK) in stable. optimal control solutions provide an automated design procedure – we have only to decide what ﬁgure of merit to use. 18. due to R.. In this sense. The system dynamics are then written as: x = (A − BK)x + Kxdesired .121 18 18.2 FullState Feedback For the derivation of the linear quadratic regulator. The closedloop system has exactly as many outputs as inputs: n. which are characterized by transfer matrices instead of transfer functions. Optimal controllers. and serves as the external input to the closedloop system.e. according to some ﬁgure of merit. we assume the plant to be written in statespace form x = Ax + Bu. If A and C represent Los Angeles and Boston. and the “Bmatrix” of the closedloop system is K. we present here one based on dynamic programming. The column dimension of B equals the number of channels available in u. Bellman. turn out to generate only stabilizing controllers for MIMO plants. implemented as u = −K(x − xdesired ). . The “Amatrix” of the closed loop system is (A − BK). The key observation is best given through a loose example: Suppose that we are driving from Point A to Point C. suitablydamped locations in the complex plane.
The point is a subtle one: the optimization problem from Las Vegas to Boston is easier than that from Los Angeles to Boston. and nine paths total from the Blevel to D.122 18 LINEAR QUADRATIC REGULATOR B1 A B2 B3 C1 C2 C3 D n=3 s=2 for example. for example. from each Bnode. beginning in Boston. this same arbitrary point X would be along the new optimal path as well. Consider now traveling from point A (Los Angeles) to Point D (Boston). C3 . A bruteforce approach sums up the total distance for all the possible paths. there are many paths to choose from! Assume that one way or another we have found the best path. If we were to now solve the optimization problem for getting from only Las Vegas to Boston.3 By way of notation. First. we say that the path from A to B1 is AB1 . and that a Point B lies along this path. This operation involves nine additions of two 3 Apologies to readers not familiar with American geography. The dynamic programming approach has two steps. B2 . We now add some structure to the above experiment. B3 . Let X be an arbitrary point east of Las Vegas. and the Cnodes to the terminal point D. Example: Nodal Travel. . Suppose there are only three places to cross the Rocky Mountains. equivalent to eighteen additions of two numbers. and three places to cross the Mississippi River. and the idea is to use this property backwards through time to evolve the optimal path. There are nine unique paths from A to D. C1 . pick the best path to D. and picks the shortest one. C2 . B1 . B2 Dopt . The comparison of numbers is relatively cheap. B3 Dopt . There are three possible paths from B1 to D. Store the best paths as B1 Dopt . Suppose that all of the paths (and distances) from A to the Bnodes are known. say Las Vegas. In terms of computations. we could summarize that this method requires nine additions of three numbers. as are those from the Bnodes to the Cnodes.
is J= 0 ∞ l(x(t). The gap widens vastly. The closedloop system thus is intended to reject disturbances and recover from initial conditions. (188) . s = 5. however. First we deﬁne an instantaneous penalty function l(x(t). that is. AB2 + B2 Dopt . The combined path with the shortest distance is the total solution.. Finally. which is to be greater than zero for all nonzero x and u. The cost associated with this penalty.4 Dynamic Programming and FullState Feedback We consider here the regulation problem. n river crossing points). u(t)) = min {l(x(t).18.. u(t)) is made by considering all the possible control inputs u in the time interval (t. 18. u(t)) = t ∞ l(x(τ ). Second. u(τ ))dτ. the return at time t is constructed from the return at t + δt.g. t + δt). u (189) The minimization of V (x(t). the optimal return is the cost of the optimal trajectory remaining after time t: V (x(t). There are several necessary deﬁnitions. this second step involves three sums of two numbers. As suggested by dynamic programming. this example gives only a mild advantage to the dynamic programming approach over brute force. Needless to say. of keeping xdesired = 0. if there are s layers of nodes (e. and . compute the distance for each of the possible paths from A to D.g.e. the integral over time of the instantaneous penalty.4 Dynamic Programming and FullState Feedback 123 numbers. and each has width n (e. u(t + δt))} . while the dynamic programming procedure involves only (n2 (s − 1) + n) additions. u(t))dt. brute force requires 6250 additions. We have directly from the dynamic programming principle V (x(t). and total optimization is done in twelve additions of two numbers. along an optimal trajectory. u(t)). as one increases the dimensions of the solution space.. In the case of n = 5. constrained to the optimal paths from the Bnodes onward: AB1 + B1 Dopt . or AB3 + B3 Dopt . the brute force approach will take (sns ) additions. dynamic programming needs only 105. but not necessarily follow ytrajectories. rivers or mountain ranges). In general. (187) i. u(t))δt + V (x(t + δt).
If V is smooth and has no explicit dependence on t. as written. u(t)) + (Ax(t) + Bu(t))δt. Substitution of this form into the above equation. u) + xT P (Ax + Bu) . This suggested form for the optimal return can be conﬁrmed after the optimal controller is derived. The matrices Q and R are to be set by the user. −→ (190) ∂x dt ∂V = V (x(t). u). 2 where P is a symmetric matrix. and positive deﬁnite. ∂x V (x(t + δt). 5 4 . and setting the derivative with respect to u to zero gives Positive deﬁniteness means that xT P x > 0 for all nonzero x.124 18 LINEAR QUADRATIC REGULATOR the diﬀerential component due to l(x. u(t + δt)) = V (x(t). t + δt) cannot aﬀect V (x(t). ∂x (191) We next make the assumption that V (x. then ∂V dx δt + h. u(t)) + Now control input u in the interval (t.o. so inserting the above and making a cancellation gives 0 = min l(x(t).t. u (192) (193) We ﬁnally specify the instantaneous penalty function. The LQR employs the special quadratic form 1 1 l(x. u) = xT Qx + uT Ru. and xT P x = 0 if x = 0. u) = xT P x. and represent the main “tuning knobs” for the LQR. u(t)). u(t)) + u ∂V (Ax(t) + Bu(t)) .45 It follows that ∂V = xT P −→ ∂x 0 = min l(x. 2 2 (194) where Q and R are both symmetric and positive deﬁnite. u) has the following form: 1 V (x.
18.5 Properties and Use of the LQR Static Gain. Hence. The MARE is easily solved by standard numerical tools in linear algebra. the order of the closedloop system is the same as that of the plant. whose solution P is needed to compute the optimal feedback gain K. 18. . This is in good agreement with the practical guidelines for control system design. (195) The gain matrix for the feedback control is thus K = R−1 B T P . The LQR achieves inﬁnite gain margin: kg = ∞. 2 2 All the matrices here are symmetric except for P A. 2 2 leading to the ﬁnal matrixonly result 0 = Q + P A + AT P − P BR−1 B T P. Robustness. we can make its eﬀect symmetric by letting 1 1 xT P Ax = xT P Ax + xT AT P x. implying that the loci of (P C) (scalar case) or (det(I + P C) − 1) (MIMO case) approach the origin along the imaginary axis. and eliminating u in favor of x. which is not a dynamical system. we have 1 1 0 = xT Qx − xT P BR−1 B T P + xT P Ax.5 Properties and Use of the LQR 125 0 = uT R + x T P B uT = −xT P BR−1 u = −R−1 B T P x. (196) This equation is the matrix algebraic Riccati equation (MARE). Inserting this solution back into equation 194. The LQR generates a static gain matrix K. since xT P Ax = xT AT P x. The LQR also guarantees phase margin γ ≥ 60 degrees.
The angular separation of n closedloop poles on the arc is constant. The remaining poles assume a Butterworth pattern. When R >> C T Q C. it is not the states x which are to be minimized. The optimal control must always stabilize the closedloop system. whose radius increases to inﬁnity as R becomes smaller and smaller. First. and equal to 180◦ /n. Behavior of ClosedLoop Poles: Expensive Control. Behavior of ClosedLoop Poles: Cheap Control. since y = Cx. and the auxiliary matrix Q weights the plant output. . This part is akin to the highgain limiting case of the root locus. When R << C T Q C. the cost function is dominated by the control eﬀort u. The Butterworth pattern refers to an arc in the stable lefthalf plane. The expensive control solution puts stable closedloop poles at the mirror images of the unstable plant poles. and there is no penalty for using large u. An angle 90◦ /n separates the most lightlydamped poles from the imaginary axis. so that the closedloop poles approach the openloop plant poles in a manner consistent with the scalar root locus. In the case of a completely stable plant. we set the weighting matrix Q = C T Q C. the gain will indeed go to zero. and at the mirror images of the unstable plant zeros. so there should be some account made for unstable plant poles. however. the cost function is dominated by the output errors y. In this case. as shown in the ﬁgure. poles are placed at stable plant zeros. There are two groups of closedloop poles. and so the controller minimizes the control action itself. In many cases.126 18 LINEAR QUADRATIC REGULATOR Output Variables. but the output variables y.
18.5 Properties and Use of the LQR 127 X φ /2 φ n=4 φ = 45 degrees X φ φ X φ /2 X .
the Kalman Filter designs an optimal H. Given statistical properties of real plant ˆ ˆ disturbances and sensor noise. (198) There are n states.2 Problem Statement We consider the statespace plant model given by: x = Ax + Bu + W1 ˙ y = Cx + W2 . This state estimation is the task of a modelbased estimator having the form: ˙ x = Aˆ + Bu + H(y − C x) ˆ x ˆ (197) The vector x represents the state estimate. H operates on the estimation error mapped to the plant output y. and may not allow reconstruction of all the states. and l outputs. B is n × m. In real applications. we derived the linear quadratic regulator as an optimal solution for the fullstate feedback control problem.1 KALMAN FILTER Introduction In the previous section. W1 . so that A has dimension n × n. and a correction term with the estimator gain matrix H. The inherent assumption was that each state was known perfectly. whose evolution is governed by ˆ the nominal A and B matrices of the plant.128 19 KALMAN FILTER 19 19. since C x = y . m inputs. the measurements are subject to disturbances. and C is l × n. y C Bu y + H + + + A x 1/s x 19. The plant subject to two random input signals.
Hence. The Kalman ﬁlter design provides H that minimizes the scalar cost function J = E(eT W e). There are two main steps for deriving the optimal gain H.3 Step 1: An Equation for Σ 129 and W2 . The estimation error may be deﬁned as e = x − x.˙ 19. is deﬁned as Σ = E(eeT ). (203) where W is an unspeciﬁed symmetric. An important assumption of the Kalman Filter is that W1 and W2 are each vectors of unbiased. (199) (200) (201) Here δ(t) represents the impulse (or delta) function. and V2 is an l × l diagonal matrix of intensities. W1 + HW2 is considered an external input. ˙ W2 denotes sensor noise. and that all the n + l channels are uncorrelated. positive deﬁnite weighting matrix. A related matrix. since it drives x directly. W1 represents disturbances to the plant. We begin with . (204) 19. (202) The eigenvalues of the matrix A − HC thus determine the stability properties of the estimation error dynamics. The second term above. if E(·) denotes the expected value. V1 is an n × n diagonal matrix of intensities. It can then be veriﬁed that ˆ e = [Ax + Bu + W1 ] − [Aˆ + Bu + H(y − C x)] ˙ x ˆ = (A − HC)e + (W1 − HW2 ). which corrupts the measurement y. the symmetric error covariance. independent white noise.3 ˙ Step 1: An Equation for Σ The evolution of Σ follows from some algebra and the convolution form of e(t). E(W1 (t1 )W1 (t2 )T ) = V1 δ(t1 − t2 ) E(W2 (t1 )W2 (t2 )T ) = V2 δ(t1 − t2 ) E(W1 (t)W2 (t)T ) = 0n×l .
130 19 KALMAN FILTER ˙ Σ = E(eeT + eeT ) ˙ ˙ (205) T T T T T T = E[(A − HC)ee + (W1 − HW2 )e + ee (A − C H ) + T T e(W1 − W2 H T )]. T T The ﬁnal expression for E(e(W1 − W2 H T )) is symmetric. It is independent of the initial condition e(0).e. leading to ˙ Σ = (A − HC)Σ + Σ(AT − C T H T ) + V1 + HV2 H T . using the property that e(t) = e(A−HC)t e(0) + 0 t e(A−HC)(t−τ ) (W1 (τ ) − HW2 (τ )) dτ. 2 2 To get from the ﬁrst righthand side to the second. and depends on the (as yet) unknown estimator gain matrix H. and the ﬁnal result follows from = 1 δ(t − τ )dτ = . The fact that W1 and HW2 are uncorrelated leads to the third line. we note that the initial T T condition e(0) is uncorrelated with W1 − W2 H T . . 2 0 i. and therefore appears in Equation 205 twice.. The last term above can be expanded. (206) t This equation governs propagation of the error covariance. We have T T T T E(e(W1 − W2 H T )) = e(A−HC)t E(e(0)(W1 − W2 H T )) + t 0 t e(A−HC)(t−τ ) E((W1 (τ ) − HW2 (τ )) T T (W1 (t) − W2 (t)H T )) dτ = 0 t e(A−HC)(t−τ ) E((W1 (τ ) − HW2 (τ )) T T (W1 (t) − W2 (t)H T )) dτ = 0 e(A−HC)(t−τ ) (V1 δ(t − τ ) + HV2 H T δ(t − τ )) dτ 1 1 V1 + HV2 H T . the written integral includes only half of the impulse.
Next.4 Step 2: H as a Function of Σ We now make the connection between Σ = E(eeT ) (a matrix) and J = E(eT W e) (a scalar). it follows that ∂J /∂H = 0. We now introduce an auxiliary cost function deﬁned as J = trace(ΣW + ΛF ). the constraint we invoke is the evolution of Σ.. J = J. we enforce Λ = ΛT . which give the derivative of a trace with respect to a constituent matrix: ∂ trace(−ΛHCΣ) = −ΛT ΣC T ∂H ∂ trace(−ΛΣC T H T ) = −ΛΣC T ∂H ∂ trace(ΛHV2 H T ) = ΛT HV2 + ΛHV2 . (208) (207) where F is an n × n matrix of zeros.4 Step 2: H as a Function of Σ 131 19. the correct choice of H achieves an extremal value. Lagrange multipliers provide an ingenious mechanism for drawing constraints into the optimization. and it can be veriﬁed that J = E(eT W e) = trace(ΣW ). (210) . Note that since F is zero.e. ∂H Proofs of the ﬁrst two are given at the end of this section. satisﬁed if H = ΣC T V2−1 . We need the following lemmas. so minimizing J solves the same problem. Then the condition ∂J /∂H = 0 leads to 0 = 2Λ(−ΣC T + HV2 ). The trace of a matrix is the sum of its diagonal elements. and the previous two lemmas.19. i. and Λ is an n × n matrix of unknown Lagrange multipliers. since the values are arbitrary. the last lemma uses the chain rule. Equation 206: ˙ J = trace ΣW + Λ(−Σ + AΣ − HCΣ + ΣAT − ΣC T H T + V1 + HV2 H T ) (209) If J is an optimal cost.
C. and inﬁnite gain margin. we obtain ˙ Σ = AΣ + ΣAT + V1 − ΣC T V2−1 CΣ. suggesting a duality with the LQR problem. R > 0 V1 ≥ 0. which minimizes the squarednorm of the estimation error. Inserting this back into Equation 206. Furthermore. V1 . 19. for all the channels together or independently. The evolution of Σ is always stable. and the same analysis and numerical tools can be applied to both methodologies. (211) Equations 210 and 211 represent the practical solution to the Kalman ﬁltering problem. V2 ]. This is dual to the stability guarantee of the LQR loop. (212) The Kalman Filter is guaranteed to create a stable nominal dynamics A−HC. which might as well be the identity.5 Properties of the Solution The solution involves a matrix Riccati equation. leading to a more common MARE form of Equation 211: 0 = AΣ + ΣAT + V1 − ΣC T V2−1 CΣ. when the plant is statecontrollable. The steadystate solution for Σ is valid for timeinvariant systems. . the KF loop achieves 60◦ phase margin. and depends only on the constant matrices [A. like the LQR. Notice also that the result is independent of the weighting matrix W .132 19 KALMAN FILTER Duality of Linear Quadratic Regulator and Kalman Filter Linear Quadratic Regulator Kalman Filter x = Ax + Bu ˙ x = Ax + Bu + W1 ˙ y = Cx + W2 ˙ u = −Kx x = Aˆ + Bu + H(y − C x) ˆ x ˆ ∞ T T T 2J = 0 (x Qx + u Ru)dt J = E(e W e) Q ≥ 0. V2 > 0 K = R−1 B T P H = ΣC T V2−1 P A + AT P + Q − P BR−1 B T P = 0 ΣAT + AΣ + V1 − ΣC T V2−1 CΣ = 0 Hence the estimator gain matrix H can be written as a function of Σ. This is in fact the case. like the LQR. as long as the plant is fully stateobservable.
• A small disturbance V1 . Closedloop poles go to the stable plant poles.19. small disturbance. dual to cheapcontrol problem. • A large uncertainty Σ creates large H. x • A large disturbance V1 . large disturbance. • V2 >> V1 : poor sensors. V2 is the intensity of the sensor noise. weighting the model dynamics Aˆ + Bu more. and small sensor noise V2 creates a large H. or the mirror image of unstable plant zeros. Recall that V1 is the intensity matrix of the plant disturbance. placing emphasis on the corrective action of the ﬁlter. H small. and the mirror images of the unstable plant poles. dual to expensivecontrol problem. H >> 1. Some closedloop poles go to the stable plant zeros.6 Combination of LQR and KF 133 The qualitative dependence of the estimator gain H = ΣC T V2−1 on the other parameters can be easily seen. 19. The limiting closedloop poles of the Kalman ﬁlter are similar. and Σ is the error covariance matrix. For the plant given as x = Ax + Bu + W1 ˙ y = Cx + W2 . we put the Kalman Filter and controller gain G together as follows: . so that the ﬁlter’s correction is dominant.6 Combination of LQR and KF An optimal output feedback controller is created through the use of a Kalman ﬁlter coupled with an LQR fullstate feedback gain. and large sensor noise V2 creates a small H. This combination is usually known as the Linear Quadratic Gaussian design. or LQG. and dual to those of the LQR: • V2 << V1 : good sensors. The remaining poles follow a Butterworth pattern whose radius increases with increasing V1 /V2 .
7 19. x no longer ˆ represents an estimated state. Separation Principle: The eigenvalues of the nominal closedloop system are made of up the eigenvalues of (A − HC) and the eigenvalues of (A − BK). 19. on loopshaping via loop transfer recovery. It can be hooked up to receive tracking error e(s) = r(s) − y(s) as an input instead. ˆ (213) (214) C(s) B y + H + + C φ = (sIA) 1 φ K u There are two central points to this construction: 1.134 19 KALMAN FILTER ˙ x = Aˆ + Bu + H(y − C x) ˆ x ˆ u = −K x. separately. We use the output error as a control input in the next section. In this case.7. . Output Tracking: This compensator is a standalone system that. tries to drive its input y to zero. but rather an estimated state tracking error. as written. See proof below. so that it is not limited to the regulation problem alone. 2.1 Proofs of the Intermediate Results Proof that E(eT W e) = trace(ΣW ) E(eT W e) = E n n i=1 n ei n j=1 Wij ej = i=1 j=1 Σij Wji .
where the second form is a sum over i of the ii’th elements.3 Proof that ∂ trace(−ΛΣC T H T ) ∂H n Aijo Bko i i=1 (BA)ko jo (BA)To ko −→ j (BA)T AT B T .7.7 Proofs of the Intermediate Results 135 the transpose of W being valid since it is symmetric.2 Proof that ∂ trace(−ΛHCΣ) ∂H trace(AHB) = trace n n n l Aij k=1 l j=1 = i=1 j=1 Aij k=1 Hjk Bki . Now consider the diagonal elements of the product ΣW : Σ11 W11 + Σ12 W21 + · · · · · · Σ21 W12 + Σ22 W22 + · · · · → ΣW = · · ··· n n trace(ΣW ) = i=1 j=1 Σij Wji . Now ∂ trace(AHB) = ∂Hjo ko = = ∂ trace(AHB) = ∂H = 19. 2 = −ΛT ΣC T Hjk Bkl .19. . the il th element 19. the il th element j=1 Aij Hlj = i=1 j=1 Aij Hij . = −ΛΣC T 2 trace(AH T ) = trace = trace n n n j=1 n T Aij Hjl .7.
we can write the above in ˆ another form: d dt x e A − BK BK 0 A − HC x e = . 2 ∂H 19. .7. However. the determinant of an upper triangular block matrix is the product of the determinants of each block on the diagonal: det(sI −A ) = det(sI −(A−BK))det(sI −(A−HC)).4 Proof of the Separation Principle Without external inputs W1 and W2 .136 19 KALMAN FILTER where the last form is a sum over the ii’th elements. Using the deﬁnition of estimation error e = x − x. then its eigenvalues are the roots of det(sI − A ) = 0. If A represents this compound Amatrix. and hence the separation of eigenvalues follows. It follows that ∂ trace(AH T ) = Aio jo −→ ∂Hio jo ∂ trace(AH T ) = A. the closedloop system evolves according to d dt x x ˆ A −BK HC A − BK − HC x x ˆ = .
The LQR tuning matrices Q and R would be picked heuristically to give a reasonable closedloop response. It cannot be directly mapped to the intuitive ideas of loopshaping and the Nyquist plot. then an LQG design approach. respectively. and further is amenable to output tracking. As noted previously.137 20 20. First. Supposing that we have an estimator gain H which creates an attractive loop function L(s). since it does not aﬀect the error dynamics of the ﬁlter. Even if one could steer clear of such pathological cases. If the sensor noise and disturbance properties of the plant are indeed wellknown. ˆ In short. There are two reasons to avoid this kind of direct LQG design procedure. speciﬁcally to perturbations at the output y .1 LOOP TRANSFER RECOVERY Introduction The Linear Quadratic Regulator(LQR) and Kalman Filter (KF) provide practical solutions to the fullstate feedback and state estimation problems. which are at the root of feedback control. where φ(s) = (sI − A)−1 . there do exist plants for which there is no robustness guarantee for an LQG compensator. This follows from the estimator evolution equation ˙ x = Aˆ + Bu + H(y − C x) ˆ x ˆ and the ﬁgure. that is. or Cφ(s)BC(s) ≈ Cφ(s)H. The procedure is termed Loop Transfer . the KF loop is an ideal candidate for a loopshaping design. combining the LQR and KF into an output feedback compensator. (215) It will turn out that the LQR can be set up so that the an LQGtype compensator achieves exactly this result. the KF loop has good robustness properties. we would like to ﬁnd the compensator C(s) that establishes P (s)C(s) ≈ Cφ(s)H. We now reconsider just the feedback loop of the Kalman ﬁlter. The KF has openloop transfer function L(s) = Cφ(s)H. Note that we have not included the factor Bu as part of the ﬁgure. although the LQR and KF each possess good robustness properties. may yield good results. however. a second problem is that this design technique has no clear equivalent in frequency space.
First. Now Q = C T C = C T W T W C = (W C)T W C. In this regard. the KF loop has sensitivity function S(s) = (I + Cφ(s)H)−1 and complementary sensitivity T (s) = (I + Cφ(s)H)−1 Cφ(s)H. ρ→0 where K is the LQR gain matrix. LTR is useful as a SISO control technique. so that the Kalman ﬁlter loop itself has good performance and robustness properties. for which W T W = I. and so in this limit . and has two main parts. First recall the gain and Riccati equations for the LQR: K = R−1 B T P 0 = Q + P A + AT P − P BR−1 B T P. The Riccati equation becomes 0 = ρ(W C)T W C + ρP A + ρAT P − P BB T P = 0. 20. and W is an orthonormal matrix. we will show (roughly) that √ lim ( ρK) = W C. In the limit as ρ → 0.138 y 20 LOOP TRANSFER RECOVERY φ (s) ^ x H C Recovery (LTR). we pick suitable parameters of the LQR design. where I is the identity matrix.2 A Special Property of the LQR Solution Letting Q = C T C and R = ρI. it must be the case that P → 0 also. but has a much larger role in multivariable control. Secondly. one carries out a KF design for H. so that the LQG compensator satisﬁes the approximation of Equation 215. The condition σ(W1 (s)S(s)) + σ(W2 (s)T (s)) < 1 is suﬃcient for robust performance with multiplicative plant uncertainty at the output.
given at the end of this section. however. • The design plant has no unstable zeros.. See Athans for more details and references on this topic.e. This matrix may be absorbed into W through a matrix inverse. we note that the above property is true only for LQR designs with minimumphase plants.. First. with the deﬁnitions φ(s) = (sI − A)−1 and X(s) = (φ−1 (s) + HC)−1 = (sI − A + HC)−1 . The result of the last line establishes that the plant must be square: the number of inputs (i. Finally. we develop C(s). The proof of the LTR result depends on some easy lemmas. with W an orthonormal matrix. rows of C).e. then the limiting LQG controller C(s) satisﬁes ρ→0 lim P (s)C(s) = Cφ(s)H. The LTR method is limited by two conditions: • The plant has an equal number of inputs and outputs. and so does not need to be written. rows of K) is equal to the number of outputs (i. .e. The LTR method can be in fact be applied in the presence of unstable plant zeros. 20. 2 Note that another orthonormal matrix W could be used in separating K T from K in the last line.20. i.3 The Loop Transfer Recovery Result 139 ρ(W C)T W C ≈ = = = WC ≈ P BB T P (B T P )T B T P (R−1 B T P )T RR(R−1 B T P ) ρ2 K T K −→ √ ρK. the recovered function will exhibit reasonable limitations inherent to unstable zeros.3 The Loop Transfer Recovery Result √ The theorem is stated as: If limρ→0 ( ρK) = W C (the above result). those with only stable zeros (Kwakernaak and Sivan). Instead.. but the recovery is not to the Kalman ﬁlter loop transfer function.
then use Lemma 3 → (I + KX(s)B)−1 KX(s)H √ √ √ ( ρI + ρKX(s)B)−1 ρKX(s)H. In the last expression. then use Lemma 3 → (I + Cφ(s)H)−1 Cφ(s). Next we invoke the result from the LQR design. Now we look at the product CX(s): C(SI − A + HC)−1 C(φ−1 (s) + HC)−1 .140 20 LOOP TRANSFER RECOVERY C(s) = = = = = = = K(sI − A + BK + HC)−1 H K(X −1 (s) + BK)−1 H. Finally it follows that limρ→0 P (s)C(s) = Cφ(s)H. both properties of orthonormal matrices. gives lim C(s) = ((I + Cφ(s)H)−1 Cφ(s)B)−1 (I + Cφ(s)H)−1 Cφ(s)H = (Cφ(s)B)−1 Cφ(s)H = P −1 (s)Cφ(s)H. reintroduced into the limiting compensator. then use Lemma 2 → K(X(s) − X(s)B(I + KX(s)B)−1 KX(s))H KX(s)H − KX(s)B(I + KX(s)B)−1 KX(s)H (I − KX(s)B(I + KX(s)B)−1 )KX(s)H. we used the assumption that W is square and invertible. ρ→0 CX(s) = = = = = This result. then use Lemma 2 → C(φ(s) − φ(s)H(I + Cφ(s)H)−1 Cφ(s)) (I − Cφ(s)H(I + Cφ(s)H)−1 )Cφ(s). to eliminate √ ρK: lim C(s) = (W CX(s)B)−1 W CX(s)H = (CX(s)B)−1 CX(s)H. as desired. with ρ → 0. ρ→0 .
• Design for crossover frequency.5 Three Lemmas Lemma 1: Matrix Inversion (A + BCD)−1 = A−1 − A−1 B(C −1 + DA−1 B)−1 DA−1 . for example. • Scale the plant outputs (and references). 20. in depth and pitch control of a large submarine. Proof: . the plant input channels must be augmented with the necessary additional poles (at the origin). Then. one meter of depth error cannot be compared directly with one radian of pitch error. Often. For example. • Integrators should be part of the KF loop transfer function. Quantitative uncertainty models are usually at the cost of a lengthy identiﬁcation eﬀort. the bandwidth of is a more intuitive design parameter than is. the integrators are moved from the plant over to the input side of the compensator. The bandwidth of the controller is roughly equal to the frequency at which the (recovered) loop transfer function crosses over 0dB. and the compensator C(s) is constructed.4 Usage of the Loop Transfer Recovery 141 20. so that one unit of error in one channel is as undesirable as one unit of error in another channel. The tracking errors will accrue as desired. The LQR design step is thus trivial. Since the Kalman ﬁlter loop has only as many poles as the plant. with Q = C T C and R = ρI. • Scale the plant inputs in the same way. the highfrequency multiplicative weighting W2 . One Newton of propeller thrust cannot be compared with one radian of rudder angle. if no steadystate error is to be allowed. by using the limiting cheapcontrol LQR design.20.4 Usage of the Loop Transfer Recovery The idea of LTR is to “recover” a Kalman ﬁlter loop transfer function L(s) = Cφ(s)H. once the KF design is completed. Some speciﬁc techniques are useful.
2 2 .142 20 LOOP TRANSFER RECOVERY (A + BCD)(A + BCD)−1 = I A(A + BCD)−1 = I − BCD(A + BCD)−1 (A + BCD)−1 = A−1 − A−1 BCD(A + BCD)−1 = A−1 − A−1 BCD(I + A−1 BCD)−1 A−1 = A−1 − A−1 B(D−1 C −1 + A−1 B)−1 A−1 = A−1 − A−1 B(C −1 + DA−1 B)−1 DA−1 . Lemma 2: Short Form of Lemma 1 (X −1 + BD)−1 = X − XB(I + DXB)−1 DX Proof: substitute A = X −1 and C = I into Lemma 1. Lemma 3 I − A(I + A)−1 = (I + A)−1 Proof: I − A(I + A)−1 = (I + A)(I + A)−1 − A(I + A)−1 = (I + A − A)(I + A)−1 = (I + A)−1 .
The time constant τ is equal to the time required for y(t) to reach the value k(1 − e−1 ) = 0. and that the methods are only a small subset of what is available. Give estimates of the two plant transfer . consider the plant transfer function P (s) = k/(τ s + 1).1 SYSTEM IDENTIFICATION Introduction Modelbased controller design techniques. timedomain simulation. The basic idea is to express what is observed as a time signal whose Laplacedomain equivalent can be recognized. The process of generating workable models from observed data is the goal of system identiﬁcation. Example: Two raw step responses in heading were recorded for two different vessels. Good controllers usually have some reasonable robustness guarantees.2 Visual Output from a Simple Input For loworder plants which can tolerate impulse or step input. however. 21. We discuss in this section four fundamental but useful techniques for approximate system identiﬁcation of singleinput. The ﬁrst vessel was strongly unstable and had to be powered down abruptly after one second. If diﬀerent inputs give diﬀerent linear model coeﬃcients. such as the LQR and LTR. require a plant model. the smoothed data can still be ﬁtted to the curve y(t) = 0. In any event. The gain k is evident as the maximum value taken by the measured output. Systems with order three or higher will usually be more diﬃcult to assess with this visual technique. The user then has the choice of ignoring the nonlinearity. simulations with the nonlinear plant should always be performed to assess the robustness of the control strategy. then it is likely that nonlinear terms are playing a role. especially with the help of step functions parametrized on damping ratio ζ. Similar estimates can be made for secondorder systems.58(et − 1). It should be noted that the area of system identiﬁcation is a very rich one. a ﬁrstorder lag with gain k. or developing a controller which takes speciﬁc account of the it.143 21 21.632k. We have y(s) = k/s(τ s + 1) for the step input u(s) = 1/s. which motivates identiﬁcation with simple methods. and time constant τ . Except for the last approach. a great amount can be learned through step and impulse responses. singleoutput plants. for example if the operating point is controlled very closely. As an example. the methods are limited to linear models. and therefore y(t) = k(1 − e−t/τ ).
3. s(s − 1) y(s) = ˆ ˆ Since y(s) = P (s)u(s).144 1.2 0 1 −0.5.4 0.5 0. First. The undamped natural frequency ωn is related to ωd as follows: . undamped frequency ωn . so that the damping ratio is simply ζ = δ/2π 0.1) 1.58/(s − 1). and ζ are to be determined. and the second overshoot (same side) is about 0. The time trace of the ﬁrst system looks like the Laplace transform pair: 1 1 (e−at − e−bt ) ↔ y(s) = b−a (s + a)(s + b) y(t) = for the values b = 0 and a = −1.3/0. Next.10.5 0 0.7s.1.8 0.175.7 3. with respect to the steady value. the damped natural period is about 1.8 1 0 1 2 3 seconds 4 5 functions P (s).5 21 SYSTEM IDENTIFICATION 1 0.6 seconds 0.58 . = 2 2 u(s) s + 2ζωn s + ωn where gain k. where u(s) = 1/s. Finally. 0. the ﬁrst overshoot is about 0. assuming that the step input was of magnitude one.5. leading to the damped natural frequency ωd = 2π/1. so let k = 0.70rad/s.4 0 0. Thus for the experimental data.2 0. we note that the steady value of y(t) is about 0. The ratio is often written as the logarithmic decrement δ = ln(0. The second trace looks like a secondorder response of the form 2 y(s) kωn .6 0. we have P (s) = 0.
1 s2 21.g. Certain plants which cannot admit a step or impulsive input will tolerate a sinusoidal input. Two sources of error can corrupt the output signal y(s). and a gain.21. For each such test at a speciﬁc frequency. input signal u(s). + 1. water waves).28s + 14.3 Transfer Function Estimation – Sinusoidal Input The main idea of transfer function estimation is that P (s) can be estimated by simply dividing the measured output by the measured input. the estimated transfer function approach will succeed.e. we have ˆ P (s) = 7. Inserting these into the template above. in the frequency domain: y(s) ˆ P (s) = . A sensor which has noise in the frequency range of the system dynamics is problematic for obvious reasons. however: real disturbances and sensor noise. The data are plotted. u(s) (216) In applications. In some cases. which in many devices is negligible for low frequencies. periodic.. were implemented on a vessel operating near its cruising speed. while the next section generalizes to broadband input. disturbances may be bandlimited (e. compute the magnitude and phase relating the input to the output. and if these occur in a frequency range far away from the dominant dynamics. Bode plots are ﬁgures of transfer function magnitude and phase as functions of frequency. zeros. the input signal u(s) is known quite accurately because it is generated by a computer. along with the magnitude and phase of the transfer function . i. Conduct as many ˆ tests as are necessary to build a Bode plot of the plant estimate P (s).75rad/s. Example: Sinusoidal rudder angles trajectories. or used directly in a loopshaping or Nyquist plot approach. The idea then is to drive the plant with a narrowband. This section discusses the use of periodic inputs to create a Bode plot. which can be either parameterized in terms of poles. of 10deg amplitude (u(t) = 10deg). A similar argument holds for sensor noise.0 ..3 Transfer Function Estimation – Sinusoidal Input 145 ωn = √ ωd 1 − ζ2 3.
4 Transfer Function Estimation – Broadband Input Many times one needs to deal with experimental data that is broadband. either as the result of a closedloop run.21 which holds reasonably well at low frequencies. Note that the experimental magnitude and phase in this example deteriorate at the higher frequencies.093 . dB 10 −2 10 −1 10 rad/s 0 10 1 10 2 21. + 0. 0 −10 −20 −30 −40 0 angle(y(s)). but elements of spectral analysis are necessary.83s + 0. or of signiﬁcant disturbances.146 21 SYSTEM IDENTIFICATION ˆ P (s) = s2 0. . This is a property of almost all physical systems. deg −50 −100 −150 −200 −3 10 y(s)/u(s). In this case. frequencydomain analysis can still be used. The phase angle of y(s) is taken with respect to the input signal u(s). and an indicator that the plant model cannot be trusted above a certain frequency range.
• The Nyquist rate depends only on the sampling time step dt. We review several points for dealing with the DFT calculation: • The m’th DFT point is a summation of complex unitmagnitude vectors (e−j2πnm/N ) times the original data (x(n)). To improve the frequency resolution of the DFT. the ﬁrst frequency is 0. An additional multiplication of the DFT result by 2 will give peaks which are of about the right magnitude to be compared with the timedomain signal. a common approach is to zeropad the end of the real data.e.21.4. (217) for m = [1. called the Nyquist rate. What happens near the Nyquist rate depends on whether N is odd or even. N ].4 Transfer Function Estimation – Broadband Input 21. and so on. Dividing the DFT result by N returns the signal to its real magnitude. With this scaling. • The sampling theorem limits our usable frequency range to only onehalf of the sampling rate. It operates on a data vector x(n) of length N : N −1 X(m) = n=0 x(n)e−j2πnm/N . These correspond to the frequency range ωm = 2π m − 1 . the discrete Fourier transform (DFT) is a standard.1 Fourier Transform of Sampled Data 147 For the purpose of analyzing transfer functions. .. The points higher than π/dt correspond to negative frequencies. but the frequency vector accompanying the DFT can be made arbitrarily long by increasing the number of data points. optimized tool. X(3) = X ∗ (m − 1). dt N (218) i. and the last frequency is slightly less than the sampling rate. and a complexconjugacy holds: X(2) = X ∗ (m). the value X(1) is twice the true DC value. • The DFT generates a vector of N complex points as outputs.
The second rationale for windowing is based on the discrete nature of the transform. all of which are available through the Matlab function spectrum() for example. It is common to use segments which do not overlap. The DFT magnitude at a speciﬁc ω(m) is in fact an average of continuous frequencies from the neighborhood of the point. 21. but it is not a necessity. If N is prime. Because the DFT provides frequency data at only N/2 unique frequencies. if the window goes to a small value near its tails. • If possible. These can be reduced by taking separate DFT’s of subsections of data (perhaps zeropadded to maintain frequency resolution). no windowing at all is usually referred to as applying a rectangular window. window: 1 2πk 1 − cos 2 N w(k) = • The DFT of a given signal is subject to bias and variance. One of the simplest is the Hann. and the magnitude of the side lobes (tall side lobes bring in frequencies that are far away from ω(m)).2 Estimating the Transfer Function In continuous time. see below. This averaging idea is especially important in transfer function veriﬁcation. the estimated transfer function follows from .148 21 SYSTEM IDENTIFICATION • The data x(k) should be multiplied by a smoothing window w(k). the DFT should be run on a number of samples N which is a multiple of a large power of two. the DFT will take a long time to execute.4. will ensure a fair spectral analysis of a time series. There are many smoothing windows to choose from. First. Gibb’s eﬀect (the spectral signature of a discontinuity between x(1) and x(N )) can be minimized. The above steps. and then averaging them. when the plant input y(t) and output u(t) are transformed into frequency space. or cosine. for two reasons. Smoothing windows are generally chosen to achieve a tradeoﬀ of two conﬂicting properties relating to the average: the width of the primary lobe (wide primary lobes bring in frequencies that belong to the adjacent bins ω(m− 1) and ω(m + 1)). there is the possibility that a component in the real data lies in between one of these DFT frequencies.
compute two covariances and a crosscovariance: ∗ Γp (m) = Up (m)Up (m) uu Γp (m) = Yp (m)Yp∗ (m) yy ∗ Γp (m) = Yp (m)Up (m). we now want to include the same approach in our estimation of P (s). spectral analysis of a signal beneﬁts by using multiple segments and averaging. The procedure is: • For the p’th segment of data. compute the transfer function estimate ∗ ˆp (m) = Yp (m)Yp (m) .4 Transfer Function Estimation – Broadband Input 149 y(s) ˆ P (s) = . which assesses the quality of the ﬁnal ˆ estimate P (m): Γyu (m)Γ∗ (m) yu . Γyy (m)Γuu (m) Coh(m) = . u(s) (219) As noted above. yu • Construct average values of the transfer function estimate and the other quantities: ˆ P (m) Γuu (m) Γyy (m) Γyu (m) = = = = ˆ avgp (Pp (m)) avgp (Γp (m)) uu avgp (Γp (m)) yy avgp (Γp (m)) yu • Compute the coherence function.21. P ∗ Up (m)Up (m) • For the p’th segment of data.
The NelderMeade simplex method is easy to use. At the outset. the minimization must proceed iteratively. The method has its main strength in the fact that it applies to any model that can be simulated. The simulation operation can be written this way: y (t) = G(θ. both of which are reasonably assumed to be random processes. and we plan to look through the threedimensional parameter space θ = [m. Since the normed error is a complicated function of both u(t) and θ.150 21 SYSTEM IDENTIFICATION If the coherence is near zero. the three programs listed comprise a working Matlab set for identiﬁcation of a ﬁrstorder. u(t)). there is no clear relation between the input and the output. Some notes on use: • In this example. and the ˆ system identiﬁcation problem is to minimize ˆ(t) − yobs (t). running the simulation generates a new y . the coherence will deteriorate at high frequencies and also at any frequency where disturbances or noise occur. nonlinear system. b. k]. including those of high order and with signiﬁcant nonlinearities. With real data. we need to come up with some structure of the plant model. . Since the program simulate always uses the global variable theta as the parameters. the same simulation generates the “observed” data and the simulated response. then the segmental crosscovariances Γp (m) yu are sporadic and have cancelled out. we must be careful about setting theta in the calling programs. A fair guess for the actual dynamics has the form my + by + ky = u(t). say. where  ·  y here indicates the Euclidean norm. if the coherence is near one. For a given parameter vector θ. the method is computationally expensive and gives no guarantee of a useful solution. This result could be caused by either disturbances or sensor noise. and can be invoked with the Matlab function fmins. or even of convergence. Consider. Alternatively. 21. for example. and computing the norm gives a scalar measure ˆ of goodness. driven by the input force u(t). and uncoupled to the input signal.5 TimeDomain Simulation The timedomain simulation approach tweaks the parameters of a simulation so that its output matches the observed output. This can be based on physics in many cases. On the other hand. the case of a mass mounted on a spring and a dashpot. As an example. then the crosscovariances are in agreement and a real inputoutput relationship exists.
randn(2. and a better starting value can be given.g. plot(t_obs. but we are still at the mercy of the minimization algorithm. err = err + 1000*(1sign(mass)). . Starting from diﬀerent initial guesses for θ may help ﬁnd better results.. spline(t_raw. the data is interpolated to the same time scale as the observed data.’. u_obs) .’.1) . and a very complicated function. clf . y_obs = . t_obs. y_raw.5 TimeDomain Simulation 151 • After a simulation run is complete. % true parameter vector 0:dt:20*dt . % time step [1 2] . in order to compute the error. theta = theta_final . The error calculation.g. In many instances. e. y_sim = spline(t_raw. dt = . however.. t_obs. can be easily augmented by a term which penalizes invalid parameters. t_obs) . figure(1) . y_raw. hold off . %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% clear all . % observed output [theta_final] = fmins(’get_err’. disp(sprintf(’final theta(1): %g.21. [0 max(t_obs)]. y_raw] = ode45(’simulate’. % observed time vector ones(length(t_obs).3 theta = t_obs = u_obs = [t_raw. [0 max(t_obs)]. • The initial guess for theta is a random vector. 0) . % observed input y_raw] = ode45(’simulate’. 0) . clear global .1)) . disp(sprintf(’final theta(2): %g. theta_final(1))) . theta is roughly known. the Simplex method will take over from this point. y_obs. theta_final(2))) . y_sim. t_obs) . global u_obs t_obs y_obs dt theta . • The Simplex method may head into invalid parameter space. negative mass. [t_raw. e. • There is no guarantee that a global minimum will be found or even exists. however.
0) . ind = floor(t/dt) + 1 . y_raw.152 21 SYSTEM IDENTIFICATION %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% function [err] = get_err(theta_arg) . y_raw] = ode45(’simulate’.y) . global y_obs t_obs theta . theta = theta_arg . tau = theta(2) .’. % we have to choose which u_obs to use: % a zeroorder hold as implemented. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% function [ydot] = simulate(t. global u_obs dt theta . y_sim = spline(t_raw. err = norm(y_sim .y_obs) . ydot = ( k*u_obs(ind) . [t_raw. [0 max(t_obs)]. k = theta(1) . err)) . %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% . disp(sprintf(’error: %f. t_obs) .y^3 ) / tau .
both of which can provide highaccuracy absolute Cartesian navigation.1 Acoustic Navigation Consider a transponder A. z) and one equation: (xA − xB )2 + (yA − yB )2 + (zA − zB )2 = R2 . 22. located remotely from the vehicle. There is also a class of distributed sensor systems. notably compasses. these generally involve an array of communicating elements. the elapsed time T between a tranmission and consequent reception at A is T = Tt + 2R/cw .153 22 CARTESIAN NAVIGATION The bulk of our discussion on maneuvering and control has assumed that the necessary system states can be measured. which can transmit an acoustic signal. We present the fundamental concepts behind two methodologies in this second class: the global positioning system (GPS) and acoustic navigation. the job of B is just to transmit a signal whenever it receives one. rate gyros. paddle wheels. and also measure. with a (short) predictable response time Tt . . inertial navigation units. The range R follows by inversion: (T − Tt )c2 . then a given measurement of R places A on a sphere around B. Next. where cw is the speed of sound in water. The marine engineer is in fact faced with choices between many diﬀerent basic sensor packages. the time to receive a reply. about 1450m/s. and depth guages. Thus. These listed sensors are selfcontained and rely primarily on the physical properties of the natural environment. 2 R= Suppose that the location of B is known. y. with microsecond accuracy. place a responder B at a distance R away from A. It is a case of three unknowns (x. for example.
Very large systems utilize frequencies of 1015kHz (for a placement accuracy around 25m). then we have a functional set of measurements for acoustic navigation with just two responders. but still three unknowns. (Ultra) ShortBaseline The geometry of SBL or USBL puts very short baselines between the responders compared to the target distance. with a working range of 100500m to the target. For instance. then the intersection circle lies in a vertical plane.e. with baseline lengths on the order of 10100m. Better and better performance can be obtained by increasing the number of responders. and do not allow for abrupt crossovers. positional accuracy will be very poor since the two solution spheres are tangent. the ﬁxed net is often attached to a vessel or other structure. With three responders. Here there are three equations and three unknowns. if we know which side A is on. LongBaseline Longbaseline systems typically involve a larger responder net. Finally. There are many other pieces to the approach. the addition of a depth sensor to our suite would allow us to pin A’s location at one of two points on the circle. and may have tenkilometer baselines. and handling various geometric conﬁgurations that give rise to poor or degenerate solutions. In the latter . the intersection of a sphere (responder D) and a circle (responders B and C) is two points. and C lie in a nearly horizontal plane. but the nonlinearity of the equations leads to the nonuniqueness in the solution. A fourth responder or a depth transducer would be needed to completely constrain the solution. and the target ranges are similar to the baseline lengths. for example. There are two equations. B. Frequent sources of error in longbaseline systems are variations of cw . The above discussion is a minimum conceptual explanation of acoustic navigation.154 22 CARTESIAN NAVIGATION The introduction of a another responder C (typically listening and responding at a diﬀerent frequency than B) places A on the intersection of two spheres. and also in multipath. including an account of the variation of sound speed cw with water depth. obtaining the Cartesian locations of the responder network. and consequently of the baselines and spheres. i. named for the length of the baselines relative to the target (transponder A) range. The wavelength of a 100kHz signal is about 1cm and 5cm is a reasonable estimate of the accuracy for these systems. The baseline is the line connecting responders B and C. There are two common conﬁgurations used for acoustic navigation.. when A is near this baseline. When A. a circle. Typical frequencies in use are around 100kHz.
2 Global Positioning System (GPS) 155 B A3 A2 B A1 A* 2 A* 3 A C C Figure 6: Left: General conﬁguration of a longbaseline acoustic system. Sometimes. A∗ is an additional solution to the tworesponder 3 problem. these signals are typically eliminated by rejecting those receptions which are outside a very tight and slowly moving window on travel time T . in contrast. The accuracies achieved with GPS are therefore strongly dependent on the accuracy of timekeeping. Right: General conﬁguration of a shortbaseline system. A3 is wellposed. false signals can be caused by reﬂections oﬀ the seaﬂoor and the surface. with target transponder A and ﬁxed responders B and C. The position A1 is on the baseline and has poor accuracy in the direction normal to the baseline.22. but with one fundamental diﬀerence: only the oneway travel time from each satellite is measured at the target. It is similar in concept to the acoustic navigation systems described above.2 Global Positioning System (GPS) The GPS system is the most powerful system publicly available for absolute position reference above water. and the only receptions available are via multipath! 22. condition. . bottom topography can shield a direct acoustic path.
among other items. With reference to a rotating Earth. This single point is near the centroid of the three approximate points.000km. the range measurements from two sources locate the target on one of two points (the intersection of two circles). The target time base can be kept up to date by . that triangulation and navigation are performed at the target. the spatial frequency is doubled.3m range error. the speed of light c = 3 × 108 m/s implies that a 1ns timing error will cause a 0. Introduce a third satellite now so that the intersections of the circle pairs now occur at six points.156 22 CARTESIAN NAVIGATION The core of the system is a network of 24 satellites arranged in six planes (4 per plane) inclined at 55 degrees from the equatorial plane. and indicate the approximate true solution. and the six lines are spaced 60◦ apart in longitude. the time of its transmission. and the threedimensional location of the satellite (the ephemeris). In practical terms. ˜ The estimated range circles are too large. The accuracy of satellite position is better than 10m. geosynchronous orbit is much higher. using a combination of leastsquares analysis of past data (1 week). Three of these are very close.000km. from many satellites. The time base is continually monitored and updated from a ground station. It is from these two pieces of information. at around 40. so that the range estimate is ˜ r + r = c(t + t). each of the satellites carries four atomic clocks on board. good to 1ns per day. and the estimated location of the target too far from the baseline. The orbit lines over a nonrotating earth are thus nearsinusoids that reach lattitudes of 55◦ N and 55◦ S. the arguments for three dimensions are the same. triangulation of the type described for acoustic navigation is possible. at an altitude of about 20. For this reason. The ephemeris of each satellite is also monitored and updated from the ground station. so that the longitudinal distance between tracks is 30 degrees. Suppose that the target clock ˜ is too slow by an amount t. The trick is to ﬁnd the ˜ correction for t that puts the three close solutions onto a single point. For the ideal case that the time bases of the target and the satellites are exactly aligned. We now come to the last thorny issue: how to make target receivers that don’t require atomic clocks! The solution is very clever. and accurate to within 1ns. or 1800 nautical miles. Each satellite transmits a regular signal which contains. The altitude is chosen carefully to correspond with a 12hour orbit. In two dimensions. and a Kalman ﬁlter to predict the future ephemeris. and represents the best position solution. and illustrated for the case of twodimensions.
. and inaccuracies in the receiver and multipath (≈ 2m). ionosphere and troposphere electromagnetic variations (≈ 2. essentially reducing a threedimensional to a twodimensional problem. The idea here is that a stationary target can detect the eﬀects of S/A (since it is not moving) and then transmit the corrections over a small geographical area.5m). Major sources of error include: clock base and satellite navigation (≈ 2. The position speciﬁcation for GPS is 25m. GPS is also subject to selective availability or S/A.22. at the 95’th percentile. This is a remarkable feat. which are then applied to their own satellite navigation processing. an independent altitude measurement will enhance the accuracy of GPS.2 Global Positioning System (GPS) 157 performing this check for every set of signals. Hence. Finally.5m). Selective availability is controlled by the United States Department of Defense. the threedimensional timing problem is solved with four satellites. given that it is an absolute measure over the entire surface and atmosphere of our 6000kmradius Earth. by providing highaccuracy navigation in local areas. the addition of a slowvarying random component in the satellite ephemeris data. The advent of diﬀerential GPS solved most of the S/A concerns of American allies. as with the acoustic navigation systems described. Many current GPS receivers are capable of decoding these local corrections. and degrades the position speciﬁcation to 100m (typical). the twodimensional timing problem is solved with three satellites. Diﬀerential GPS typically provides 12m accuracy.
158 22 CARTESIAN NAVIGATION .
S. Jr. CA. Society of Naval Architects and Marine Engineers: Jersey City. Wiley: New York.H. Coordinate systems and rigidbody dynamics. Guidance and Control of Ocean Vehicles.C. propulsion. CA. J. FL.S. Dynamics of Mechanical and Electromechanical Systems. E. Euler angles and quaternions. April 1973. T. FL. Comprehensive modern treatment of marine dynamics and control. General reference. L. J. D.L. Marine Propellers and Propulsion. A method for estimating static aerodynamic characteristics for slender bodies of circular and noncircular cross section alone and with lifting surfaces at angles of attack from 0◦ to 90◦ . Includes a chapter on transverse and azimuthing thrusters. Hoerner and H.I. Elsevier: New York.F. Principles of Naval Architecture.F. NJ. S. Data on drag of bodies and wings.159 23 REFERENCES Short notes indicate how the reference is related to the text. 1997. Crandall. Blevins.V. 1975.V. Karnopp.V. 1986. Marine Hydrodynamics. Fossen.D. Carleton.. and lifting surfaces. E. PridmoreBrown. Hoerner Fluid Dynamics: Bakersﬁeld. Kurtz.F. Junkins and J. Krieger: Malabar. and D. 1994. Optimal Spacecraft Rotational Maneuvers. Newman. S. R. Hoerner Fluid Dynamics: Bakersﬁeld. Extensive added mass formulas. Extensive data on lift of bodies and wings. Fluid Dynamic Drag. Formulas for Natural Frequency and Mode Shape. 2nd Revision. Turner. 1984. 1985. Lewis. MIT Press: Cambridge.. The inﬂuence of thruster dynamics on underwater vehicle behavior and their . Krieger: Malabar. 1988.D. Classical treatment of stability. J. ed. Borst. Fluid Dynamic Lift. Borst. Hoerner and H. NASA Technical Note D7228.N.C. Jorgensen. 1975. ButterworthHeinemann: Boston. 1994.
25:146159. No. C. Applied Nonlinear Control. Whitcomb.J. IEEE J. 1990. 1986.A. D. P. S.J. Acceleration and steadystate propulsion dynamics of a gas turbine ship with controllablepitch propeller.M. and Systems. pp 340352. Doyle. Feedback Control Theory. D. Loopshaping concepts. Rock.C. Proc. 1994. General control reference. PrenticeHall: Englewood Cliﬀs. Tannenbaum. FlowInduced Vibrations. A.R. Francis. 1979. NJ. 10.160 23 REFERENCES incorporation into control system design. . 1970. Toward an improved understanding of thruster dynamics for underwater vehicles. Stateoftheart in thruster modeling. Luenberger. R. 1992. Accurate fourquadrant nonlinear dynamical model for marine thrusters: Theory and experimental validation. 94:283308. Philadelphia. R.J. Introduction to Dynamic Systems. Healey.R. K. Cody. Grosenbaugh.J. Sauders College e Publishing. Yoerger. and A. and J. Slotine. 1995. Ogata. Oceanic Engineering. SISO linear control. Slotine and W. B. IEEE Symposium on Autonomous Underwater Vehicle Technology. Modern Control Engineering. Oceanic Engineering. IEEE J. Van Nostrand Reinhold. Brown. Rubis. C. NJ.J. 1991.P. Cooke. J. 1972. Harper. S.E. W. New York. 1986. Macmillan: New York.G. Proc. SNAME Trans. Circuits. Rubis and T.M. 2000.. J. Blevins. SNAME Annual Meeting. MA. 1990. General reference on transforms. R. Control Engineering: A Modern Approach. Bachmayer.R. and M. Governing ship propulsion gas turbine engines. Cable strumming. B´langer.G.D. Wiley: New York. J.L. E. PrenticeHall: Englewood Cliﬀs. 2nd Edition. Li. D. Miles. 15:167178. Siebert. L. MIT Press: Cambridge. and J. Secondorder nonlinear thruster model. Firstorder nonlinear thruster model.A. Signals.
Martin. Kalman ﬁlter practical aspects. 1996. and M. Parkinson. 1995.L. Optimal Control. Kwakernaak and R.E. 1986. 2nd Edition. pp 13131324. Hearn and A.A. 1999. A.161 Dynamic programming and LQR derivation. and B. Proc. Athans.T. 1981. Vetterling. M. Function minimization. General reference on LTR. F. ed.L. 71:11771186. IEEE Trans. S. Automatic Control. J. IEEE.W.V. B. Gilbert. Lewis and V. Numerical Recipes in C. Wiley: New York. Halsted: New York. L. Doyle and G. 1986. Linear Optimal Control Systems. Spectral Analysis in Engineering. Wiley: New York.W. Athans. H. Teukolsky. Transfer functions from spectral analysis. Stein. 26:416. Proc. UK. A tutorial on the LQG/LTR method. 1983. No. Seminal paper on LTR. Cambridge University Press: Cambridge. Press. Syrmos. Multivariable control of a submersible using the LQG/LTR design methodology. More tutorial LTR. American Control Conference. R. Flannery. 1992. Vol. Proc. . Kalman ﬁlter derivation and properties. Sivan. NAVSTAR: Global positioning system – Ten years later. Applied Optimal Estimation. MIT Press: Cambridge.P. W. 1988. American Control Conference. 1972.C. G. Gelb. There is another such issue: IEEE Proc. Metcalfe. S. Multivariable feedback design: concepts for a classical/modern synthesis. 1. Valavani. 87.H. The whole issue is in fact dedicated to global positioning systems. pp 12891296.. W.
162 23 REFERENCES .
Estimate the net force that the radial bearings would have to support for a 1mm mass imbalance. 000rpm (revolutions per minute). Because of high rotational speed. For example. The gas generator turbine of a LM2500 marine engine rotates at p = 10. and that there is no signiﬁcant sideslip. Kinematics. say. with Ixy = Ixz = Iyz = xg = zg = 0. The following question pertains to the inertial forces experienced by this spinning rigid body. and has a mass of approximately 200kg. 1. it can be quite useful for preplanning trajectories. yg = 1mm with respect to the spin axis. which can be evaluated separately. 2. Hint: Break the trajectory up into three parts.163 24 PROBLEMS Many of the following problems use data presented by Fossen and in Principles of Naval Architecture. Although deadreckoning is not recommended as a method for longterm navigation. consider a surface vessel heading proﬁle as follows: t=0 0 < t <= 60s 60s < t < 240s 240s < t < 300s t = 300s φ π/4 φ(t) = 0 rad φ(t) = t/60 × π/4 rad φ(t) = π/4 rad φ(t) = (300 − t)/60 × π/4 rad φ(t) = 0 rad y φ x t = 0s t = 60s t = 240s t = 300s What is the total distance traveled in the Cartesian x direction during these ﬁve minutes? You may assume that the forward speed of the vessel is u = 1m/s. . the turbine is exceedingly sensitive to lateral balancing. Inertial dynamics.
˙ θ(0) = 1. using several examples. etc. Coeﬃcients and system modeling. Give physical interpretations for what would make such a device stable or unstable. The system equation can be written as two ﬁrstorder equations: d dt ˙ θ θ −B/A −C/A 1 0 ˙ θ θ = . response sensitivity to B and C. Nr . y O θ x Uο . Discuss. θ(0) = 0). ¨ ˙ (c) If we consider the diﬀerential equation A¨(t) + B y(t) + Cy(t) = 0. (b) Write Nθ . B. ±3}. using the MATLAB ODE solver ode45. If θ is the angle of the vane with respect to the wind direction. B. y ˙ the condition for stability is that A. Consider a weather vane in a wind of velocity Uo . For example. (d) Create a numerical model of this system. (a) Write the singledegree of freedom (N ) linearized equations of motion about the ﬁxed axis 0. Simulate the system response to nonzero initial conditions (e. Nθ˙ .. ±3.164 24 PROBLEMS 3.. and C must have the same sign. C} = {1. look at the range {A. which are related to the aerodynamic coefﬁcients. and Nθ in terms of Nv .g. Nr . Express this requirement in terms of the derivatives in the previous question.
(a) Set x(t) = est . s1 . ﬁnite negative? F O F O F O A u A u A u 5. x ˙ wherein all the coeﬃcients are positive and real (i. Consider the vibration of a massspringdashpot system: m¨ + bx + kx = 0. and ﬁnd two solutions for s in terms of m. complex s) in terms of the standard parameters . Give the linear hydrodynamic coeﬃcient at two diﬀerent operating conditions. using the quadratic formula.165 4. and x(0) = 0.e. Secondorder system. and k. b. small. Hydrodynamics. physical). origin O and A: is it zero. What do you think will happen when s1 = s2 ? (c) Write this x(t) for subcritical b (that is. ﬁnite positive.. The ﬁgure below shows some characteristic ﬂuid force curves versus a motion parameter. The solution pairs depend on whether b exceeds a critical value: what is the critical value and how do the solutions change? (b) From the initial condition x(0) = xo . and s2 . in terms of xo . ﬁnd the coeﬃ˙ cients for the general solution x(t) = c1 es1 t + c2 es2 .
i. Make a sketch of the timedomain result. for the same initial conditions. (a) Write the equation of motion for roll (K). Give your answer in terms of a real exponential multiplied by a single sine or cosine. i. x ˙ This equation deﬁnes a system. Submarine roll dynamics and control. What happens when the roots si are far apart vs.e. in both the subcritical and supercritical damping cases? Include sketches. with input u. This submarine has antirolling ﬁns to ensure stability. 6. both in frequency (Laplace) space... both s are real.8tm per degree of ﬁn rotation δ. (b) If the automatic control law δ = k1 ψ is used. write the transfer function for this system. what range of k1 ensures stability? . where ψ is the roll angle. Using the Laplace transform. at a forward speed of 5m/s. A submarine has weight 1200t (tons) and the center of gravity is 0. and in the timedomain. and make a sketch. (d) Sketch the response x(t) for supercritical b.5m above the center of buoyancy (What can you conclude?). The rolling motion can be assumed to be decoupled from the other motions. the impulse response. The control hydrodynamic derivative is Kδ = −2.e. when they are close together? (e) Consider now the case where an input acts to drive the mass from zero initial conditions: m¨ + bx + kx = u. Be sure to cover both the subcritical and supercritical damping cases.166 24 PROBLEMS ωn = k (undamped natural frequency) m b ζ = √ (damping ratio) 2 km ωd = ωn 1 − ζ 2 (damped natural frequency). (f) What is the step response of this system. and output x.
Steering. but assume it is nearly symmetric. Submarine stability via slender body theory. Indicate areas where the vessel turns against the rudder action. Consider a submarine moving forward at speed U . the submarine is neutrally buoyant and stable with the x and y axes horizontal. Does surge decouple from pitch and heave? Write out the coupled equations in matrix form. Draw curves of rudder angle δ vs. What does this tell us about the magnitude of the buoyant force and where it acts with respect to the center of weight? (a) Derive the equations of motion for the submarine moving at speed U. restoring pitch moment for small pitch θ. (f) Can the submarine be stable in pitch without feedback control? Can the submarine be stable in depth without feedback control? 9.. (e) Write out the complete linearized equations of motion.167 (c) If the speed suddenly drops to 2. (b) Derive the hydrostatic. Omit nonlinear and memory eﬀects.5m/s. where it rapidly tapers down to a point. and be sure to include the hydrostatic moment. (d) Expand the ﬂuid forces and moments in terms of the motions. how does the range of stabilizing k1 change? 7. If . (c) Linearize the inertial terms in the equations of motion. so that you can omit certain hydrodynamic terms. 8. (a) An underwater vehicle hull has the shape of a circular cylinder. except at the very ends. yaw rate r for a dynamically stable surface vessel. etc. Submarine pitch/heave dynamics. and restricted to small motions in the vertical (x−z) plane. • At rest. Explain your choices. and then for an unstable vessel. The submarine is not symmetric in the x − z plane due to the sail. Assume that: • The submarine is symmetric in the x − y plane and yG = 0.
and is this hull stable in pitch? Recall that the aerodynamic center is the equivalent location of the observed lift force that creates the observed moment. (b) It turns out that in order to match the lift and moment of the hull in experiments.0. as desired? 10. Using the lifting surface formulas. for U = 10m/s. (a) With l/d = 7. and speed is 2m/s. Slender body theory. using slenderbody theory. use slenderbody theory to express the following coeﬃcients: Zw . ˙ all referenced to the body midpoint.168 24 PROBLEMS the radius is 0. 7m 10m 15m 11. a ﬂat tail of radius 0.5m. length is 17m. Lifting surfaces. Determine the corrected value for Mw .4m has to be included in the slenderbody estimate. and compare again to the exact result. and in addition compute Zw . (b) Perform the slender body calculation also for a sphere. and Mw .6/rad is to be applied at the tail to make the device just stable in pitch. and compare it with the exact results from the table below (Blevins). Where is the aerodynamic center with respect to the center of the hull. (c) A ﬁn with predetermined lift slope of 3. approximate the crossbody added mass. Consider a long ellipsoidal body of length l and diameter d. (c) What is the added mass in the inline direction? . How big is the required ﬁn area that brings the net Mw to zero. Mq˙ . estimate the coefﬁcient Kv of a submarine sail as shown below.
355 1.908 1.0: 0.000 2b 2a 12.008 0. The submarine is a rotationally symmetric ellipsoid with L/D = 7.500 2.2: 3. Mw . for a small submarine.6: 0.075 0.148 0.210 5. L = 35m.2: 0.6: 0.1: α = 6.021 ∞: 0.894 7.0: 0.500 2.0: 0. The body .143 0.933 10.0: 0.1: α = 0.704 5.0: 0.0: 0.0: 0.036 10.000 Ellipsoid longitudinal added mass: 4αρπab2 /3 where a/b = 0. Submarine pitch stability with various methods.960 ∞: 1. using several diﬀerent methods. We will calculate the pitch derivative due to an angle of attack.0: 0.059 7.0: 0.169 Sphere added mass: 2ρπa3 /3 Ellipsoid crossbody added mass: 4αρπab2 /3 where a/b = 0.0: 0. We ﬁx the body origin to the midpoint as shown below.
20. and give a corrected slenderbody value for Mw of the body. DESTABILIZING . Force and moment referenced to body midpoint: Z = −0. 13. Cy = 1. (f) Use the Hoerner result and your result from wing theory to write a net Mw for the body plus ﬁns. (d) Apply Jorgensen’s approximate formulas to estimate Mw for the body. geometric aspect ratio of 3. (e) Use the experimental data from Hoerner (p. (a) Apply wing theory to characterize Mw due to the ﬁns alone.2: Symmetric body of revolution with L/D = 6. make linearizations where necessary.170 24 PROBLEMS is appended with two ﬁns as shown in the ﬁgure: each has an area of 1m2 .7.5ρU 2 D2 Cydb tan(w/u). (g) What is the location of the net aerodynamic center? If it has an unstable location forward of the midpoint. and has its longitudinal center of action 1m forward of the body stern. for the body alone? Hoerner p. Cm = 0. what increase in ﬁn area would bring it back to the midpoint? (h) What is the diameter of a ﬂat stern that will allow the slender body theory to give the same moment as the Hoerner (experimental) data.5ρU 2 D2 LCnb tan(w/u). (b) Apply slender body theory to estimate Mw for the body alone. 13. for the body alone? (i) What is the diameter of a ﬂat stern that allows the approximate formulas of Jorgensen to give the same moment as the Hoerner data.2) below to estimate Mw for the body.53. (c) Linearize the Munk moment. M = 0.
with mL >> M . . what is the maximum speed for the system? Assume the normal drag coeﬃcient of the cable is Cn = 1. the ﬁrst undamped natural frequency in heave. what is the dynamic tension amplitude at the top (s = L) of the cable? The formula for this dynamic tension (with mL >> M ) ˜ is T = EAP k sin(ks)/cos(kL). If this angle is considered equal to the critical angle φc .5second period. (b) If the vessel heaves with a P = 2m amplitude and a 4. Cable mechanics (hard!).2. You are asked to assess the operational envelope of a cable/vehicle system which has been installed on a vessel. The vehicle is streamlined. and a density of ρc = 5000kg/m3 . an eﬀective extensibility of E = 100 × 109 P a. (a) The towing angle at the vessel cannot exceed 25deg from vertical. and a density of 5000kg/m. (a) Calculate ω1 . with diameter 3cm. k = ω m/EA. does the cable unload at the surface. A towing cable is steeljacketed. and m is the cable mass per unit length. The cable is steeljacketed. with diameter 3cm. for the heaving conditions above? 14. and has negligible mass (M ). for reasons of deck and crane geometry. Cable dynamics. where A is the crosssectional area of the cable. The vehicle it tows is streamlined. If we are towing at low speed and L = 4000m depth. an eﬀective extensibility of 100 × 109 P a.171 Each fin: 1m 2 area AR = 3 1m L/D = 7 x y L = 35m 13. and has a mass (material plus added) of 100kg. (c) Taking into account the static tension induced by inwater weight.
If the fastest waves that the surface vessel responds to have period 4s. where ˜ c is a constant to be found. and wn is the inwater weight of the cable. express the natural frequency as a function of the vehicle depth. ﬁnd c in terms of Q and z(s = L).172 24 PROBLEMS (b) Considering the undamped. uncoupled axial dynamics. and sketch the curve. ˜ 2 ∂z ∂z z2 The derivatives can be transformed from s to z coordinates via the chain rule: ∂q ˜ ∂ q ∂z ˜ = · ds ∂z ∂s ∂2q ˜ ∂ 2 q ∂z ˜ = 2 ds ∂z 2 ∂s 2 + ∂ q ∂ 2 z ∂z ˜ · · . =T ∂t2 ∂s ∂s (m + ma ) where ma is the added mass of the cable per unit length. ∂z ∂z∂s ∂s (d) The equation above has a solution of the form q = cJ0 (z). If q (s = L) = Q. sin φ 1 (the cable is √ nearly vertical). uncoupled lateral dynamics follow the equation 2 ∂2q ¯ ¯ ∂ q + wn sin φ ∂q . and then use the substitution z = 2 γs to arrive at the Bessel equation ∂2q ˜ ∂q ˜ +z + z 2 q = 0. we impose a ˜ harmonic input at the top. ˜ ∂s ∂s where γ = (m + ma )ω 2 /wn . what is the “threshold” operating depth? (c) The undamped. indicating a resonance frequency ωn ? ˜ . Simplify this expression by writing ¯ ¯ T = wn s (the vehicle weight eﬀect is small). t) = q (s) cos ωt gives ˜ 2 ˜ ˜ ¯∂q ¯∂ q T 2 + wn sin φ + wn γ q = 0. that is. What condition makes q blow up. Separation of variables q(s. and J0 (z) is the zero’th order Bessel function of the ﬁrst kind.
all you have to do is divide r (s)/δ(s) by another factor of s to account for the fact that dφ/dt = r . especially when x is large. C = [0 1]. −z1 is the zero (there is one real zero).13 1. Include at least the three lowest modes in your sketch. (a) The nondimensional system with states x = [v . δ(s) (s + p1 )(s + p2 ) where K is a constant gain.2U/d. · · ·. Vessel heading control.42 −4.173 (e) Recalling that strumming occurs when ωn 0. You should ﬁnd that strumming is hard to avoid in deep towing applications! 15. the transfer function φ(s)/δ(s) is crucial.. and then use the quadratic formula if necessary to express your answer as s + z1 r (s) =K .25). . deployment depths for which the strumming would be centered. and [−p1 . 2. and discuss the trends for the higher modes. Are both poles stable? (b) For the purposes of autopilot design.8 −2. so you do not need to perform any transformations in this question. i. n = 0. r ] evolves according to dx /dt = Ax + Bδ. Fortunately.3 . where A= −0. B= −0. Sketch all the poles and zeros of the plant transfer function P (s) = φ(s)/δ(s) on a root locus plot.4 . 1. construct a graph of towing velocities vs. write the transfer function r (s)/δ(s) = C(sI − A)−1 B.e. perform the matrix multiplications. −p2 ] are the two poles of the secondorder system. If the output is yaw rate r . J0 (x) = 0 for for x π(n − 0. Use matrix inversion for the 2 × 2 matrix (sI − A).90 −0. All the calculations below are to be made in nondimensional coordinates. We consider the yaw/sway dynamics of a highspeed container ship.
(b) Is the DSRV vehicle openloop stable.000118 Izz = 0. and brings to P (s)C(s) one additional pole at the origin. showing all the dimensional state variables versus dimensional time. (d) Assuming that all the states can be measured accurately. without control action δ? You can assess this either by ﬁnding the eigenvalues of your Amatrix above.000130 Zw = −0. Z ].0m Mq = −0. Demonstrate a rough PID design by devising two controller zero locations that will attract the closedloop poles into the lefthalf plane. Draw the path that you expect each closedloop pole to follow.0439 Zw = −0. and L. U . where K is a 1 × 4 gain matrix.156/U 2 Zq = −0.000146 ˙ Mδ = −0. a fullstate controller δ = −Kx can be used. The coeﬃcients governing the pitch/heave dynamics of a Deep Submergence Rescue Vehicle (DSRV) are given below. Fullstate feedback of a submarine. 16.0m/s L = 15. You do not need to calculate the PID gains that go with your proposed zero locations. q .174 24 PROBLEMS (c) A PID controller C(s) will work to stabilize the heading angle. or by computing the stability parameter C .0364 U = 2. where the input u = δ.0113 Mq˙ = −0. and two arbitrary zeros. and the state vector is ˙ x = [w . create a simulation using Matlab. your approximation for Z should include both w and θ. Under this control. Ixx = 0. θ.0128 Mθ = −0.0315 ˙ Zδ = −0. using the factors ρ/2. the system dynamics are governed by . Z here is the elevation of the vehicle in inertial ˙ coordinates.00193 xG = 0 m = 0.0112 Mw = −0. that is. (c) In preparation for controller design.00157 Mw = 0.0277 (a) Write the linearized (nondimensional) dynamics in the matrix form x = Ax + Bu. as the control gain increases from zero to very large values. Make a graph of the openloop step response.0175 Zq˙ = −0. They are nondimensionalized as in the lecture notes.
i. a highperformance positioning system sets the horizontal position of the cart uo .175 x = (A − BK)x. i.e. u = uo .. Use the Matlab function place to put the four closedloop poles at g(−0. g = 1. u l l Note that the acceleration of the cart is now considered to be the input to the plant.e. y = φ. so that if A − BK has all negative eigenvalues.31i) and g(−0. we assume that all of the mass (m) is concentrated at the distal end of an arm of length l. The inverted pendulum shown below is often used as a simple model for rocket ﬂight. For this problem. Demonstrate that your closedloop design is stable against nonzero initial conditions.. ˙ the system is stable.95 ± 0. ¨ (a) If the observation is taken to be the angle of the bar φ.. e. e. φ m g l uo A Lagrangian derivation of the dynamics gives: 1 ¨ g φ = sin(φ) − cos(φ)¨o . What are the eﬀects of increasing or decreasing g? Note that the pole locations suggested above are with respect to the dimensional system. write the statespace representation of the dynamics linearized about the point φ = 0. using dimensional time.g.81i).59 ± 0. and can also illustrate the dynamic behavior of an unstable ocean vessel which is propelled from the stern. a barge being pushed by a tugboat. Nyquist stability.. .g. 17.
176 24 PROBLEMS (b) Show that this plant is fully state observable and controllable.0m U 0.000456 xG 8m/s Yv ˙ −0. P (s)? Use this plant model for the rest of this question. ∞]rad/s. if any? (e) A stabilizing controller is not hard to ﬁnd.00258 Nv ˙ −0. (c) Now make the assumption that l = 1.00705 0. 2. for which you will need to make explicit calculations.00792 Izz 175. there are no hidden loops or other features. These three frequencies are all that you need to sketch the overall shape. (f) Invoke the Nyquist stability criterion to conﬁrm that the closedloop system is stable. c + jd (c + jd) (c − jd) c2 + d2 Be sure to include the complexconjugate points for −ω on your plot.0116 Yr 0.2. The parameters governing the surface maneuvering of a highspeed container ship are given below for reference: m L Yr ˙ Yδ Nv 0. s+2 C(s) = Create a rough Nyquist plot of the loci of P (s)C(s).000419 0. you may ﬁnd the following identity useful: a + jb (a + jb) (c − jd) (ac + bd) + j(bc − ad) = = . and g = 1 (it can be achieved through a proper nondimensionalization) What is the transfer function of the plant.0000 Nr ˙ −0. In the case of ω = 2. (d) Where are the plant poles and zeros.2. Root locus and loopshaping. (g) About how many degrees of phase margin does this design provide? What reduction in lowfrequency gain can be tolerated? Is it a reasonable design? 18.00242 −0.00385 Nr 0. and a suggested one is: −4(2s + 1) . for the frequencies ω = [0.05 −0.00126 .00222 Nδ −0.0000 Yv −0.
to create a system model of the openloop transfer function P (s)C(s). (c) Apply the kp you selected to P (s)C(s).4 A= . using the rootlocus and loopshaping techniques. and then use the Matlab functions feedback() to create the resulting feedback system. r ] evolves ˙ according to dx /dt = Ax + Bδ. The following steps create two heading autopilots. Give a root locus plot. For the purposes of autopilot design. and . which will be quite far to the left.numC). where −0.3. B= . In addition to the Matlab commands listed below. be sure that you equalize axis scaling for your plots in the complexplane.42 −4. the origin coincides with the center of added mass.13 1. numPC = conv(numP. (b) Using τd = 2 and τi = 6 as suggested values. by using axis(’equal’). τi s C(s) = kp 1 + τd s + The actual numerical values for kp . using the plant above and a PIDtype controller: 1 .90 −0. or ss(). for numerators.7.8 −2. that puts the three slow poles in the following sector: 1) minimum undamped frequency (nondimensional) of 0.. so that Yr = ˙ Nv = 0. and τi are to be found in the next step. The relevant output is yaw rate: C = [0 1] and D = 0..177 Note that the center of vessel mass is located aft of the origin.g.3 −0. Also. e. You don’t need to show the fourth. you will ﬁnd very useful the convolution function conv() which can be used to combine systems. and 3) minimum damping ratio 0. for this model. (a) Use the Matlab command tf(). however. the transfer function φ(s)/δ(s) is needed. fast pole.5. use the Matlab command rlocus() and then rlocfind() to select a controller gain kp . τd . The nondimensional system with states x = [v . with your pole locations clearly marked on top of the trajectories taken as kp varies. 2) maximum frequency of 0..
how does it diﬀer. for the rootlocus design. and what would L(s) have to contain to make it a PID? (h) The controllers you just designed are in nondimensional time coordinates. Make a visual estimate of the gain and phase margins. This design has inﬁnite gain margin and 90 degrees phase margin. . invert the plant to come up with a compensator: C(s) = L(s)/P (s). (f) As above. and D gains for use on a real time scale.178 24 PROBLEMS step() to plot the closedloop system response to a step input in desired heading.0. and plot the closedloop step response. .I. luckily. where ωc = 2. (e) An alternate approach for controller design of this stable plant is loopshaping: For the openloop function L(s) = ωc /s. (d) Use the Matlab command nyquist() to make a Nyquist plot of P (s)C(s) for your design. LQR. 19. choose the positive one. the expression for p22 is messy. (g) The loopshaping control is not quite a PIDcontroller. create the feedback system. (a) What is the control gain K in terms of ρ? Hints: There are two solutions for p12 . Also. The plant is an undamped oscillator with undamped poles at ±j. Consider the statespace system and LQR design: A = B C D Q R = = = = = 0 −1 1 0 [1 0]T [0 1] 0 CT C ρ. give the P. Note that the plant output is position for this problem. you won’t need to use it.
the KF Riccati equation yields H = [100 14]T . 0 0]. ˙ for which the output is y = x. You can get the characteristic equation for the poles as det(sI − 2 (A − BK)) = 0. and V2 = 0. giving the frequency and damping ratio of the Butterworth pattern in terms of ρ. Consider the ﬁrstorder. LQR. (c) Now look at the expensivecontrol problem. and compute the eigenvalue (there is only one) of the closedloop system. . 20. Write the leadingterm approximations for P and then K. for the case Q = C T C and R = ρ. (a) Solve the LQR Riccati equation for P . with ρ → ∞. C = [0 1] .179 (b) Determine the limiting approximations for K with ρ very small and very large – these are the cheap control and expensive control problems. (c) Derive the limiting closedloop pole locations for ρ −→ 0. and ﬁnd the closedloop system eigenvalue. unstable system governed by x = x + u. LQG/LTR. for the choices V1 = [100 0 . For the inverted pendulum plant (see previous question) with statespace matrices A= 0 1 1 0 . The statespace matrices are A = B = C = 1. write the approximations for P and K. (a) Compute the openloop transfer function for the Kalman ﬁlter loop: L(s) = C(sI − A)−1 H. if M is a scalar instead of a matrix. Your answer should give a positive P as a function of ρ alone. and then make it ﬁt the form s2 + 2ζωn + ωn = 0. 21. Note that eig(M ) = M . with ρ → 0. Again.01. (b) Consider the cheapcontrol problem.
This entire design is made in nondimensional coordinates. (a) Plant Modeling and Characteristics .027 .1415 . and what kind of tracking performance should we expect from the associated LTR design? (c) Now consider the closedloop transfer function.03 .006326 . The parameters for the linearized sway/yaw motions of a swimmer delivery vehicle are given below.180 24 PROBLEMS (b) Even though this L(s) is unstable.016 . Yr = 0. What is the characteristic equation for the closedloop system. Nrdot = 0.0 . and about what damping ratio has the Kalman ﬁlter provided? 22.L] U = 4. What is the lowfrequency gain for this L(s). Ndelta = 0. Nr = 0. which you can write as S(s) = L(s)/(1 + L(s)) (the sensitivity). m = 0. Yrdot = 0. . You are asked to develop an LQG/LTR controller for this plant. Yv = 0.055 .0034 . .3 . Nvdot = 0. all nondimensional except [U.013 . % m Izz = 0. Ydelta = 0. % Parameters. Nv = 0. xg = 0. its magnitude plot conﬁrms that it is a reasonable loopshaping design. Please make sure you answer all the questions.0074 . and include a listing of your code. and it is suggested that you compose a single Matlab script to perform the steps in sequence. .1 . Yvdot = 0. LQG/LTR design. % m/s L = 5.
Construct a statespace plant model. and postmultiplying it by C: ˙ x = (A − BK)x + BKC T ydesired y = Cx. 0. ii. B. the input to the closedloop system is xdesired and the output is x. For the choices V1 = I3×3 and V2 = 0. and make a plot of the closedloop step response. V1 .00001]. Be sure to give the numerical values of H. . C matrices. B. C. for given A. given design matrices A. and V2 . Using the Matlab command lqr(). ii. i. to take rudder angle δ as an input and give heading angle φ as an output. (b) LQR and KF Designs i.001. This compression can be achieved in one step by premultiplying the system by C T . How do the gains and step responses change as you make ρ smaller and smaller? Note that the fundamental closedloop LQR system is ˙ x = (A − BK)x + BKxdesired y = x. Q. Where are the poles of your plant model? Is this model stable? iv. Compute and list the controllability and observability matrices. There should be three states in your model. and R = ρ. The Matlab command lqe() can be used to generate the Kalman ﬁlter gain H. and R matrices.e.01.. φdesired = 1]. list K and plot the closedloop step responses for the choices ρ = [0. compute H. With the choices Q = C T C. Your plot should show speciﬁcally the output φ. is the plant statecontrollable and stateobservable? iii. Show a plot of your plant’s step response. 0.1. for an input of xdesired = [vdesired = 0. with one input channel and one output channel. Please provide the numerical values for the A.181 i. you can compute the LQR feedback gain K. rdesired = 0.
and small at high frequencies. you should set this to I3×3 . L(s) should be large at low frequencies. the product P (s)C(s) → L(s). versus log(frequency). consistent with the rules of loopshaping. Make a closedloop step response plot for the smallest value of ρ.. The closedloop KF system is as follows: ˙ x = (A − HC)ˆ + Hy ˆ x y = C x. in Matlab. the input is the measurement y and the output is an estimated version of it. As ρ → 0. Demonstrate this by computing P (s)C(s) for the three diﬀerent values of ρ above. You may ﬁnd the Matlab command freqresp() helpful. and its output u is the control action to be applied as input to the plant. ˆ ˆ i. and crossover frequency of the openloop KF system L(s) = C(sI − A)−1 H.g. the compensator C(s) has the following state space representation: ˙ z = (A − BK − HC)z + He u = Kz.e. . e.. i. How does it compare with the KF step response of part 2b)? In real LTR applications. so that the input to the compensator is the tracking error e = r − y. you may simply multiply the systems. ii. and overlaying the respective P (s)C(s) over the plot of part 3a). The total openloop transfer function is the P (s)C(s).182 24 PROBLEMS Note that the lqe() command asks for a disturbance gain matrix G. ˆ (c) Loop Transfer Recovery The LQG compensator is a combination of the KF and LQR designs above. With normal negative feedback. Make a log(magnitude) plot of the KF openloop transfer function L(s) = C(sI − A)−1 H. iii.. the particular values of V1 and V2 can be picked to control the lowfrequency gain. sysPC = sysP * sysC . y .
up) axes of an underwater vehicle. A triaxial accelerometer package is aligned with the [x. Note that under the x ¨ ¨ inﬂuence of gravity alone. (a) Ignoring the coupling of Euler small angles.81m/s. (b) Given that the acceleration due to gravity is 9. At a particular instant in time.z] (fwd.y.80. the three raw accelerations from the strain guages are [¨ = −0.73] meters/second. y = 0. what estimates can you give for the roll and pitch angles? I suggest that you draw the measurement vector in the coordinates of the vehicle frame.183 23. z = 9. port. and then consider the orientation of the body which would lead to it. how do we know that these are reasonable (but not foolproof!) angle calculations to make? . this sensor will always report that the vehicle is accelerating upwards. Inertial navigation.96.
5kg/s 70000N m 60Hz 32 0. lies properly on the seaﬂoor.07ρ 0.19 5585m3 0. which is being paid out constantly and sinks slowly under its own weight. so the vessel will need to slow down and speed up frequently. Unexpected deviations in speed from the calculated trajectory are likely to create loops or kinks in the cable.97 .025 0.1 1.876 0. or induce large tensions. The bathymetry for the path is wellknown but variable.522 0.184 24 PROBLEMS 24. length L draft h beam b displacement surge added mass ma thrust reduction factor t wetted surface area Aw towed resistance coeﬃcient Cr wake fraction at propeller w propeller diameter D pitch/diameter ratio P/D rotative eﬃciency ηr zerospeed thrust coeﬃcient kt (J = 0) slope of kt with J zerospeed torque coeﬃcient kq (J = 0) slope of kq with J rotational moment of inertia Ip engine parameter a b c d maximum fuel rate fm maximum torque Qm maximum speed nm gearbox reduction ratio λ eﬃciency ηg vessel 145m 9m 22m 17750m3 0.6 0.0833 2. You are asked to carry out modeling and control system design for the surge dynamics of an oceanic cablelaying vessel.03 × 105 kgm2 0.1 0.173 .22 7.0939 1. LQG/LTR design: term project.21m 1. This vessel needs to have very good speed control so that the cable.208 2.0025 0.
. propeller rotational speed. (c) Construct a linear approximation for the plant dynamics around the operating condition fo /fm = 0. Plot the pairs [fo /fm . make a contour plot which gives Qe /Qm (the fraction of maximum torque developed) as a function of f /fm (the fraction of the maximum fuel rate) and ne /nm (the fraction of the maximum rotation rate of the engine). and then demonstrate that it will work with a simulation of the real nonlinear system.g. f the fuel rate in kg/s. Qm fm nm fm where Qe represents the engine torque in N m. and the eigenvalues. e. and vessel linear speed. and engine system are as given in the table. The vessel is powered by a single gas turbine engine.95.. For instance. The parameters for the vessel. propeller. driving one propeller. (a) Make a map of the gas turbine characteristics. C matrices for your plant. the engine cannot develop negative torque. engine torque. where the calculated torque exceeds Qm .185 It is desired to create a linear controller for the nominal surge motion. and the output is δu. engine. and ne the rotation rate of the engine in Hz. B. If the plant input is δf .8. neo /nm ] on the characteristic plot from Part 1. and assess whether the vessel. δ here indicates the perturbed value from the steady state. in which you show: advance ratio seen by the propeller. within a neighborhood of the nominal speed. list the A. Similarly. (b) Make a table of some steady operating conditions for fo /fm in the range of 0. Be sure to include clipping on your contours. The gas turbine torquespeed characteristic ﬁts the relation: f ne f Qe =− a +b + c +d . and propeller are well matched to operate over a range of fo . f = fo + δf .05–0. engine rotational speed.
. When this is done. for example. with input f . An integral action can be added quite easily in the LTR design technique. but you cannot easily change its shape. This is to be kept as a separate controller design. This is not acceptable in the long term. so that an “augmented plant” is created. You should prepare for this part: a plot of log(L(s)) versus log(ω). This choice corresponds with the closedloop system being able to overcome waves with period 10seconds or longer. You don’t . a listing of your V1 and V2 choices. This augmented plant has one more state than the original plant: A B 0 0 Aaug = Baug = [0 1]T Caug = [C 0] The idea is to then carry out the KF design and LTR as before.[1 0]) . or move it sidetoside. First. Note that for this scalar design. given that the vessel is laying out a cable. and the A. B. the actual vessel speed will vary from the nominal value. First. (e) In the event of a current or wind disturbance. C matrices of your LQG compensator. (f) For the purpose of demonstrating your controllers. with the augmented plant model. Secondly. you will be able to move the curve L(s) up or down on the Bode plot. construct a nonlinear simulation of the plant system. Present a ﬁgure of L(s) and the other quantities as you gave them in Part 4. The addition of an integrator channel to the plant or compensator can be accomplished with the command sysPaug = sysP * tf([0 1]. add an integrator to the input of the plant. recover this loop shape with an LQGtype controller. using a small control penalty. choose the ﬁlter gains V1 and V2 to make the crossover frequency (where the magnitude of L(s) = C(sI −A)−1 H passes through unity) about 0.186 24 PROBLEMS (d) Create an LTRtype controller for the plant. move the integrator from the augmented plant into the compensator.6rad/s.
If the compensator has states z. try . it is mostly to verify that the steady conditions you calculated above. Cc ]. Demonstrate four properties. showing timedomain simulations for each: • The ﬁrst controller. np ]. reject some lower frequency disturbances. The compensator states z are propagated by a set of system matrices [Ac . For example. • The ﬁrst controller does. You will propagate the ﬁrst two states as usual with the nonlinear equations and the fuel rate f as input. In the absence of a controller.187 need to show anything for this task. np . Bc . are actually steady conditions from the point of view of the whole simulation. however. which you just designed. Recall that you are implementing a controller for operation about a nominal condition. with f = 0. thus modelling the eﬀects of steady wind or current. A suggestion is to let uprime = uprime + k*sin(10*omega_c*t). does not completely reject steady disturbances. the part needed to keep us near the steady condition. the steady condition speed. The output of the compensator is then fuel rate f . (g) Augment your simulation above with the compensator dynamics. We are interested mainly in disturbance rejection. where u is the speed of the vessel and r is a reference speed. this line would drive the system with an additional acceleration of km/s2 . This adds an oscillating acceleration at ten times the crossover frequency. Note that the input to the compensator is an error signal: e = r − u. be sure to then add on 0. make a new state vector [u.8fm . I suggest you implement a disturbance as an acceleration: uprime = uprime + k. z]. • The ﬁrst controller does not reject high frequency disturbances. You may wish to use global variables in Matlab so that you don’t have to type everything in more than once. without integral action. and your simulation has states [u. for which you will just set r = uo . where omega_c is the crossover frequency.8fm . whenever your compensator generates f from the error signal. This line is to be added after computing the usual parts of uprime.
• The second controller. velocity u. rejects steady disturbances. Use the same approach as above for steady disturbances (h) What is the propeller response during the highfrequency disturbances? Is it reasonable or do we need to slow down the controller bandwidth? . The eﬀect of an acceleration disturbance onto the observed output. scales with the inverse of the frequency. with integral action.188 24 PROBLEMS uprime = uprime + k*sin(omega_c/10*t)/100. Note the extra factor of 100 is included so that the result can be compared directly with the highfrequency disturbance case above.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue listening from where you left off, or restart the preview.