Chapter 5: Standard Models for Dynamic Systems

• So far we have developed system models that are collections of either first-order ODEs (i.e., an electrical circuit
with a single capacitor or single inductor), and/or second-order ODEs (i.e., mechanical systems)
• “Standard forms” of dynamic systems will facilitate the solution of the system response, using either analytical
or numerical methods (i.e., MATLAB/Simulink)
– State-variable equations
– State-space representation (SSR)

– Linearization

– Input-output (I/O) equations

– Transfer functions

– Block diagrams

• It is important to remember that all standard model forms represent the same governing dynamics for a system
(i.e., the models do not change the dynamics)
 5.2 State-Variable Equations
• One method for representing system models is to use state variables >> A set of variables that completely
define (or describe) a system
– The number of state variables is equal to n, which is the order of the system modeling ODEs.

• State: the smallest set of dynamic variables that completely define a dynamic system
– These variables are called state variables and are usually the physical variables of the system (displacement, velocity
for mechanical systems; voltage, current for electrical systems; pressure, temperature for fluid/thermal systems)
– If one knows the state of a system, then any quantity can be determined (such as energy, momentum, etc.). However,
the choice of state variables is not unique.

• The standard convention is to define xi , i = 1, 2, 3, … n as the state variables (the “dynamic variables”), and

uj, j = 1, 2, … r as the input variables (or, “control variables”)

 State-Variable Equations (2)

• The state-variables equations are a collection of n first-order ODEs, where the all right-
hand sides must be functions of the state variables or input variables:
x 1  f1 ( x1 , x 2 ,..., x n , u1 , u2 ,..., u r )
x 2  f 2 ( x1 , x2 ,..., x n , u1 , u2 ,..., u r )

x n  f n ( x1 , x2 ,..., x n , u1 , u2 ,..., u r )

• The RHS functions fi can either be linear or nonlinear functions

– If all RHS functions are linear, then we can use “state-space representation” (SSR), a convenient
vector-matrix format (next sub-section)
 State-Variable Equations:
Example 5.1
• Determine the state-variable equations for the system modeled by the following ODEs,
where z and w are the dynamic variables and v is the input
2
z  0.8 z  0.4 w  0.2 zw 0
4 w  3w  0.1w3  6 z 8v

• Order of the system?  Number of state variables required

• (worked on board)
2
z  0.8 z  0.4 w  0.2 zw 0
4 w  3w  0.1w3  6 z 8v
 State-Variable Equations:
Example 5.2 (Practice Problem)
• Derive the state-variable equations for the mass-spring-damper mechanical system with
3
a nonlinear spring force f k ( z ) k1 z  k3 z

mz  bz  k1 z  k 3 z 3  Fa Mechanical system

(nonlinear)

Do Example 5.3 as home practice problem

f k ( z ) k1 z  k 3 z 3

mz  bz  k1 z  k 3 z 3  Fa
 5.3 State-Space Representation (SSR)

• For modern, complex dynamic systems, we often encounter systems with multiple inputs and multiple
outputs (MIMO)
– IF the system is linear, we can set the state-variable equations in a convenient vector-matrix format, which we will
call a state-space representation (SSR)
– The SSR vector-matrix format that is extremely well-suited for computer simulation (MATLAB and Simulink)

• State vector: For a system of order n, we require n state variables to completely define the system
– We arrange these state variables in a column vector called the state vector, x

– The state vector is not a physical vector (like a 3-component force vector) but rather a convenient grouping of the
important dynamic variables.

• State space: An n-dimensional “geometric space” that contains the state vector x
 State-Space Representation (2)

• In order to obtain a complete state-space model, we must obtain n first-order, linear differential equations
• For example, if our system is LTI and third order (n = 3) with two inputs (r = 2), then our state-variable
equations must have the form:

x 1 a11 x1  a12 x 2  a13 x3  b11 u1  b12 u 2

State equations:
x 2 a 21 x1  a 22 x 2  a 23 x3  b21u1  b22 u 2 linear combinations
of xi and uj
x 3 a31 x1  a32 x 2  a33 x3  b31u1  b32 u 2
 State-Space Representation (3)
• Next, we need equations for the system outputs, so we write the outputs as linear combinations of the states
and inputs

• Suppose that we have two outputs (m = 2; perhaps these variables are measured by two sensors):

y1 c11 x1  c12 x 2  c13 x3  d11 u1  d12 u 2 Output equations:

linear combinations
y 2 c 21 x1  c 22 x 2  c 23 x3  d 21u1  d 22u 2 of xi and uj

• For time-invariant systems, the coefficients a, b, c, and d are constants

 State-Space Representation (4)
• Finally, note that we can write the state and output equations in a compact vector-matrix form because the
equations are linear

• First, assemble all variables (states, outputs, inputs) into column vectors with proper dimensions:

 x1   u1   y1 
x  u  y 
x  2  u  2  y  2 
     
     
 xn   ur   ym 
“state vector,” n x 1 “input vector,” r x 1 “output vector,” m x 1

• Note that the left-hand side of the state equations is the first-time derivative of the state vector x
 State-Space Representation (5)
• We can assemble the a, b, c, and d coefficients of the state and output equations into
four matrices:

 a11 a12  a1n   b11 b12  b1r 

a a 22  a 2 n  b b22  b2 r 
A   21 B   21
• Key to SSR          
   
 a n1 an2  a nn   bn1 bn 2  bnr 
n: member of states n x n (square) matrix: n x r matrix:
“system” or “state” matrix “input” matrix
r: number of inputs
m: number of outputs  c1n 
 c11 c12
 d11 d12  d1r 
[A]n×n c  c 2 n  d
C   21
c 22 d 22  d 2 r 
D   21
[B]n×r          
   
[C]m×n  c m1 cm2  c mn   d m1 d m2  d mr 
m x n matrix: m x r matrix:
[D]m×r “output” matrix “direct-link” matrix
 State-Space Representation (6)
• Finally, we can write the state and output equations in vector-matrix form:

x  Ax  Bu “state equation”

y Cx  Du “output equation”

• Both equations constitute a state-space representation (SSR)

– Note that the inner/outer dimensions of all vectors and matrices must match
 State-Space Representation:
Example 5.6
• Obtain a complete state-space representation (SSR) of the electromechanical DC motor
– A tachometer is used to measure the angular velocity of the rotor (  ), and an ammeter is used to
measure current in the armature circuit ( I )

LI  RI ein (t )  K b Dynamic Variables : I ,  , 

DC motor
model
J  b  K m I  TL x I , x  , x 
1 2 3

• The electrical and mechanical ODEs are 1st-order and 2nd-order, respectively
– Consequently, n = 3 and the system requires three state variables
LI  RI ein (t )  K b

J  b  K m I  TL
LI  RI ein (t )  K b

J  b  K m I  TL
 State-Space Representation:
Example 5.6 (3)
• Complete the derivation on the board… the SSR result is

 x 1    R / L 0  K b / L  1 / L 0 
x  x 2   0 0 1  x   0 0  u
 x 3   K m / J 0  b / J   0  1 / J 
A B State equation

 y1   0 0 1  0 0
y     x  u Output equation
 y 2   1 0 0  0 0
C D
• Check and verify the dimensions of the matrix-vector multiplication!
 5.4 Linearization

• Linearization is a method for converting a nonlinear equation (or model) into a linear model
• Linearization relies on a Taylor-series expansion about a nominal (or reference) operating point (x*
and u*)
• We rely on linearization because there exists a wealth of linear analysis tools and there is no
general, unifying theory for solving nonlinear systems
– Many systems (such as hydraulic systems) are highly nonlinear

• A linearized model is reasonably accurate and representative of the (true) nonlinear model as long
as the dynamic variables do not deviate “too far” from the operating point
 5.4 Linearization

A A
A

A is operating point or point of linearization

df ( x  x0 ) d 2 f ( x  x0 ) 2
f ( x)  f ( x0 )   2
dx x0 1! dx x0
2!
Taylor’s Series expansion df
f ( x)  f ( x0 )  ( x  x0 )
dx x0

 f ( x) m x  x
0
 Linearization: 3 Basic Steps

1. Choose (or solve for) the nominal operating point (or states) to linearize the system about
• The operating point may be given, or it may be an equilibrium (steady-state) point that can be obtained from the
governing nonlinear model. Typically, the nominal operating point will be a static state (constant). Call this
nominal state/input x* and u*

2. Re-define the governing equations in terms of the nominal variables and the perturbation (or deviations)
variables with respect to the nominal values
• For example, use the convention δx = x – x* for the perturbation (deviation) from the nominal state x*, and δu =
u – u* for the input perturbation from nominal input u*

3. Expand all nonlinear terms in a Taylor series about the nominal value and retain only the first-order

(linear) terms
 Linearization: Process
• Suppose we have a first-order nonlinear system with one input u

x  f ( x, u ) RHS function f is nonlinear

Step 1: The nominal input u*(t) results in nominal state x*(t) which is the nominal operating point

Step 2: The perturbation variables are δx = x – x*, and δu = u – u*. Re-write the NL equation in terms of
perturbation and nominal variables:

x  x *  f x  x*, u  u *

Step 3: Expand the RHS in a Taylor series about x* and u*

f 1 2 f f 1 2 f
x  x *  f x*, u * x  2
x 2
   u  2
u 2

x * 2! x * u * 2! u *

Cancel!
Nominal state solution f f Linear model in terms of δx and δu
x  x  u
x * u * The solution will yield δx(t)
 Linearization: Summary
• Suppose we have a second-order system with two nonlinear state-variable equations, and one input u

x1  f1 ( x1 , x2 , u ) Both RHS functions f1 and f2 are nonlinear

x 2  f 2 ( x1 , x2 , u )

Step 1: The nominal input u*(t) results in nominal states x1*(t) and x2*(t)
Therefore, the nominal state solutions x* are the nominal operating point

Step 2: The perturbation variables are δx1 = x1 – x1*, δx2 = x2 – x2*, and δu = u – u*

Step 3: After the Taylor-series expansion (to first order), the resulting linearized
system can be put into SSR:
 f1 / x1 f1 / x2   f1 / u 
 x A x  B u where A = 
f 2 / x2 
and B =  
 f 2 / u 
 f 2 / x1
The matrices A and B are numerically evaluated at the nominal states x* and u*
Note that the solution to the linearized system gives us δx(t), the perturbation state vector. In order to estimate the state
solutions, we have x(t) = x*(t) + δx(t)
 Linearization: Example 5.9
Perform the linearization about the static equilibrium state x∗ that results from the
x  2 x  0.4 x  0.3u  f ( x, u )
3
nominal input u∗ = 2.

Step 1: The nominal input u*(t) results in nominal states x1*(t) and x2*(t). Therefore, the nominal state solutions x* are
the nominal operating point

Step 2: The perturbation variables are δx1 = x1 – x1*, δx2 = x2 – x2*, and δu = u – u*

Step 3: After the Taylor-series expansion (to first order), the resulting linearized system can be put into SSR:

Solution on whiteboard
 Linearization: Example 5.10

CP Qin  KT P  Patm 0

 Qin Qin* ; P P*
*2
Q Nominal pressure Or
P*  in2  Patm Operating point
KT
CP Qin  Qout
 P  P  P* ;  Qin Qin  Qin*
Qout KT P  Patm f f
P  P  Qin Taylor’s series
P * Qin *
CP Qin  KT P  Patm
f  KT f I
P  1 Q  KT P  Patm  f P, Qin 

P * 2C P  P
*
;
Qin

C
in *
C C atm

Over time the outflow Qout will balance the constant input flow Q*in f  KT2

P 0 >> Pleasure reaches a constant value >> P* P * 2CQin*
 Linearization: Example 5.10

f  KT2 f I
 
P * 2CQin* Qin *
C

f f  KT2 I
P  P  Qin  P  P  Qin
P * Qin * 2CQin* C

Study all relevant example problems from the book

 5.5 Input-Output (I/O) Equations
• Consider a single-input, single-output (SISO) dynamic system, represented by the generic “block diagram”
below:

• An input-output (I/O) equation is a differential equation that only involves the input variable u, output
variable y, and their derivatives

• The general form of an I/O equation (for an nth-order system) is

an y ( n )  an  1 y ( n  1)    a2 y  a1 y  a0 y bmu ( m )    b1u  b0u

For systems with two or more inputs, the right-hand side of the I/O equation
will include additional input terms

• If we have a system with m output variables, we will have m I/O equations, one for each output variable.
Therefore, unlike the state-variable equations, we can solve each I/O independently from the others.
 I/O Equations: Simple Example
• Consider again the 1-DOF mechanical system (mass-spring-damper with input force Fa ):

mz  bz  kz Fa

• Let’s define the output variable as position, y = z and input variable as u = Fa

• Therefore, our system model equation becomes

my  by  ky u I/O Equation

• Since this equation only involves output y and input u, it is the I/O equation

• Systems with multiple states/outputs require significant algebra to derive the I/O equations
 Differential Operator
• For more complicated systems with multiple outputs, developing the I/O equations involve much more effort
than obtaining the SSR

• A useful method is to simplify the algebra is to employ the differential operator or “D operator,” defined
below: d
D
dt
• Therefore, time-derivatives can be written as powers of operator D, for example

Dy  y D 2 y  y etc

• We can use the D operator to manipulate the governing dynamical equations in order to obtain the desired I/O
equation
 Differential Operator: Example 5.13

Derive the I/O equation with the displacement

of mass m1 as the output, and Fa(t) as the input
variable

Applying the D operator

See the and practice the remaining

part from the textbook
 5.6 Transfer Functions
• Transfer functions (TFs) are a convenient way to represent and analyze the input-output (or cause-and-effect)
relationship of a SISO (LTI) dynamic system
– Transfer functions are based on Laplace transform methods

• Laplace transforms are an operation method that can be used to solve linear differential equations (ODEs)

• Laplace transformations convert linear differential equations into algebraic equations in terms of the Laplace
variable “s”

• Transforming a time function f(t) from the time domain to the Laplace (or “s”) domain is defined by the
integral:


L  f (t ) F ( s) 0 f (t )e  stdt Laplace transform
 Transfer Functions (2)
• We can use the Laplace transformation operation to transform standard time functions (step function,

impulse, sinusoid, etc) into the Laplace (or “s”) domain, and therefore Laplace tables can be constructed in

order to aid analytical calculations

• Chapter 8 presents a brief overview of Laplace transform theory and the solution of differential equations

using Laplace transform methods

• Transfer functions are commonly used in block diagrams (“simulation diagrams”) and numerical algorithms

such as MATLAB and Simulink

• However, we may develop (and use) transfer functions without formal use of the Laplace transformation

method
 Developing Transfer Functions
• In order to develop the transfer function in the Laplace domain (without dealing with Laplace transforms),
consider the following third-order I/O equation:

a3y  a 2 y  a1 y  a0 y b1u  b0 u

• Now, consider an exponential input, u(t) = U(s)est, where s = s + jω is a complex variable and U(s) is a
complex function (recall that j is the imaginary number, j   1 )

• Using the Euler identity, e j cos   j sin  the exponential function is

e st et e jt et cos t  j sin t 

• Therefore, est will be complex (real + imaginary parts)

– Since input u(t) is strictly real, U(s) will be the complex conjugate of est
 Developing Transfer Functions (2)
• Recall from differential equation theory, that if the input is u(t) = U(s)est, the particular solution
(response due to the input) will also be an exponential function, y(t) = Y(s)est

• Now, substitute the exponential input and output equations into the I/O differential equation, with the
knowledge that y s 2Y ( s )e st and y sY ( s )e st etc…

a s
3
3
 a 2 s 2  a1 s  a 0 Y ( s )e st b1 s  b0 U ( s )e st
• Finally, form the ratio Y(s)/U(s), or output/input (note est cancels) :

Definition: The transfer function of an LTI

Y ( s) b1 s  b0 system is the ratio of the Laplace transform of
 3 2
G ( s ) the output, Y(s), over the Laplace transform of
U ( s ) a3 s  a 2 s  a1 s  a0
the input, U(s), with zero initial conditions.

“Transfer Function” G(s)

 “Time-Domain” Transfer Function
• In order to demonstrate a version of the transfer function in the time domain, consider again the same third-
order I/O equation:
a3y  a 2 y  a1 y  a0 y b1u  b0 u
• Next, apply the D operator notation:

a3 D 3 y  a 2 D 2 y  a1 Dy  a0 y b1 Du  b0 u
or, factoring out y(t) and u(t)

a D
3
3
 a 2 D 2  a1 D  a 0 y (t ) b1 D  b0 u (t )

• Finally, form the ratio y(t)/u(t), or output/input:

y (t ) b1 D  b0 “Time-domain” TF is the same

 as formal definition of the TF
u(t ) a3 D 3  a 2 D 2  a1 D  a0
if we replace D with s
 Transfer Functions: Notes
• The prior slide shows that we can simply interchange the s and D symbols to produce transfer
functions (purely for the sake of representing dynamic systems)
• It should be stressed that while MATLAB and Simulink very often rely on TFs (which use the
complex variable s) to represent I/O differential equations, their numerical solution techniques are
based on the time-domain differential equations and not the Laplace-transform solutions
• Some comments on transfer functions:
– Transfer functions are limited to LTI systems

– TF is a mathematical model of a particular system (I/O equation)

 Transfer Functions:
Example 5.14
• Derive the transfer function for the given I/O equation of a series RLC circuit, where current is
the output variable y and source voltage is the input variable u

y  8 y  10 y 4u I/O equation for RLC circuit

• Apply the D operator to all terms in the I/O equation:

D 2 y  8 Dy  10 y 4 Du

• Form the ratio of output/input: y (t ) 4D

 2
u (t ) D  8 D  10

• Replace D with s 4s Transfer

G( s)  2
s  8s  10 function
5.7 Block Diagrams
 5.7 Block Diagrams

• Block diagrams are standard pictorial or graphical representations of interconnected

systems
– Each dynamic system or component is a block, usually a transfer function
– Blocks are connected by signal paths, which represent the flow of input and output signals and
computations
– Signal flow into a block represents a math operation, usually multiplication

• Simulink is based on block diagrams, which are constructed from “drag-and-drop” menus
in the Simulink environment
 Standard Block-Diagrams Components
• Multiplication by a constant (or gain), K

• Integration:

Simulink notation
 Standard Block-Diagrams Components (2)
• Dynamic system represented by transfer function G(s)

3s  2
For example, if the transfer function is G( s)  2
s  4 s  12
the corresponding I/O equation is
y  4 y  20 y 3u  2u

Differentiation in the time domain (multiply by D) is equivalent

to multiplying by “s” in the Laplace domain
Remember:
Integration in the time domain (dividing by D) is equivalent to
dividing by “s” in the Laplace domain
 Standard Block-Diagrams Components (3)

• Summing junction (adding and/or subtracting signals)

Signs must be provided at the summing junction

 Block Diagram: Example 5.17
Consider the electromechanical DC motor system. Draw the complete block diagram of the DC motor using
transfer functions to represent the armature circuit and mechanical rotor dynamics. The desired system output is
angular velocity of the rotor, 𝜔.
• Draw the block diagram for the DC motor system:

LI  RI ein (t )  K b Armature circuit dynamics

Two coupled, linear
first-order ODEs
J  b K m I  TL Rotor dynamics

where ein(t) = source voltage (an input)

Kb = back emf constant, Km = motor torque constant,
TL = load torque (an input)
L = inductance of coil, R = resistance of coil
I = current (“dynamic variable”)
w = angular velocity of motor (“dynamic variable”)
J = motor inertia, b = motor friction coefficient
 Block Diagram:
Example 5.17 (2)
Armature circuit dynamics
• Work example on the board… see textbook for details
LI  RI ein (t )  K b
• Resulting block diagram (verify that blocks match the model!) Rotor dynamics

J  b K m I  TL
 5.8 Standard Input Functions
• The subsequent chapters will emphasize obtaining the output or system response to a desired input
function

• System performance is often characterized by the system’s dynamic response to any number of
standard input functions

• We can think of these “standard” input functions as “test input signals” for characterizing the
system’s dynamic response

• Many standard input functions have a basis in the realistic (expected) input for a dynamic system
 Step Input
• The step input shows a sudden, instantaneous change (step) from one constant value to another
constant value
U(t)
Unit-step input
1
0 for t 0
U (t ) 
1 for t  0 t
0

• Many times, the step input is initially zero and instantaneously steps to a constant value

• Step input with magnitude of A : u (t )  AU (t )

 Ramp Input
• A ramp input function increases linearly with time at a constant rate

u(t)

Unit-ramp input u (t ) t Slope = 1

t
0
• A general ramp input is u (t ) at
– where a is the slope, which could be positive or negative
 Ramped Step Input
• The input “ramps” up from zero to a constant (step) value

• The ramp-up component shows a linear increase with time

• The ramped step input might be a more realistic representation of a step input, since most
physical inputs require a finite time interval to change from one constant value to another
constant value

Ramped step input u(t)

A
at for 0 t  A / a
u (t ) 
A for t  A/ a
t
0
 Pulse Input
• The input has a constant (finite) magnitude over a finite time interval (the “pulse duration”),
and is zero before and after the pulse is applied

Pulse input
u(t)
 0 for t 0 A

u (t )  A for 0  t  t1
 0 for t t1

t
0 t1
 Impulse Input
• The input has a magnitude that lasts for a very short duration

• We can think of the impulse input as a pulse function where the pulse duration goes to zero in
the limit, i.e., dt  0

• Suppose a constant force F = 150 N is applied for a very short duration Dt = 0.1 sec (a pulse
input)
– Therefore, the impulse input has a “strength” of 15 N-s, which is the area of the original
pulse function
– Note that the units of an impulse input contain units of time; for a force input, an impulse has
units of N-s.
 Impulse Input (2)

• Pulse force inputs with an impulse of 15 N-s

 Impulse Input (3)
• Mathematically, an unit impulse input is represented by the Dirac delta function, δ(t), which has
infinite magnitude but infinitesimally small-time duration, and hence the “area” of the pulse is
unity
 (t) 0, for t 0

  (t)dt 1
 f(t), N

f (t ) 15 (t ) N

t
0
 Sinusoidal Input
• A sinusoidal input function is a repeating, periodic input that can be represented by sine
and/or cosine functions
Harmonic input fuction (frequency is 0.5 Hz)
3

Sinusoidal input 2

u (t )  A sin t  B cos t 1

Input
0

A and B are amplitudes -1

and w is the frequency in rad/s

-2

Frequency is sometimes expressed -3

0 1 2 3 4 5
in cycles/sec or hertz (Hz) Time, sec

w = 2p rad/s = 1 Hz
 Chapter 5: Summary
• We have introduced and discussed the standard forms for representing the mathematical models of
physical systems:
– State-variable equations
– The state-space representation (SSR)
– Input-output (I/O) equations
– Transfer functions
– Block diagrams

• Of course, the governing system dynamics (i.e., the ODEs) are not altered if we choose to represent
the system in one or more different standard forms
• We also presented the basic steps for the linearization process which allows us to develop a linear
model that represents the system dynamics in the vicinity of a nominal operating point