CEE490b Feb.

11, 2002

Review of Random Loading Concepts - II

Assuming a periodic signal with the form: x (t ) = a sin ωt

T
1
The mean is: x (t ) = x =
T ∫ a sinω t dt = 0
T → L arg e 0

∫ (x(t ) − x ) dt
T
2 2 2 1 2
The variance is: x (t ) = x = σ = x
T
T →L arg e 0
T
1 2
= ∫
T 0
a sin 2 ω t dt

T
a2 1
=
T ∫ 2 (1 − cos 2ω t ) dt
0

a2 T a2
= =
T 2 2

1
The Standard Deviation (RMS) is: σ x = x2 = a
2

The Peak Value is: x (t ) max = xˆ = a = 2σ x

The 2 is a “Peak factor” g = 2

Identical expressions result from assuming a periodic signal of the form:

x(t ) = a cos ωt

The Probability Density Function, p x (x ) is:

the probability that x1 ≤ x ≤ x1 + dx = p x ( x1 )dx

dθ x = x1
= or the proportion of time between θ and θ 1 over the total time
θ
where θ = sin −1 (x / a )

d ( x / a) dx dx
dθ = = =
2 2 2 2
1− x / a a 1− x / a a − x2
2

Page 1 of 6
CEE490b Feb. 11, 2002

1.2

for x1 ≥ 0 , 0 ≤ θ ≤ π
dθ 1 dx
p x ( x1 )dx = = p x (x )
π π a 2 − x12 1

for x1 ≤ 0 , π ≤ θ ≤ 2π
dθ 1 dx
p x ( x1 )dx = = 0.8

π π a − x12
2

1 dx 0.6
Therefore; p x ( x ) =
π a2 − x 2

1 0.4
At x = 0; p x ( x ) = , and at x = a;
πa
p x (x ) → ∞ ,
0.2

1
πa
0

a 0 a
The Autocorrelation Function is:

T
1
∫ sin(ωt + φ ) sin(ω (t + τ ) + φ )dt
2
R(τ ) = a
T 0
T → l arg e
T
1
= a 2 ∫ sin(ωt + φ )[sin(ωt + φ ) cos ωτ + cos(ωt + φ ) sin ωτ ]dt
T 0
T T
a2 a2
T ∫0 T ∫0
2
= sin (ω t + φ ) cos ωτ dt + sin(ωt + φ ) cos(ωt + φ ) sin ωτ dt

(due to orthogonality of sin & cos = 0)

T
a2 1
Therefore: R(τ ) =
T ∫ 2 (1 − cos 2(ωt + φ )) cos ωτdt
0
2
a T a2
= cos ωτ = cos ωτ
T 2 2

The normalized autocorrelation function is normalized by the variance, σ 2

R(τ ) (a 2 / 2) cos ωτ
R ′(τ ) = 2 = = cos ωτ
σ a2 / 2

Page 2 of 6
CEE490b Feb. 11, 2002

Relationship Between Input and Output

The mean-square response of the output (i.e. response) was shown in Eq. 8.32 to
be composed of two parts; the first attributable to the quasi-steady “Background”
component and the other to the “Resonant”:

f ∞
1 o 1
y ≅ 2 ∫ S p (f )df + 2 ∫ S p (fo ) H (f ) df
2 2

k o k 0
fo
(8.32)
1 1 πf
= 2 ∫ S p (f )df + 2 S p (fo ) o
k 0 k 4D

A similar, but more convenient approximation to this expression recognizes the fact
that the integral of the input spectrum from 0 up to the natural frequency fo is
approximately equal to the integral over the entire spectrum of input, when fo is
reasonably high, compared with the dominant frequencies of the input spectrum.
This is the case for most applications involving wind, for example. The equation for
the variance reduces to:
1 2 1 πf
y2 ≅ 2
σ p + 2 S p (fo ) o
k k 4D
σ2  π foS p (fo ) 
≅ 2p 1 + 
k  4D σ p2 

The Non-dimensional Force Spectrum

The Resonant Response

The Background Excitation

(a “Static” relationship between Input and Output)

σ y2 = σ yB
2 2
+ σ yR

or, the variance of the Total Response equals the sum of the variance
of the Background Response plus the Resonant response

2 2
In many situations, especially when the damping is low, σ yB << σ yR and then:

1 π
σ y2 ≈ σ yR
2
≈ f o S p (f o )
k 2 4D

Page 3 of 6
CEE490b Feb. 11, 2002

Properties of Power Spectra

S x (f1 )df = the contribution to the total variance of x (i.e. σ x2 ), due to variations of
x(t ) with frequencies of f1 ≤ f ≤ f1 + df
∞
∴ ∫ S x (f )df = σ x2
0

The Relationship between the PDF and the SDF

PDF – Probability Density Function, p x (x )

SDF – Spectral Density Function, S x (f )

If x(t) is Gaussian, then

( x − x )2
−
1 2σ x2
p( x ) = e (Eq. 8.11)
2π σ x
where

∞
σ x2 = ∫ S x (f )df (Eq. 8.18)
0

When: y (t ) = a x (t )

S y ( f ) = a 2 S x (f ) and σ y2 = a 2σ x2

As an example, with seismic excitation where the force is p(t ) = −my&&g (t ) ,

then S p (f ) = m 2 S y&&g (f ) .

If x(t) is a time-varying quantity:

p x (x ) provides information on the amplitude content of x(t)

S x (f ) provides information on the frequency content of x(t)
R x (τ ) provides information on the correlation of x with itself at time lag, τ

Page 4 of 6
CEE490b Feb. 11, 2002

Description of Periodic Functions

Fourier Series – Provides information on the variation in the frequency domain at

discrete harmonics of the frequency, ω , 2 ω , 3 ω , ……

As T → ∞ this results in a Fourier Integral, providing information at all values of ω :

∞
1
∫ A(ω )e
iω t
x (t ) = dω
2π −∞
∞
a Fourier Transform Pair
∫ x(t )e
− iω t
A(ω ) = dt
−∞

If P (t ) = Po e iωt
Po
y (t ) = H (ω )e iωt , the effect of the load at one harmonic of ω
K

then if there are many harmonic loads:

∞
1
y (t ) = ∑ c r H (f r )e i 2πf1t
−∞ K

∞
1
∑c
2 2
y2 = r H (f r )
K2 −∞

∞ ∞
If T → ∞ , 1/ T → df , ∑ → ∫,
0
rf1 = f r → f and H (f r ) → H (f )
0

∞ ∞
1
S (f ) H (f ) df
2
σ y2 = y 2 = ∫ S y (f )df = 2 ∫ p
0 k 0

See Fig. 8.12 for the relationship between S y (f ) and S p (f ) .

Page 5 of 6
CEE490b Feb. 11, 2002

Fig. 8.12 The relationship between spectrum of input and spectrum of output

Closed form solutions of this integral are difficult, so approximations can be made
with minimal loss of accuracy.

f ∞
1 o 1
σ = y ≅ 2 ∫ S p (f )df + 2 ∫ S p (fo ) H (f ) df
2 2 2
y
k o k 0
f
1 o 1 πf
≅ 2 ∫ S p (f )df + 2 S p (fo ) o
k 0 k 4D

2
σ p2  π f o S p (f o ) 
y ≅ 2 1 + 
k  4D σ p2 

Page 6 of 6