Scribd
Upload
51 views13 pages

Monte Carlo Simulation Technique: Random Number Generators

This document discusses Monte Carlo simulation techniques. It describes: 1) How Monte Carlo simulation uses random numbers to estimate the probability of failure for performance functions. 2) Common random number generators like linear congruential generators that are used to generate pseudo-random numbers needed for Monte Carlo simulations. 3) Methods for transforming uniformly distributed random numbers into other distributions like normal and lognormal using inverse transform and Box-Muller techniques.

Uploaded by

Saurabh Tripathi
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
51 views13 pages

Monte Carlo Simulation Technique: Random Number Generators

This document discusses Monte Carlo simulation techniques. It describes: 1) How Monte Carlo simulation uses random numbers to estimate the probability of failure for performance functions. 2) Common random number generators like linear congruential generators that are used to generate pseudo-random numbers needed for Monte Carlo simulations. 3) Methods for transforming uniformly distributed random numbers into other distributions like normal and lognormal using inverse transform and Box-Muller techniques.

Uploaded by

Saurabh Tripathi
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 13

26-10-2020

Monte Carlo Simulation Technique


• This technique is used to estimate the
probability of failure of a performance
function.
• A simple and straight forward approach
• Based on generation of random numbers
• ‘N’ Random numbers are generated for a
variable based on its probabilistic
characteristics (mean, standard deviation,
and distribution)

Random number generators


• Random number vs. pseudo-random
number
• U(0,1) – Uniformly distributed continuous
RV between 0 and 1.
• Uniform random-number generators
• Transformation technique to generate
random numbers following other
distributions

1
26-10-2020

Uniform random number


generators
• Arithmetic generators are commonly used
– Employ sequential methods
• Each number is determined by one or several of its
predecessors according to a fixed mathematical
formula
– Carefully designed generators can produce
number that appear to be independent
random variates.
• Sequential numbers are not truly random, being
derived from previous numbers in some
deterministic fashion – Pseudo-random numbers

Uniform random number


generators
• Properties of a good arithmetic uniform
random number generator:
– numbers generated should be appear to be
independent and uniformly distributed
– should be fast and not require large amounts
of storage
– should have the ability to reproduce a given
stream of random numbers exactly.
– should have a very long period

2
26-10-2020

Reproducing the random


numbers
• Useful when attempting to compare the
responses of two different designs

Arithmetic generators
• Linear congruential generators (LCGs)
– Introduced by Lehmer (1951)
– A sequence of integers Z1, Z2, … are defined
by the recursive formula:
• Zi=(aZi-1+c)(mod m)
• Where m is the modulus, a is a multiplier, c is an
increment, and all three parameters are positive
integers.
• =(aZi-1+c)(mod m) means the whole remainder of
(aZi-1+c) after dividing it by m
• 15 mod 4=3 & 17 mod 5=2
• Zi is an integer and varies between 0 and m-1.

3
26-10-2020

Linear Congruential Generators


• The sequence starts with Z0 is a positive
integer, a seed
• Ui=Zi/m
• Ui can have 0, 1/m, 2/m, …, (m-1)/m
• m should be a large number, to generate
continuous uniform distributed random
numbers
• Z1 is obtained from Z0, Z2 from Z1, …

Linear Congruential Generators


• For fixed values of multiplier, a, increment,
c, and modulus, m, the same sequence of
Zi values is produced for same seed value,
Z0.
• So long as the seed value is known, the
recursive formula produces a given stream
of pseudo-random numbers exactly.
• Sequence of Ui is independent and
uniformly distributed if a,c, and m are
correctly selected.

4
26-10-2020

Periodicity
• If Z0=3 produces Z1=746, then whenever
Zi-1=3, then Zi=746.
– This property is called periodicity
– {83,83,83,…} – periodicity=1
– {94,4832,325,94,4832,325,..} – Periodicity=3
– Undesirable and common random-number
generators suffer from this.
• Generate random numbers with the
following values: Z0=21, a=25, c=55, and
m=96.

Periodicity
• Generate random numbers with the
following values: Z0=21, a=25, c=55, and
m=96.
– 21,4,59,90,1, …
– Uniformly distributed random numbers
between 0 &1 [0,1):
4/96=0.042, 59/96=0.614, 90/96=0.9375,
• Maximum periodicity of LCG is m, and will
occur only if a,c, and m are selected
carefully. This is called full periodicity.

5
26-10-2020

Periodicity of a generator
• A generator has full period if its period is
m.
• A generator which is full period will
produce exactly one of each possible
values, {0,1, …., m-1), in each cycle.
• If the generator is good, all of these
possible values will appear to occur in
random order.

Selection of a,c & m


• LCG has full period if the following three
conditions are fulfilled.
– The only positive integer that exactly divides
both m and c is 1
– If q is a prime number (divisible only by itself
and 1) that exactly divides m, then q exactly
divides (a-1). This must be true for all prime
factors of m; example: m=96, a=25.
– If 4 exactly divides m, then 4 exactly divides
(a-1)

6
26-10-2020

Example
• Is Zi=(25Zi-1+55) (mod 96) is a full period
generator?

Minimal standard generator


• Minimal standard (MS) generator with
constants:
A=75, c=0 & M=231-1, has a periodicity of m-1 or
about 2 x 109, with requirement that the seed 0
must never be used.
Once Zi=0 – generator will return zeros.
This form of the LCG with c=0 is called a
multiplicative LCG.
Zi+1=aZi (mod m)
Suffer from serial correlation, some correlation
between successive values.

7
26-10-2020

Inverse transform method


• Generate u ~ U(0,1)
• Return x=F-1(u)

Example
• Generate 10 uniformly distributed random
numbers between 20 and 50.

8
26-10-2020

Monte Carlo Simulation


• If X is N(m,s), then random numbers for X are
generated as:

Where F-1 is the inverse of the CDF of a standard normal variable

Monte Carlo Simulation


• If the random varibale, X, is lognormal
distributed with parameters lX and zX, then
random numbers for X are generated as:

9
26-10-2020

Inverse Transformation
FU(u) FX(x)

1.0 1.0

0.5 0.5

0.0 0.0
U fU(u) fX(x) X

U 1 ui 0 xi X

Example – Normal RV
• Generate 2 random variates for bearing
capacity following normal distribution with
mean 500 kPa and standard deviation 5
kPa, using inverse transformation
technique.

10
26-10-2020

Example – Lognormal RV
• Generate 2 random variates for bearing
capacity following lognormal distribution
with mean 500 kPa and standard deviation
5 kPa, using inverse transformation
technique.

Radial transformation method


• Box and Muller (1958)
– Useful to generate random numbers following
distributions which have no simple closed-
form solutions, such as normal.
– Exact
– Simple to use

11
26-10-2020

Box-Muller (1958)
• If X is a normally distributed with mean mX
and standard deviation sX, then
realizations of X can be generated as:
1. Generate u1~U(0,1) and u2 ~U(0,1)
2. Form g1=sqrt(-2lnu1) cos(2pu2) and
g2= sqrt(-2lnu1) sin(2pu2)
3. Form x1=mX+sXg1 and x2= mX+sXg2
This method generates realizations of X1
and X2, which are independent normal
random variates.

Example
• Generate two random variates following
normal distribution with mean 12 and
standard deviation 4, using the following
two uniformly distributed random numbers
between 0 and 1. Use Box-Muller method.
• u1=0.89362
• u2=0.42681

12
26-10-2020

Example
To estimate the settlement due to consolidation of a
homogeneous saturated soil deposit, the information on the
coefficient of consolidation, cv is a very important. It can be
calculated as:
𝑘
𝑐𝑣 =
𝑚𝑣 𝛾𝑤
Where, k is the hydraulic conductivity, mv is the coefficient
of volume compressibility, and gw is the unit weight of water.
Assume k and mv are statistically independent lognormal
random variables with means of 1.3 x 10-7 m/min and
0.0011 m2/kN, respectively. Both have coefficients of
variation of 10%. Using the above data, estimate P(cv>1.2 x
10-5 m2/min).

13

You might also like