Sie sind auf Seite 1von 34

INSE 6310:

Systems Engineering Maintenance


Management
Concordia Institute of Information System Engineering (CIISE),
Concordia University, QC, Canada

Winter 2016
Lecture 2
Arash Mohammadi
Office: S-EV 9187
Email: arashmoh@encs.Concordia.ac
Homepage: http://users.encs.concordia.ca/~arashmoh/

1/15/2016

HO-2

1/ 34

Outline of this Class


Objectives:
Probability and statistics basics.
Reliability concepts.
Life-time distributions.

References:
Check the following materials on the course website
Hand out #2 (HO-2)
Appendix 1 of the book.

1/15/2016

HO-2

2 / 34

Probability and Statistics Basics

1/15/2016

HO-2

3 / 34

Classical Set Theory


1. Population, Sample-Space, or Universal Set
All the possible outcomes of a random experiment.
A set that contains all elements of interest, typically denoted by S
Discrete Sample-Space
(e.g., Rolling a dice:
Continuous Sample-Space
(e.g., Failure time of an equipment:
Arrival Time:

2. Events (subset)
Any subset of a Sample-Space is known as an event E.
E occurs when the outcome of the random experiment is contained in E.

3. Venn Diagram
Graphical representation of sets within the frame of reference of Sample-Space.
Convenient to describe various set operations.

1/15/2016

HO-2

4 / 34

Operation of Events (Sets)


Intersection of Events

Set Operations

Union of Events

Complement of an Event

Mutually Exclusive Events

1/15/2016

HO-2

5 / 34

Concepts of Probability Theory


Interpretations of Probability
1.

Classical: The probability of an event E is equal to the number of outcomes composing


that event (n) divided by the total number of possible outcomes (N).

2.

Relative Frequency: Probability of an event is defined in terms of the proportion of


times the event occurs in a long series of identical trials

4
7
4
8
12

3.

10
10
11
5
5

6
10
3
4
6

7
4
2
8
3

5
4
8
7
7

10
7
6
4
9

4
8
10
11
10

6
8
9
9
11

5
7
3
5
3

6
7
4
5
2

Subjective: The probability of a particular person (to be used when there are no other
ways to obtain probability measures)

1/15/2016

HQ-2

6 / 22

Axioms of Probability

Example
There is a deck of 54 card including jokers.
1. What is the probability that the first card drawn is a diamond? 2. A Jack? 3. A
diamond Jack? 4. Either a diamond or a Jack?

1/15/2016

HO-2

7 / 34

Example
Classical View (Equally Likely assumption)
S: All possible 54 cards
A: The first card is a Diamond
B: The first card is a Jack

1/15/2016

HO-2

8 / 34

Concepts of Probability Theory


Conditional Probability
Calculate probabilities when some partial/ additional information regarding the
result of the experiment becomes available.
If the event B occurs then in order for A to occur, it is necessary that the actual
occurrence be a point in both A and B.

Multiplication Rule:

Independent Events:

1/15/2016

HO-2

9 / 34

Concepts of Probability Theory


Law of Total Probability
Partition: A partition of the sample space S is a collection of n disjoint
events
where the union is S, i.e.,

A partition divides any set A into disjoint pieces as follows

As the partition consists of disjoint sets, we have

Applying multiplication rule we have:

1/15/2016

HO-2

10 / 34

Concepts of Probability Theory


Bayes Theorem
Conditional probability coupled with the Law of Total Probability.
Assume we have a partition

and we know probabilities

Prior Probability:
Likelihood Function:
Posterior Probability:
1/15/2016

HO-2

11 / 34

Random Variable

Particular outcome of the a random experiment is not known in advance,


therefore, we use a Random Variable (r.v.) to associate a number with the
outcome of a random experiment.

A RV is a real-valued function defined over the elements of the Sample-Space.

A RV associate a number to each outcome in the Sample-Space.

We denote a RV by an uppercase letter X and the numerical value that it can


take by a lowercase letter x.
Discrete RV: A random variable is said to be discrete if its sample
space is countable.
Continuous RV: A random variable defined over a continuous
Sample-Space.

1/15/2016

HO-2

12 / 34

Discrete Random Variable

Let X be a discrete r.v. defined over a sample space


Probability Mass Function (PMF):
where

The cumulative distribution function (CDF):


where

Expected value and Variance:

1/15/2016

HQ-2

13 / 34

Discrete Random Variable

Example: In five flips of a coin, we define the following r.v.

f(x)

F(x)

X: The number of heads obtained.

X (number of heads)

X (number of heads)

1/15/2016

HQ-2

14 / 34

Continuous Random Variable

Let X be a Continuous r.v.,

Probability Density Function (PDF):

The cumulative distribution function (CDF):


32.8

148.7

53.9

53.4

28.2

45.6

106.5

65.3

188.0

37.6

15.1

18.2

85.8

101.4

116.5

96.5

110.6

55.1

104.9

22.2

22.4

82.3

44.6

5.1

9.2

48.2

130.8

20.6

34.6

57.4

77.9

38.1

44.9

26.7

67.5

129.2

44.2

94.3

211.3

181.2

46.7

89.1

93.9

46.9

22.3

52.3

74.2

35.4

58.7

19.6

28.2

144.7

32.7

8.9

91.5

70.8

35.6

64.0

26.3

161.6

83.5

30.5

97.3

21.1

88.6

48.8

23.6

38.6

24.1

71.7

124.8

120.1

147.1

72.0

31.2

28.8

178.0

16.6

20.2

21.5

30.1

82.2

60.3

102.7

82.1

65.4

89.8

41.9

57.0

44.8

68.9

103.0

41.9

78.1

38.6

35.0

75.7

28.5

53.9

49.4

Failure data
1/15/2016

HQ-2

15 / 34

Continuous Random Variable

1/15/2016

HQ-2

16 / 34

Reliability Concepts

1/15/2016

HQ-2

17 / 34

Life Data Analysis

Life data could be the time the product operated successfully or the time the
product operated before it failed.

In life data analysis (Weibull analysis), we attempt to make predictions about the
life of all products in the population by fitting a statistical distribution (model) to
life data.

Life Data Analysis Requires The Practitioner to:


Gather life data for the product.
Select a lifetime distribution that will fit the data and model the life of the
product.
Estimate the parameters that will fit the distribution to the data.
Generate plots and results that estimate the life characteristics of the product.
Important Life Characteristics:
Reliability or probability of failure at a specific time
Mean Life or Mean Time to Failure (MTTF)
Failure rate.

1/15/2016

HQ-2

18 / 34

Time to Failure
Failure: Termination of the systems ability to perform its required function
under the given condition for specific period of time.
T: Time to failure which is a random variable
The probability of a failure
occurring in (ti-1,ti)

History of the failure times.


Relative frequency histograms.

1/15/2016

HQ-2

19 / 34

Lifetime Distribution (Life Data Model)


Time to failure (T) is a continues random variable.
We use probability density functions (PDFs) to characterize time to
failures.
The probability of a failure
occurring in (ti-1,ti)

1/15/2016

HQ-2

20 / 34

TIME to FAILURE
We characterize the time to failure with the following four different functions:
T: Time to failure which is a random variable
Probability Density Function (PDF):
Cumulative Distribution Function (CDF):

Reliability (Survival) function:


Hazard function

Mean Time to Failure:

1/15/2016

HQ-2

21 / 34

Reliability and Hazard Function

Hazard function h(t) and


Reliability R(t) are related to
each other.
Hazard function is enough to
uniquely describe the failure
distribution of the system.

1/15/2016

HQ-2

22 / 34

Bathtub Curve

1/15/2016

HQ-2

23 / 34

Effects of Overhaul on h(x)


An example of the interesting questions we want to address:
What is the optimal time (tp) for performing preventive replacement to
minimize the expected cost per unit time?

1/15/2016

HQ-2

24 / 34

Life data Distributions

1/15/2016

HQ-2

25 / 34

Component lifetime distributions


The four well-known probability density functions in practice are as follows:
Exponential Distribution
Normal Distributions (Gaussian Distribution)
Hyper-exponential distribution
Weibull distribution

1/15/2016

HQ-2

26 / 34

Exponential Distribution
Failure is due to random causes such as sudden excessive loading
Electronic components
Complex industrial plants
PDF:
CDF:
Reliability:
Mean time to failure:

Hazard Rate (Failure Rate Function):

1/15/2016

HQ-2

27 / 34

Exponential Distribution
Hazard function h(x) is constant
over time.
Equipment deterioration is
independent of the age of the
system.
The CDF

Hazard Function

Reliability
1/15/2016

HQ-2

28 / 34

Exponential Distribution
Exponential distribution is memoryless:
Consider T an exponentially distributed random variable with > 0.
For any two real numbers a, b > 0 , we have

Exponential distribution has the memoryless property which means


that the chance of survival in the next interval is independent of the
current age of the system.

1/15/2016

HQ-2

29 / 34

Normal Distribution
Failure is due to the additive effect of a large number of small and
independent random variations
Light bulbs
Bus engines

Mean time to failure:


Variance of time to failure: 2

1/15/2016

HQ-2

30 / 34

Normal Distribution
Hazard function h(x) is
increasing in time.
Equipment deteriorates over
time and as it ages.
The CDF

Hazard Function

Reliability
1/15/2016

HQ-2

31 / 34

Normal Distribution Properties


Normal Table: Areas in the upper
tail of the standard normal distribution.

If X is a random variable from normal


distribution with mean and standard
deviation :

1/15/2016

HQ-2

32 / 34

Normal Distribution
Linear Combination of two Normal R.V.

Addition of n i.i.d. normal r.v.s

Parameter Estimation:

1/15/2016

HQ-2

33 / 34

Whats Next?

Topics of the Next

1/15/2016

Hyper-Exponential
Weibull Distributions
Parameter Estimation
Different Type of Data

HQ-2

34 / 34

Das könnte Ihnen auch gefallen