ELE5TEL 1

Telecommunications Engineering

Dr. David Tay

Room BG434
x 2529
[email protected]

DT
ELE5TEL 156

System Reliability
• Virtually all telecommunication systems will fail at some
point in time and needs to be repaired / serviced or re-
placed.

• Important to have an objective measure, i.e. metric, of

the reliability of the system.

• Requires concepts from probability and statistics.

Reliability Function

• The state variable X(t) describe the operational state of

a system or an item:

X(t) = 1 if system/item is functional or working

X(t) = 0 if system/item has failed or not working

• TTF: Time-To-Failure which is a random variable de-

noted by T . The probability density function is denoted
by f (u).

• The CDF (Cumulative-Density-Function) is given by

Z t
F (t) ≡ Pr(T ≤ t) = f (u)du
0

DT
ELE5TEL 157

• The reliability of a system/item at time t is the proba-

bility the system/item does not fail, i.e. still working, at
time t. The reliability or survivor function R(t) is
then given by
Z t
R(t) = 1 − F (t) = 1 − f (u)du
0

Example with Gaussian distribution

DT
ELE5TEL 158

Availability

• A system/item is available if it is ready or present for

immediate use - important in contracts.

• Measures of availability can be specified instantaneously

or as an average over a period of time, eg. life cycle of
system/item.

• Instantaneous availability:

A(t) = Pr(X(t) = 1)

• Average availability:
t2
1
Z
Aaverage = A(t)dt
t2 − t1 t1

• Aaverage gives the percentage the system/item is working

with respect to the total time the system/item has been
in operation.

• Aaverage can be a crude / rough measure that does not

tell the whole story.

DT
ELE5TEL 159

• Examples with the same Aaverage

The time in takes to repair failed system is an important

factor.

DT
ELE5TEL 160

• MTBF (Mean-Time-Between-Failure): the average amount

time a system/item is functioning.

• MTTR (Mean-Time-To-Repair): the average amount of

time it takes to repair/replace a system/item. The MTTR
gives an indication of the maintainability of a sys-
tem/item which is defined as how quickly a system can
be repaired.

• Average availability can also be calculated as:

MTBF
Aaverage =
MTBF + MTTR

• Instantaneous unavailability is defined as the

U (t) = Pr(X(t) = 0)

• Average unavailability:

MTTR
Uaverage = 1 − Aaverage =
MTBF + MTTR

• Important to note that MTBF and MTTR are average

values and the actual failure / repair time can be quite
different if the variance in the probability distribution is
large.

DT
ELE5TEL 161

MTTF

MTTF: Mean-Time-To-Failure is defined as mean (or aver-

age) value of TTF of a system/item:
Z ∞ Z ∞
MTTF = E[TTF] = tf (t)dt = R(t)dt
0 0

where the second equality follows by applying the relation-

ship f (t) = −R′ (t) and using integration by parts.

MTTF and MTBF is usually used interchangebly to mean

the same thing. MTBF is used in the system/item that can
be repaired.
Example: A system has a life-cycle of 2 years during which
it failed 3 times. The failure happened after 60 days, 200
days, and 450 days of service. The time it took to repair
the failures were 7 days, 3 days and 5 days. Determine (i)
MTBF, (ii) MTTR and (iii) Aaverage .
Solution:
(i) TBF1 = 60 days, TBF2 = 200 - 60 - 7 = 133 days, TBF3
= 450 - 200 - 3 = 247 days, TBF4 = 730 - 450 - 5 = 275
days.

60 + 133 + 247 + 275

MTBF = = 179 days
4

DT
ELE5TEL 162

(ii)
7+3+5
MTTR = = 5 days
3
(iii)
179
Aaverage = = 0.973
179 + 5

Failure Rate

The failure rate z(t) is given by

f (t)
z(t) =
R(t)

which is the conditional probability per unit time

Pr(t < T ≤ t + ∆t | T > t)/∆t

The probability is for the case the system/item fails within

the interval ∆t given that it did not fail before time t.
The failure rate z(t) gives the failures per unit time (usually
hours) as a function of time and is generally not constant.
The common failure rate follows a bathtub curve

DT
ELE5TEL 163

High failure rates at the beginning and end of the life cycle.

Exponential Distribution
• Used extensively to model reliability in telecommunica-
tion systems.

• Fairly accurate representation of electronic component

DT
ELE5TEL 164

TTF.

• Has memory-less feature: after a system/item has been

repaired, it does not remember the previous TTF, ie.
the next TTF is independent of the previous TTF. The
system/item behaves like it is ”as good as new”.

• PDF is given by

f (x) = λe−λx for x ≥ 0

and f (x) = 0 for x < 0. The MTTF is given by:

Z ∞
1
MTTF = E[X] = x f (x) dx =
0 λ

DT
ELE5TEL 165

Example of PDF

• CDF is given by
Z x
F (x) = f (u)du = 1 − e−λx for x ≥ 0
0

DT
ELE5TEL 166

Example of CDF

• Reliability

R(t) = 1 − F (t) = e−λt for t ≥ 0

DT
ELE5TEL 167

• Failure rate
f (t)
z(t) = =λ
R(t)

Example: A component has a MTTF of 100 hours can be

modelled with an exponential distribution. If it is required
to have a reliability of at least 50%, what is the maximum
recommended operating time?
Solution: λ = 1/100 hr−1

R(t) = e−λtmax = 0.5

1
tmax = − loge 0.5 = −100 loge 0.5 = 69 hours
λ

Weibull Distribution

• Extremely flexible distribution in the field of reliability

engineering.

• Has a shape parameter for accurate modelling for TTF

or TTR distribution.

• PDF given by:

α α−1 −(λt)α
f (t) = αλ t e for t ≥ 0

• The parameter α > 0 is called the shape parameter.

DT
ELE5TEL 168

• The parameter λ > 0 is called the scale parameter.

• Mean (not so easy to calculate) given by

Z ∞ Z ∞
αλα tα e−(λt) dt
α
E[T ] = tf (t)dt =
0 0
 
1 1
E[T ] = Γ 1 +
λ α
where Γ is the Gamma function given by
Z ∞
Γ(t) = xt−1 e−x dx
0

The Gamma function is the generalization of the facto-

rial. When t is a positive integer Γ(t) = t.(t − 1). . . . 2.1.

• Careful selection of α and λ either from expert knowledge

or empirical data allows accurate modelling for TTF or
TTR.

• CDF
α
F (t) = 1 − e−(λt) for t ≥ 0

• Reliability function
α
R(t) = e−(λt) for t ≥ 0

• Failure rate
f (t)
z(t) = = αλα tα−1
R(t)
which is not a constant unlike in exponential distribution.

DT
ELE5TEL 169

• When α = 1, the Weibull distribution becomes the spe-

cial case of exponential distribution.

• Example for modelling TTR

System Modelling
• A system is made of on the interconnection of compo-
nents that interact which each other.

DT
ELE5TEL 170

• The reliability of the system is dependent on:

1. Reliability of the individual components.

2. How the components interact with each other.

• Suppose the system is composed on n components, each

with state variable Xi (t) (i = 1, . . . , n)

Xi (t) = 1 if component i is working

Xi (t) = 0 if component i has failed

• The reliability of component i is given by:

Ri (t) = E[Xi (t)] = 0 × Pr(X(t) = 0) + 1 × Pr(X(t) = 1)

which is the probability component i survives for a time

t.

• A RBD (Reliability Block Diagram) is used to show the

interconnection between components:

DT
ELE5TEL 171

DT
ELE5TEL 172

Structure Function

The structure function of a system is a function of the state

variables of the components

S(X) = S(X1 , X2 , . . . , Xn )

where the state vector of the system is

X = [X1 , X2 , . . . , Xn ]

The structure function gives the state of the system

S(X) = 1 if system working

S(X) = 0 if system has failed

Series Structure

• System will work only if all components are working.

• Structure function is given by

n
Y
S(X(t)) = X1 × X2 × . . . × Xn = Xi (t)
i=1

DT
ELE5TEL 173

• Reliability is given by
" n # n n
Y Y Y
R(S(t)) = E Xi (t) = E[Xi (t)] = Ri (t)
i=1 i=1 i=1

• The weakest link, ie. min (Ri (t)), is the most important
in determining the system reliability.

R(S(t)) ≤ min {Ri (t)}

DT
ELE5TEL 174

Example: single-thread satellite link RF chain.

The MTBF for the components are

1. Frequency converter (FC) MTBF = 95,000 hours.

2. Digital modem (DM) MTBF = 120,000 hours.

3. High-Power amplifier (HPA) MTBF = 75,000 hours.

Want to calculate system survives after 6 months of oper-

ation t = 365 × 24/2 = 4380 hours assuming exponential
distribution.
Solution: λ = 1/MTBF and R(t) = e−λt .

RF C = e−4380/95000 = 0.955

DT
ELE5TEL 175

RDM = e−4380/120000 = 0.964

RHP A = e−4380/75000 = 0.943

n
Y
R(S(t)) = Ri (t) = RRC × RDM × RHP A = 0.868
i=1

Lower reliability than individual components reliability.

Parallel Structure

• System will work if any components are working.

DT
ELE5TEL 176

• Structure function is given by

S(X(t)) = 1 − (1 − X1 ) × (1 − X2 ) × . . . × (1 − Xn )
n
Y
=1− (1 − Xi (t))
i=1

• Reliability is given by
n
" #
Y
R(S(t)) = E 1 − (1 − Xi (t)) =
i=1

n
Y n
Y
1− (1 − E[Xi (t)]) = 1 − (1 − Ri (t))
i=1 i=1

• There is n degree of redundancy in the system, ie. for

the system to fail all components must fail.

DT
ELE5TEL 177

Example: parallel satellite link RF chain.

The MTBF for the components are

1. Frequency converter (FC) MTBF = 95,000 hours.

2. Digital modem (DM) MTBF = 120,000 hours.

DT
ELE5TEL 178

3. High-Power amplifiers (HPA1 and HPA2) MTBF = 75,000

hours.

Want to calculate system survives after 6 months of oper-

ation t = 365 × 24/2 = 4380 hours assuming exponential
distribution.
Solution: Reliability of High-Power amplifier system (HPAS)
is
RHP AS = 1 − (1 − RHP A1 ) × (1 − RHP A2 )

= 1 − (1 − 0.943)2 = 0.997

Total system reliability

n
Y
R(S(t)) = Ri (t) = RF C × RDM × RHP AS
i=1

= 0.955 × 0.964 × 0.997 = 0.918

which is an improvement of the system without redundancy

(0.868).

DT