09/07/2015

S. N. Njoroge 1 S. N. Njoroge 2

Multiple Failure Mechanisms

• It’s seldom true that an item's failure is caused by a single failure
mechanism.
• Mostly, there will be more than one mechanism that causes the
failure of an item
• Most of the practical systems fail due to different failure
STATISTICAL RELIABILITY AND mechanisms
ESTIMATION FOR COMPONENTS • Let 𝑓1 (𝑡) and 𝑓2 (𝑡) be the PDF of the system due to failure
mechanism 1 and 2 respectively.
SELECTION
• The PDF of the TTF of the system caused by either of the failure
mechanisms is
S. N. Njoroge
𝑓(𝑡) = 𝑓1 (𝑡),1 − 𝐹2 (𝑡)- + 𝑓2(𝑡),1 − 𝐹1 (𝑡)-

S. N. Njoroge 3 S. N. Njoroge 4

Multiple Failure Mechanisms Multiple Failure Mechanisms - Ex

• Where 𝐹1 (𝑡) and 𝐹2 (𝑡) the are failure function for failure • Failure of an item is caused by two different failure mechanisms
mechanism 1 and 2 respectively. A and B. The time-to-failure distribution of the item due to
• The Failure Function of the item under two different failure failure mechanism A can be represented by exponential
mechanism is given by: distribution with parameter λ𝐴 = 0.002 hours. The time-to-
𝑡 failure distribution of the item due to failure mechanism B can be
𝐹(𝑡) = 𝑓1(𝑡),1 − 𝐹2 (𝑡)- + 𝑓2(𝑡),1 − 𝐹1 (𝑡)- 𝑑𝑥 represented by exponential distribution with parameter
0
λ𝐵 = 0.005 hours. Find the probability that the item will fail
before 500 hours of operation.

S. N. Njoroge 5 S. N. Njoroge 6

Multiple Failure Mechanisms - Ex Failure Rate

𝑓𝐴 𝑡 = λ𝐴 𝑒 −λ𝐴 𝑡 ; 1 − 𝐹𝐴 𝑡 = 𝑒 −λ𝐴 𝑡 • Whenever the hazard function is constant, we call it the Failure
𝑓𝐵 𝑡 = λ𝐵 𝑒 −λ𝐵 𝑡 ; 1 − 𝐹𝐵 𝑡 = 𝑒 −λ𝐵 𝑡 Rate.
𝑡
• Failure Rate is a special case of hazard function (which is the
𝐹 𝑡 = 𝑓1 𝑡 1 − 𝐹2 𝑡 + 𝑓2 𝑡 1 − 𝐹1 𝑡 𝑑𝑥
0 time dependent failure rate).
= λ𝐴 𝑒 −(λ𝐴 +λ𝐵 )𝑡 + λ𝐵 𝑒 −(λ𝐴 +λ𝐵 )𝑡 • Failure Rate is one of the most widely used contractual
𝐹 𝑡 = 1 − 𝑒 −(λ𝐴 +λ𝐵 )𝑡 reliability measures
𝐹 500 = 1 − 𝑒 − 0.002+0.005 500 = 0.9698 • By definition, it is appropriate to use Failure Rate only when the
time-to-failure distribution is exponential.
• H/W – Find the reliability due to single failure mechanism • Also, Failure Rate can be used only for a non-repairable system
and combined reliability due to multiple failure mechanisms
and compare

1
 09/07/2015

S. N. Njoroge 7 S. N. Njoroge 8

Failure Rate Mean Time To Failure (MTTF)

𝑇𝑜𝑡𝑎𝑙 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑓𝑎𝑖𝑙𝑢𝑟𝑒𝑠 𝑖𝑛 𝑎 𝑠𝑎𝑚𝑝𝑙𝑒 • Expected value of a system's time to first failure.
𝐹𝑎𝑖𝑙𝑢𝑟𝑒 𝑅𝑎𝑡𝑒 =
𝐶𝑢𝑚𝑢𝑙𝑎𝑡𝑖𝑣𝑒 𝑜𝑝𝑒𝑟𝑎𝑡𝑖𝑛𝑔 𝑡𝑖𝑚𝑒 𝑜𝑓 𝑡𝑕𝑒 𝑠𝑎𝑚𝑝𝑙𝑒 • Used as a measure of reliability for non-repairable items such as
bulbs, microchips and many electronic circuits.
• N.B • Mathematically,
∞ ∞
• Failure Rate represents the number of failures per unit time. 𝑀𝑇𝑇𝐹 = 𝑡 × 𝑓 𝑡 𝑑𝑡 = 𝑅 𝑡 𝑑𝑡
• If the failure rate is 𝜆, then the expected number of items that 0 0
fail in the interval 0 ≤ 𝑥 ≤ 𝑡is 𝜆𝑡.
• MTTF can be considered as the area under the curve
represented by the reliability function, 𝑅 𝑡 between 0 and ∞
• If the item under consideration is repairable, then the expression
represents Mean Time To First Failure of the item

S. N. Njoroge 9 S. N. Njoroge 10

Mean Time To Failure (MTTF) Mean Time To Failure (MTTF)

• MTTF varies for different failure distributions
• For example let us assume that we have two items A and B with
same MTTF (e.g. 500 days). If the TTF of item A is exponential
is that of item B is normal then there will be a significant
variation in the behaviour of these items.
• The figure in slide 11shows the CDF of these two items up to
500 days.
• MTTF is the average life of a non-repairable system.
• For a repairable system, MTTF represents the average time
before the first failure.

S. N. Njoroge 11 S. N. Njoroge 12

Mean Time To Failure (MTTF) Mean Residual Life

• It may be of interest to know the expected value of the remaining
life of the item before it fails from an arbitrary time 𝑡0 e.g. when
disposing off big equipment
• We denote this value as 𝑀𝑇𝑇𝐹(𝑡0 ),
• It represents the expected time to failure of an item aged 𝑡0
• . Mathematically,
∞
𝑀𝑇𝑇𝐹 𝑡0 = 𝑡 − 𝑡0 𝑓(𝑡|𝑡0 ) 𝑑𝑡
𝑡0
Where
𝑓 𝑡 𝑡0 = 𝑕(𝑡) × 𝑅(𝑡|𝑡0 )
the density of the conditional probability of failure at time 𝑡,
provided that the item has survived over time 𝑡0

2
 09/07/2015

S. N. Njoroge 13 S. N. Njoroge 14

Mean Residual Life - Example MTTF of a Maintained System

• The time to failure of an airborne navigation radar can be • If an item is subject to preventive maintenance after every 𝑇𝑝𝑚
represented using Weibull distribution with scale parameter units, that is, at 𝑇𝑝𝑚 , 2𝑇𝑝𝑚 , 3𝑇𝑝𝑚 , etc. The expected time to
𝜂 = 2000 hours and 𝛽 = 2.1. It was told that the age of the failure, 𝑀𝑇𝑇𝐹𝑝𝑚 , of the item is given by:
∞
existing radar is 800 hours. Find the expected value of the
𝑀𝑇𝑇𝐹𝑝𝑚 = 𝑅𝑝𝑚 𝑡 𝑑𝑡
remaining life for this radar. 0
𝑇𝑝𝑚 2𝑇𝑝𝑚 3𝑇𝑝𝑚
• (Ans: 1165.72 hours.)
= 𝑅𝑝𝑚 𝑡 𝑑𝑡 + 𝑅𝑝𝑚 𝑡 𝑑𝑡 + 𝑅𝑝𝑚 𝑡 𝑑𝑡 + ⋯
0 𝑇𝑝𝑚 2𝑇𝑝𝑚
• Try at home
• N.B. Use Gamma tables • 𝑅𝑝𝑚 𝑡 is the reliability of the item subject to preventive
maintenance
• N.B.:Preventive maintenance will not improve the reliability of
the system, if the time to failure is exponential

S. N. Njoroge 15 S. N. Njoroge 16

MTTF of a Maintained System MTTF of a Maintained System - Ex

• If the item is restored to ‘𝑎𝑠 − 𝑔𝑜𝑜𝑑 − 𝑎𝑠 − 𝑛𝑒𝑤’ state after • A solid state radar is subject to preventive maintenance after
each maintenance activity, then the reliability function between every 400 flight hours. The time to failure of the radar follows
any two maintenance tasks can be written as: exponential distribution with mean life 800 flight hours. Find
the 𝑀𝑇𝑇𝐹𝑝𝑚 of the radar.
𝑀𝑇𝑇𝐹𝑝𝑚 • Ans: 800hrs
𝑇𝑝𝑚 𝑇𝑝𝑚 𝑇𝑝𝑚
2
= 𝑅 𝑡 𝑑𝑡 + 𝑅(𝑇𝑝𝑚 )𝑅 𝑡 𝑑𝑡 + 𝑅 𝑇𝑝𝑚 𝑅 𝑡 𝑑𝑡
0 0 0
+⋯
𝑇𝑝𝑚
2
𝑀𝑇𝑇𝐹𝑝𝑚 = 1 + 𝑅 𝑇𝑝𝑚 + 𝑅 𝑇𝑝𝑚 +⋯ 𝑅 𝑡 𝑑𝑡
0
𝑇𝑝𝑚
0
𝑅 𝑡 𝑑𝑡
𝑀𝑇𝑇𝐹𝑝𝑚 =
𝐹(𝑇𝑝𝑚 )

S. N. Njoroge 17 S. N. Njoroge 18

MTTF of a Maintained System - Ex Read on Your Own

• A manufacturing company buys two machines A and B. The • Mean Operating Time Between Failures (MTBF)
time to failure of machine A can be represented by Weibull • Variance of Mean Time To Failure
distribution with 𝜂 = 1000 hours and 𝛽 = 2. The time to • MTBM – Mean Time Between Maintenance
failure of machine B can be represented by Weibull distribution
• MTBO – Mean Time Before Overhaul
with 𝜂 = 1000 hours and 𝛽 = 0.5. The maintenance manager
in charge of operation plan to apply preventive maintenance for • MTBCF – Mean Time Before Critical Failure
both the machines for every 200 hours, so that he can improve • MTBUR- Mean Time Before Unscheduled Removal
the expected time to failure of the machines. Check whether the
manager’s decision is correct.
• Ans: 𝐴 = 𝑐𝑜𝑟𝑟𝑒𝑐𝑡, 𝐵 = 𝑤𝑟𝑜𝑛𝑔

3
 09/07/2015

S. N. Njoroge 19 S. N. Njoroge 20

Reliability Estimation
Any reliability prediction methodology using time-to-failure
approach will involve the following steps:
1. Construct the reliability block diagram (RBD) of the system.
This may involve performing failure modes and effect
analysis (FMEA).
SYSTEM RELIABILITY 2. Determine the operational profile of each block in the
reliability block diagram.
CONFIGURATIONS 3. Derive the time-to-failure distribution of each block.
S. N. Njoroge
4. Derive the life exchange rate matrix (LERM) for the different
components within the system.
5. Compute reliability function of each block (item).
6. Compute the reliability function of the total system.

S. N. Njoroge 21 S. N. Njoroge 22

Reliability Block Diagram (RBD) Reliability Block Diagram (RBD)

• The RBD of a system is a logical diagrammatic illustration of • In an RBD the items whose failure can cause system failure
the system in which each item (hardware/software) within the irrespective of the remaining items of the system are connected
system is represented by a block. in series.
• RBD forms a basis for calculation of system reliability • Items whose failure alone cannot cause system failure are
measures. connected in parallel.
• Each block within a RBD can represent a component, • Depending on the item, an RBD can be represented in
subsystem, module or system. i. Series,
• The structure of a RBD is determined by the effect of failure of ii. Parallel,
each block on the functionality of the system as a whole. iii. Series-Parallel,
• A block does not have to represent physically connected iv. r-out-of-n or
hardware in the actual system to be connected in the block
v. Complex Configuration.
diagram

S. N. Njoroge 23 S. N. Njoroge 24

Reliability Block Diagram (RBD) Series Configuration

• Construction of RBD requires functional analysis of various • All the consisting items of the system MUST be available or
parts within the system. functional to maintain the required function of the system.
• Each block within a RBD should be described using time-to- • Failure of any one item of the system will cause failure of the
failure distribution for the purpose of calculating system system as whole.
reliability measures. • Series configuration is probably the most commonly
• The RBD can also have network structures (e.g. communication encountered RBD in engineering practice
systems, water network and Internet). E.g. series, parallel, ring,
star e.t.c

4
 09/07/2015

S. N. Njoroge 25 S. N. Njoroge 26

Reliability Function of Series Configuration Reliability Function of Series Configuration

• If 𝑅𝑆 (𝑡) represents the reliability function of a series system • N.B.:
with 𝑛 items and 𝑅𝑖 (𝑡) denotes the reliability function of the • It is assumed that the connecting media (such as solder joints)
item 𝑖, if 𝑇𝑇𝐹𝑖 is the time-to-failure random variable for the
item 𝑖, then the reliability function of system for 𝑡 hours of between different items is 100% reliable (unless this is
operation is given by specifically included in the RBD)
𝑅𝑆 𝑡 = 𝑃 𝑇𝑇𝐹1 ≥ 𝑡, 𝑇𝑇𝐹2 ≥ 𝑡, … . , 𝑇𝑇𝐹𝑛 ≥ 𝑡 • In most cases time (𝑡) actually represents age or utilization of
the item under consideration. It can have different units such as
𝑛 hours, miles, landings, cycles etc for different items.
𝑅𝑆 𝑡 = 𝑅1 𝑡 × 𝑅2 𝑡 × ⋯ × 𝑅𝑛 𝑡 = 𝑅𝑖 (𝑡) • One has to normalize the ‘time’ before calculating the reliability
𝑖=1 function in such cases.
• The system under consideration will maintain the required
• One method of normalizing the different life units of the items
function if and only if all the n items of the system are able to
is using Life Exchange Rate Matrix (LERM)
maintain the required function for at least 𝑡 hours of operation

S. N. Njoroge 27 S. N. Njoroge 28

Reliability Function of Series Configuration Characteristics of 𝑅𝑠 (𝑡) for a Series System

• When the life units of items are different (or different items • The value of the reliability function of the system, 𝑅𝑠 𝑡 ,for a
have different utilization), we use the following equation to find series configuration is less than or equal to the minimum value
the reliability of the series system of the individual reliability function of the constituting items
𝑅𝑆 𝑡 = 𝑃 𝑇𝑇𝐹1 ≥ 𝑡1 × 𝑇𝑇𝐹2 ≥ 𝑡2 × … .× 𝑇𝑇𝐹𝑛 ≥ 𝑡𝑛 • If represent the hazard function of item 𝑖, then the system
reliability of a series system can be written as:
𝑛 𝑛
𝑡 𝑡 𝑛
𝑅𝑆 𝑡 = 𝑅1 𝑡1 × 𝑅2 𝑡2 × ⋯ × 𝑅𝑛 𝑡𝑛 = 𝑅𝑖 (𝑡𝑖 ) 𝑅𝑠 𝑡 = 𝑒− 0 ℎ𝑖 𝑡 𝑑𝑡
= 𝑒− 0 𝑖=1 ℎ𝑖 𝑡 𝑑𝑡

𝑖=1 𝑖=1
• 𝑡𝑖 is the age of the item 𝑖, which is equivalent to age 𝑡 of the
system. That is, for the system to survive up to age 𝑡, the item 𝑖
should survive up to 𝑡𝑖 .

S. N. Njoroge 29 S. N. Njoroge 30

Series Systems - Example Series Systems - Example

• Avionics system of an aircraft consists of digital auto-pilot,
integrated global positioning system, weather and ground
mapping radar, digital map display and warning system. Apart
from the above items, the avionics system has control software.
The time-to-failure distributions of various items are given in
slide 30. Find the reliability of the avionics system for 100
hours of operation if all the items are necessary to maintain the
required function of the avionics system.
• Ans: 0.6646

5
 09/07/2015

S. N. Njoroge 31 S. N. Njoroge 32

Hazard Rate of Series Configuration

• If a system has n items, the hazard rate depends on the PDF of the MTTF of a Series Configuration
failure as seen below
• The mean time to failure, MTTF, of a series configuration,
denoted by 𝑀𝑇𝑇𝐹𝑆 , can be written as

∞ ∞ 𝑛
𝑀𝑇𝑇𝐹𝑆 = 𝑅𝑠 (𝑇) 𝑑𝑡 = 𝑅𝑖 (𝑡)
0 0 𝑖=1
• when the time-to-failure distribution is exponential
1
𝑀𝑇𝑇𝐹𝑠 = 𝑛
𝑖=1 𝜆𝑖

S. N. Njoroge 33 S. N. Njoroge 34

LIFE EXCHANGE RATE MATRIX

MTTF of a Series Configuration - Ex
(LERM)
• A system consists of three items connected in series. The time- • See Handout on email
to-failure distribution and their corresponding parameter values
are given below. Find the mean time to failure of the system.
Compare the value of 𝑀𝑇𝑇𝐹𝑆 with mean time to failure of
individual items.

S. N. Njoroge 35 S. N. Njoroge 36

Parallel Configuration Parallel Configuration

• In a parallel configuration the system fails only when all the • Parallel components are introduced when the reliability
items of the system fail. requirements for the system are very high.
• To maintain the required function only one item of the system is • The use of more than one engine in aircraft is one of the obvious
required to function examples of parallel configuration, or parallel transformers in
a power sub-station
• Parallel items will increase cost, complexity and weight of the
system. Hence, the number of parallel items required should be
carefully determined and if possible optimized

6
 09/07/2015

S. N. Njoroge 37 S. N. Njoroge 38

Characteristics of a Parallel Configuration Reliability Function of Parallel Configuration

• The system reliability, 𝑅𝑆 (𝑡), is more than reliability of the any • As the system fails only when all the items fail, the failure
of the consisting items. function, 𝐹𝑆 (𝑡), of the system is given by:
• If 𝑕𝑖 (𝑡) represent the hazard rate of item 𝑖, then the reliability
function of a parallel configuration can be written as: 𝐹𝑆 𝑡 = 𝑃,𝑇𝑇𝐹1 ≤ 𝑡 × 𝑇𝑇𝐹2 ≤ 𝑡 × ⋯ × 𝑇𝑇𝐹𝑛 ≤ 𝑛-
𝑛
𝑡
− 0 ℎ𝑖 𝑡 𝑑𝑡
𝑅𝑆 𝑡 = 1 − ,1 − 𝑒 - 𝐹𝑆 𝑡 = 𝐹1 𝑡 × 𝐹2 𝑡 × ⋯ × 𝐹𝑛 (𝑡)
𝑖=1 • The reliability function, 𝑅𝑆 (𝑡), of a parallel configuration can
then be written as:
𝑅𝑠 𝑡 = 1 − 𝐹𝑠 𝑡
= 1 − * 1 − 𝑅1 𝑡 × 1 − 𝑅2 𝑡 × ⋯ × 1 − 𝑅𝑛 𝑡 +
𝑛

𝑅𝑠 𝑡 = 1 − 1 − 𝑅𝑖 𝑡
𝑖=1

S. N. Njoroge 39 S. N. Njoroge 40

Parallel Configuration - Example Hazard Function of a Parallel Configuration

• A fly-by-wire aircraft has four flight control system electronics • Hazard function, 𝑕𝑆 (𝑡), of a parallel configuration can be
(FCSE) connected in parallel. The time-to-failure of FCSE can written as:
be represented by Weibull distribution with scale parameter −𝑑𝑅𝑠 (𝑡) 1
𝑕𝑆 𝑡 = ×
𝜂 = 2800 and 𝛽 = 2.8. Find the reliability of flight control 𝑑𝑡 𝑅𝑠 (𝑡)
system for 1000 hours of operation.
𝑛 𝑛
• Ans: 0.999991 𝑗=1*𝑓𝑗 𝑡 × 𝑖=1,𝑖≠𝑗 𝐹𝑖 (𝑡)+
= 𝑛
1− 𝑖=1 1 − 𝑅𝑖 𝑡

S. N. Njoroge 41 S. N. Njoroge 42

Example MTTF of a Parallel Configuration

• For the flight control system electronics discussed in the slide • The mean time to failure of a parallel configuration, denoted by
39, find the hazard function of the system at time 𝑡 = 100 𝑀𝑇𝑇𝐹𝑆 , can be written as:
∞ ∞ 𝑛
Ans: 8.0 × 10−8
𝑀𝑇𝑇𝐹𝑆 = 𝑅𝑠 𝑑𝑡 = *1 − ,1 − 𝑅𝑖 (𝑡)-+ 𝑑𝑡
0 0 𝑖=1
• For most of the failure distributions we may have to use
numerical integration to evaluate the above integral. However, in
case of exponential distribution we can get simple expression for
system’s MTTF as
∞ 𝑛 ∞ 𝑛

𝑀𝑇𝑇𝐹𝑆 = 𝑅𝑖 (𝑡) 𝑑𝑡 = 1− 1 − 𝑒 −𝜆𝑖 𝑡 𝑑𝑡

0 𝑖=1 0 𝑖=1
• N.B: Assuming that the TTF distribution of component 𝑖 is
exponential with mean 1 𝜆𝑖

7
 09/07/2015

S. N. Njoroge 43 S. N. Njoroge 44

r-out-of-n Systems Configuration Series and Parallel Configuration

• See handout on email • Are of two types
• Model 1. Series-Parallel Configuration
• Model 2. Parallel-Series System

S. N. Njoroge 45 S. N. Njoroge 46

Series-Parallel Configuration Series-Parallel Configuration

• System has a series structure with n items where each item has • If 𝑅𝑖,𝑗 (𝑡) denotes the corresponding reliability of the 𝑗𝑡ℎ parallel
parallel redundant items. component of the item 𝑖 component, then the reliability of item
• We assume that item 𝑖 has 𝑚𝑖 components in parallel. 𝑖 of the system is given by:
𝑚𝑖
𝑅𝑖 𝑡 = 1 − ,1 − 𝑅𝑖,𝑗 (𝑡)-
𝑗=1
• the system reliability can be written as:
𝑛 𝑛
𝑚𝑖
𝑅𝑠 𝑡 = 𝑅𝑖 𝑡 = *1 − ,1 − 𝑅𝑖,𝑗 (𝑡)-+
𝑗=1
𝑖=1 𝑖=1

S. N. Njoroge 47 S. N. Njoroge 48

Parallel-Series System Parallel-Series System

• The system has 𝑛 items connected in parallel where each item • Since item 𝑖 has 𝑚𝑖 components in series, the reliability of item
has 𝑚𝑖 components connected in series. 𝑖 is given by
𝑚𝑖
• An aircraft with more than one engine, is a typical example for
these type of configuration 𝑅𝑖 = 𝑅𝑖,𝑗 (𝑡)
𝑗=1
• where 𝑅𝑖,𝑗 (𝑡) is the reliability function of the component 𝑗 in
item 𝑖.
• Now the reliability of the parallel-series system is given by:
𝑛 𝑛 𝑛

𝑅𝑠 𝑡 = 1 − 1 − 𝑅𝑖 𝑡 =1− ,1 − 𝑅𝑖,𝑗 (𝑡)-

𝑖=1 𝑖=1 𝑗=1