ME6226, Product Reliability

Dr. Piyush Shakya

Engineering Asset Management (EAM) Group

Department of Mechanical Engineering

Indian Institute of Technology Madras
[email protected]

https://home.iitm.ac.in/pshakya/

February 29, 2024

Dr. Piyush Shakya (EAM, IITM) https://home.iitm.ac.in/pshakya/ February 29, 2024 1 / 62

Contents

1 System Reliability from Component reliabilities

2 Reliability Logic Diagrams

Dr. Piyush Shakya (EAM, IITM) https://home.iitm.ac.in/pshakya/ February 29, 2024 2 / 62

System Reliability

• A system is more complicated than the individual subsystems.

• The subsystems reliability is computed by performing failure data

analysis individually.

• The conguration in which subsystems are connected comes into

consideration.

• The congurations may be categorized into four types.

i Series
ii Parallel
iii Mixed
iv Complex

Dr. Piyush Shakya (EAM, IITM) https://home.iitm.ac.in/pshakya/ February 29, 2024 3 / 62

Series Conguration

• All subsystems should work for the system to work.

• Ej = the event that j th component doesn't fail before time t.

• P (Ej ) = Rj (t) ( j th component doesn't fail before time t).

• P (S) = P (E1 ∩ E2 ∩ E3 ∩ E4 ∩ · · · · · · ∩ En )

P (E1 ∩E2 )
• P (E2 |E1 ) = P (E1 ) (Conditional Probability)

Dr. Piyush Shakya (EAM, IITM) https://home.iitm.ac.in/pshakya/ February 29, 2024 4 / 62

Series Conguration continued...

• P (E1 ∩ E2 ) = P (E2 |E1 ).P (E1 )

P ((E1 ∩E2 )∩E3 )
• P (E2 ∩E1 ) = P (E3 |(E2 ∩ E1 ))
•

theref ore, P (E1 ∩ E2 ∩ . . . En ) = P (E1 ).P (E2 |E1 ).P (E3 |E2 ∩ E1 )
.P (E4 |E3 ∩ E2 ∩ E1 ) . . . P (En |En−1 ∩ . . . E1 )

• If all Ej are independent of each other.

• P (E1 ∩ E2 ∩ E3 . . . En ) = P (E1 ).P (E2 ).P (E3 ) . . . P (En )

• P (S) = P (E1 ).P (E2 ).P (E3 ) . . . P (En )
Yn
• Rs (t) = R1 (t).R2 (t).R3 (t) . . . Rn (t) = Ri (t)
i=1
• therefore, Rs (t) ≤ min{R1 (t), R2 (t), R3 (t) . . . Rn (t)}
Dr. Piyush Shakya (EAM, IITM) https://home.iitm.ac.in/pshakya/ February 29, 2024 5 / 62
Failure modes for a system/components

• As system fails if any of these failure mode occurs.

• Assuming all failure modes are independent.

n
Q
• The system reliability: Rs (t) = Ri (t), where Ri (t) is the proba-
i=1
bility that the ith failure mode doesn't occur before time t.

• The reliability for the independent failure mode is same as series

conguration.

• Reliability of series system is always worse than its poorest

subsystem/failure mode.

Dr. Piyush Shakya (EAM, IITM) https://home.iitm.ac.in/pshakya/ February 29, 2024 6 / 62

Failure data analysis (earlier slide)

• If the total population size is signicantly increased the histogram

of failure density (discussed earlier) will become a smooth curve.

Pk R∞
• j=1 fdj ∆t = 0 fd (ξ)dξ = 1

• fd (ξ) is probability density function and

R∞ ξ is the time to failure
(random variable).
0 fd (ξ)dξ = 1

Dr. Piyush Shakya (EAM, IITM) https://home.iitm.ac.in/pshakya/ February 29, 2024 7 / 62

Failure data analysis (earlier slide)
Pj Rt
• F (j∆t) = i=1 fdi ∆t will turn into F (t) = 0 fd (ξ)dξ

Rt R∞
• R(t) = 1 − F (t) =⇒ R(t) = 1 − 0 fd (ξ)dξ = t fd (ξ)dξ

fd (t)
• Recall from earlier Z(t) =
R(t)

Pk R∞
• MTTF= j=1 fdj (j∆t)∆t → MTTF = 0 tfd (t)dt

• The following expressions also may be proven easily

− 0t Z(ξ)dξ
R
• R(t) = e

R∞ R∞
• MTTF = 0
tfd (t)dt = 0
R(t)dt

Dr. Piyush Shakya (EAM, IITM) https://home.iitm.ac.in/pshakya/ February 29, 2024 8 / 62

System Reliability relation with failure rate
Rt
− Zs (ξ)dξ
• System reliability Rs (t) = e 0

• If subsystem/component j has hazard rate as Zj (ξ) then

Rt Rt Rt
n
Q − Z1 (ξ)dξ − Z2 (ξ)dξ − Zn (ξ)dξ
• Rs (t) = Ri (t) = e 0 e 0 ...e 0

i=1

Rt Rt P
n
− (Z1 (ξ)+Z2 (ξ)+...···+Zn (ξ))dξ − (Zj (ξ)dξ)
• Rs (t) = e 0 =e 0 j=1

Rt Rt P
n
− (Zs (ξ)dξ) − (Zj (ξ)dξ) n
P
• Rs (t) = e 0 =e 0 j=1
; =⇒ Zs (ξ) = Zj (ξ)
j=1

Dr. Piyush Shakya (EAM, IITM) https://home.iitm.ac.in/pshakya/ February 29, 2024 9 / 62

System Reliability relation with failure rate (Exponential)
(
1 − e−λt , time to f ailure (t) ≥ 0 1
1 F (t) = ; Mean=
λ;
0, time to f ailure (t) < 0
1
Variance= 2
λ
(
e−λt , t ≥ 0
2 R(t) = 1 − F (t) =
1, t<0
Rt
3 On comparing with R(t) = e− 0 Z(ξ)dξ
=⇒ Z(ξ) = λ(constant)

4 Failure rate is constant for exponential distribution.

5 If each components/failure mode follows constant hazard rate

(CHR)/CFR then Zs = λ1 + λ2 + λ3 + . . . · · · + λn = constant &
the system will be have CFR (exponential disttribution).

Dr. Piyush Shakya (EAM, IITM) https://home.iitm.ac.in/pshakya/ February 29, 2024 10 / 62

System Reliability relation with failure rate (Weibull)
−( t )β
(
( βη )( ηt )(β−1) e η , t ≥ 0
1 f (t) =
0, t<0

−( t )β
(
1−e η , t≥0
2 For Weibull distribution F (t) =
0, t<0

−( t )β
(
e η , t≥0
3 R(t) = 1 − F (t)=
1, t<0

fd (t)
4 on comparing with Z(t) = =⇒ Z(ξ) = ( βη )( ηξ )β−1
R(t)

Dr. Piyush Shakya (EAM, IITM) https://home.iitm.ac.in/pshakya/ February 29, 2024 11 / 62

System Reliability relation with failure rate (Weibull)

• If each components has Weibull failure density function such that

n
βj ξ βj −1 P βj ξ βj −1
Zj (ξ) = ηj ( ηj ) , then Zs (ξ) = ηj ( ηj )
j=1
• So the system does not exhibit Weibull type failures even though all
components show Weibull type failures.
n
P (ξ)β−1
• However, if βj = β for all j = 1, 2, 3, . . . , n; Zs (ξ) = β (ηj )β
j=1
n
βs ξ βs−1
(ξ)β−1 (η1)β
P
• ηs ( ηs ) =β
j
j=1
" #− 1
n β
P 1
• Therefore, βs = β & ηs = (ηj )β
j=1

• So the system follows the Weibull distribution only if all components

have the same shape parameter (β).

Dr. Piyush Shakya (EAM, IITM) https://home.iitm.ac.in/pshakya/ February 29, 2024 12 / 62

Mean time to failure (Approach)
 
Rt
R∞ R∞ − Zs (ξ)dξ
• MTTF = R(t)dt; Therefore, MTTFs = e 0  dt
0 0

• for n components connected in series & having constant failure rate

(but dierent for each components)

• Zs (ξ) = Zs (constant) = Z1 + Z2 + Z3 + . . . Zn
R∞ h
−1 −Zs t
i∞
• MTTFs = e−Zs t dt = Zs e
0 0

(−0+1) 1 1
• MTTFs =
Zs = Zs = j=1
P
Zj
n

• as ZS ↑ MTTF↓; Similarly may be done for other distributions.

Dr. Piyush Shakya (EAM, IITM) https://home.iitm.ac.in/pshakya/ February 29, 2024 13 / 62

Practice Problems

Question-1

If the components A has reliability of 0.99 & components B has reliability

of 0.999. If SA10 denotes a system with 10 A type components in series.
What is the reliability of SA10 , SA100 , SB10 , SB100 , and SA50 B50 ?

✓ RSA10 (t) = 0.9044,

✓ RSA100 (t) = 0.3660

✓ RSB10 (t) = 0.990;

✓ RSB100 (t) = 0.9048

✓ RSA50 B50 (t) = 0.5755,

Dr. Piyush Shakya (EAM, IITM) https://home.iitm.ac.in/pshakya/ February 29, 2024 14 / 62

Solution (Question-1)

Question-1

If the components A has reliability of 0.99 & components B has reliability

of 0.999. If SA10 denotes a system with 10 A type components in series.
What is the reliability of SA10 , SA100 , SB10 , SB100 , and SA50 B50 ?

✓ RSA10 (t) = (0.99)10 = 0.9044,

✓ RSA100 (t) = (0.99)100 = 0.3660

✓ RSB10 (t) = 0.990;

✓ RSB100 (t) = (0.999)100 = 0.9048

✓ RSA50 B50 (t) = (0.99)50 .(0.999)50 = 0.6050 × 0.9512 = 0.5755,

Dr. Piyush Shakya (EAM, IITM) https://home.iitm.ac.in/pshakya/ February 29, 2024 15 / 62

Practice Problems

Question-2

If a component has 10 failure modes that are independent and have the
same probability of failure. If the reliability of the components is 0.95.
What is the probability of failure for dierent failure modes?

✓ Ans: 0.0051

Dr. Piyush Shakya (EAM, IITM) https://home.iitm.ac.in/pshakya/ February 29, 2024 16 / 62

Solution (Question-2)

Question-2

If a component has 10 failure modes that are independent and have the
same probability of failure. If the reliability of the components is 0.95.
What is the probability of failure for dierent failure modes?

✓ If the probability of failure is F, then reliability for each failure mode

is (1 - F)

✓ Rs (t) = 0.95 = (1 − F )10

✓ Therefore, (1 − F ) = 0.9949 =⇒ F = 0.0051

Dr. Piyush Shakya (EAM, IITM) https://home.iitm.ac.in/pshakya/ February 29, 2024 17 / 62

Practice Problems

Question-3

If 5 components have MTTF respective as 1000 hrs, 2000 hrs, 3000 hrs,
4000 hrs, and 5000 hrs. What is the MTTF of the system if all the
components are congured in series and their failures follow exponential
distribution?

✓ Ans: 437.9 hrs

Dr. Piyush Shakya (EAM, IITM) https://home.iitm.ac.in/pshakya/ February 29, 2024 18 / 62

Solution (Question-3)

Question-3

If 5 components have MTTF respective as 1000 hrs, 2000 hrs, 3000 hrs,
4000 hrs, and 5000 hrs. What is the MTTF of the system if all the
components are congured in series and their failures follow exponential
distribution?

1 1 1 1 1
✓ λ1 = 1000 , λ2 = 2000 , λ3 = 3000 , λ4 = 4000 , and λ5 = 5000

60+30+20+15+12
✓ Zs = λ1 + λ2 + λ3 + λ4 + λ5 = 60000

137
✓ Zs = 60000

✓ MTTFs = 437.9 hrs

Dr. Piyush Shakya (EAM, IITM) https://home.iitm.ac.in/pshakya/ February 29, 2024 19 / 62

Practice Problems

Question-4

A product has 5 components (each following Weibull distribution) with

shape parameter 1.5. Their scale parameters are 3600, 7200, 5850, 4780,
and 9300 hours. Find the MTTF, Reliability & median time, assuming
that components are congured in series.

✓ MTTF =1664.5 hours

t
)1.5
 
✓ R(t) = exp −( 1842.7

✓ tmed = 1443.2 hours

Dr. Piyush Shakya (EAM, IITM) https://home.iitm.ac.in/pshakya/ February 29, 2024 20 / 62

Solution (Question-4)

Question-4

A product has 5 components (each following Weibull distribution) with

shape parameter 1.5. Their scale parameters are 3600, 7200, 5850, 4780,
and 9300 hours. Find the MTTF, Reliability & median time, assuming
that components are congured in series.

" #− 1
5 β
P 1
✓ ηs = (ηj )β
j=1

h i− 1
1 1 1 1 1 1.5
✓ ηs = (3600)1.5
+ (7200)1.5
+ (5850)1.5
+ (4780)1.5
+ (9300)1.5
= 1842.7

   
✓ MTTF = η Γ 1 + β1 = 1842.7 Γ 1 + 13 = 1664.5 hours
2

Dr. Piyush Shakya (EAM, IITM) https://home.iitm.ac.in/pshakya/ February 29, 2024 21 / 62

Solution (Question-4) continued ...
h i
= exp −( ηt )β = exp −( 1842.7
t
)1.5
 
✓ R(t)

✓ for tmed , R(tmed ) = 0.5

1
✓ tmed = η[−ln(0.5)] 1.5

1
✓ tmed = 1842.7 × (0.69315) 1.5 = 1443.2 hours

Dr. Piyush Shakya (EAM, IITM) https://home.iitm.ac.in/pshakya/ February 29, 2024 22 / 62

Thank You

It does not matter how slowly you go as long as you do not stop. 
Confucius

Dr. Piyush Shakya (EAM, IITM) https://home.iitm.ac.in/pshakya/ February 29, 2024 62 / 62