Total Control Over Human Error

A Reliability Model

ICOMS 2008 Conference

Mike Sondalini
Lifetime Reliability Solutions
www.lifetime-reliability.com

www.lifetime-reliability.com 1
Each Process is Part of a Series of
Processes
Process 10 Process 23

Process 25

Process 13

Process 16

Process 17
Process 2
Process 1

Process 6

Process 9
Process 24

Process 5 Process 12

Process 19
Raw Preparation Manufacture Assembly Packaging Product
Materials

R1 R2 R3 R4 R5

Process 14
Process 12

Process 15
Process 3

Process 7

Process 8

Process 4 Process 26 Process 18

Process 11
Process 20

Process 14
Process 22

Process 21

All business processes involve a series of work processes

www.lifetime-reliability.com 2
 The Process Chains the Bind Us
ss 2 ss 10
Proce Proce

Proce

Proce

Proce
Proce

Proce

Proce

Proce
ss 25

ss 13

ss 16

ss 17
Proce

ss 1

ss 6

ss 9
Process 10 Proce Process 23
ss 23

Process 25

Process 13 Proce

Process 16

Process 17 Proce
Process 2

Proce
ss 19
Process 1

Process 6

Process 9
Proce Proce ss 24
ss 1224
Process ss 5
Ra Pr Ma As Pa Pr
Process 5 w
Process
R 12 epR nuf
R se R ck R
ss 2 od ss 10

Proce

Proce

Proce
Proce

Proce

Proce
ss 14
ss 11

Process 19

ss 12

ss 15
ss 3

ss 7

ss ss
M ara act m agi Proce uct Proce

Proce

Proce

Proce

Proce

Proce
Proce
Proce ureProce bly ngProce

ss 25

ss 13

ss 16

ss 17
1 2 3 4 5 ss 22 Proce

ss 6

ss 9
at tio

8 1
Proce

Proce
ss 4 ss 26 ss 18 Proce

ss 20

ssss1419
eri n
Proce ss 23

Proce
Proce Proce ss 24
Raw Preparation Manufacture als Assembly
Proce ss 5 Packaging ss 12 Product
Materials ss 21 Ra Pr Ma As Pa Pr
wR ep R nuf
R seR ckR od

Proce

Proce Process

Proce
Proce

Proce

Proce
ss 14
ss 11

ss 12

ss 15
ss 3

ss 7

ss 8
R1 R2 R3 R4 M ara R5 act m agi
Proce uct
1
at Proce
tio
2
ureProce bly
3 4
ngProce
5
ss 2
ss 22
ss 10

Process 14

Proce

Proce
ss 4n ss 26 ss 18

ss 20

ss 14
Process 12
Process 3

Process 7

Proce
eri
Process 8

Proce Proce

Proce

Proce

Proce
Proce

Proce

Proce

Proce
als

ss 25

ss 13

ss 16

ss 17
Process 4 Process 26 Proce Process 18 Proce

ss 1 15

ss 6

ss 9
Proce ss 23
ss 21 Proce

Proce
ss 19
Process 11 Proce ss 24
ss 5 ss 12
Process 20

Process 14
Ra Pr Ma As Pa Pr
wR ep R nuf
R se
R ck
R od

Proce

Proce

Proce
Process 22

Proce

Proce

Proce
ss 11

ss 14
ss 12

ss 15
ss 3

ss 7

ss 8
M ara act m agi
Proce uct
Process 21
1
at Proce
tio
2
ureProce bly
3 4
ngProce
5 ss 22

Proce

Proce
ss 4n ss 26 ss 18

ss 20

ss 14
Proce
eri
als
Proce
ss 21

Hundreds of activities across dozens of processes – what chance have you got?
www.lifetime-reliability.com 3
 Work Process Reliability

A Job

Task 1 Task 2 Task 3 Task 4 Task 5 Outcome

R1 R2 R3 R4 R5

Rjob = R1 x R2 x R3 x R4 x R5 Eq. 1

How well a job is done is a simple matter of chance

www.lifetime-reliability.com 4
 What is the Chance of Good Work
No Situation and Task Error Rate (per task) Reliability Rate

1 Routine simple tasks Read checklist or digital display wrongly 0.001 0.999

2 Check for wrong indicator in an array 0.003 0.997

3 Fail to correctly replace printed circuit board (PCB) 0.004 0.996

4 Wrongly carry out visual inspection for a defined criterion (e.g. leak) 0.003 0.997

5 Select wrong switch among similar 0.005 0.995

6 Read 10-digit number wrongly 0.006 0.994

7 Routine task with care needed Wrongly replace a detailed part 0.02 0.98

8 Put 10 digits into a calculator wrongly 0.05 0.95

9 Do simple arithmetic wrong 0.01 - 0.03 0.99 – 0.97

10 Read 5-letter word with poor resolution wrongly 0.03 0.97

11 Dial 10 digits wrongly 0.06 0.94

12 Punch or type character wrongly 0.01 0.99

13 Complicated, non-routine task Fail to notice incorrect status in roving inspection 0.1 0.9

14 New work shift – fail to check hardware, unless specified 0.1 0.9

15 High stress, non-routine work 0.25 0.75

16 Fail to notice wrong position of valves 0.5 0.5

17 Fail to act correctly after 1 minute in emergency situation 0.9 0.1

Smith, Dr, David J., Reliability, Maintainabilitywww.lifetime-reliability.com

and Risk, Seventh Edition, Appendix 6. Elsevier, 2005 5
 The Chance of Doing a Good Job
A Job

Task 1 Task 2 Task 3 Task 4 Task 5 Outcome

R1 R2 R3 R4 R5

Rjob = R1 x R2 x R3 x R4 x R5 Eq. 1

Rjob = 1 x 1 x 0.6 x 1 x 1 = 0.6 (or 60%) ‘One poor, all poor’

Rjob = 0.9 x 0.9 x 0.9 x 0.9 x 0.9 = 0.59 (or 59%) For a 0.9 maintenance job

Rjob = 0.75 x 0.75 x 0.75 x 0.75 x 0.75 = 0.237 (or 24%) Maintenance with stress

It is human error permitted by bad work process design that

makes our series arranged processes so unreliable.

www.lifetime-reliability.com 6
 A Maintenance Job, every job, …

…is a Series Work Process

Task 1 Task 2 Task 3 Task 4 Task 10 Task 11 Task 12

The Work Environment

1 2 3 4 5 6 7 8 9 10 11 12
R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12

What is the chance of all 12 tasks being done 100% perfectly?

What if there were 50 tasks in a job?
www.lifetime-reliability.com 7
 A Parallel Arrangement
Fortunately reliability principles also give us the answer to the problems
with series processes – the parallel process
1

3 Outcome
Redundant Activity
-
-

Rpara = 1 – [(1-R1) x (1-R2) x ….(1-Rn)] Eq. 2

Rpara = 1 – [(1-0.6) x (1-0.6) x (1-0.6) x (1-0.6)] Parallel four poor chances

= 1 – [(0.4) x (0.4) x (0.4) x (0.4)] = 1 – [0.0256] and your odds improve,
= 0.9744 (i.e. 97.4%) because not all go wrong
at the same time
www.lifetime-reliability.com 8
 There are big risks here, unless …

• Can each task be made more certain? … How certain?

• Can we include redundancy and turn tasks into a parallel arrangement? …

Series Tasks
1 2 3 4 5 6 7 8 9 10 11 12
R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12

Parallel Tasks
1 2 3 4 5 6 7 8 9 10 11 12

1 2 3 4 5 6 7 8 9 10 11 12

www.lifetime-reliability.com 9
 Parallel Process Reliability
Parallel Tasks
.9 .9 .9 .9 .9 .9 .9 .9 .9 .9 .9 .9

.9 .9 .9 .9 .9 .9 .9 .9 .9 .9 .9 .9

Equivalent Series Task

.99 .99 .99 .99 .99 .99 .99 .99 .99 .99 .99 .99

Parallel System Series System Without Parallel Test

R system= 1- [(1- R1)x(1- R2)x(1-R3) ….] R system= R1 x R2 x R3 …. R system= R1 x R2 x R3 …

1- [(1- 0.9)x(1- 0.9)] 0.99 x 0.99 x … 0.99 0.9 x 0.9 x … 0.9
1- [0.1 x 0.1] (0.99)12 = 0.8864 (0.9)12 = 0.2824
1- [0.01] = 0.99
If 50 tasks (0.99)50 = 0.6 If 50 tasks (0.9)50 = 0.0052

www.lifetime-reliability.com 10
We need 0.9999 0.999 is not enough

12 task series process: 12 task series process:

R system= R1 x R2 x R3 …. R system= R1 x R2 x R3 ….
0.9999 x 0.9999 x … 0.9999 0.999 x 0.999 x … 0.999
(0.9999)12 (0.999)12
0.9988 (~ 1 error in 1000 tasks – ~ 4.5σ) 0.9881 (~ 1 error in 100 tasks – ~ 3.5σ)

50 task series process: 50 task series process:

R system= R1 x R2 x R3 …. R system= R1 x R2 x R3 ….
0.9999 x 0.9999 x … 0.9999 0.999 x 0.999 x … 0.999
(0.9999)50 (0.999)50
0.995 (5 errors in 1000 tasks – ~ 4σ) 0.95 (5 errors in 100 tasks – ~ 3σ)

www.lifetime-reliability.com 11
How do we get maintenance tasks to
be 0.9999 reliable, when …

… they don’t make 0.99 reliable peoples?

www.lifetime-reliability.com 12
 Something else is interesting with
parallel tasks …
Create Parallel Tasks with a Proof Test to Prove Compliance
1 1 .9 1 .9 1 .9 1 1 .9 1 1

.9 .5 1 .9 1 .9 1 .9 .9 1 .9 .9

Parallel Systems
Rsystem= 1- [(1- R1)x(1- R2)]
1- [(1- 1)x(1- 0.5)]
1- [0 x 0.5]
1- [0] = 1

As long as one parallel task is done right …

the whole step is right – 100% right!
www.lifetime-reliability.com 13
 The J.A.L. Job
JAL Job

Skilled
Technician 0.9 0.9 0.9 0.9 0.9

Super Clear
Instructions 0.9 0.9 0.9 0.9 0.9

Expert
Supervisor 0.9 0.9 0.9 0.9 0.9

Proof Test
Task 0.99 0.99 0.99 0.99 0.99

Rtask = 1 – [(1-0.9) x (1-0.9) x (1-0.9) x (1-0.99)]

= 1 – [(0.1) x (0.1) x (0.1) x (0.01)]
= 0.99999 (i.e. 99.999%, or 1 error per 100,000 opportunities)

Rjob = 0.99999 x 0.99999 x 0.99999 x 0.99999 x 0.99999 = 0.99995 (i.e. 99.995%)

The error rate for the whole job is now 5 errors per 100,000 opportunities.

How well a job is done is a simple matter of controlling chance

www.lifetime-reliability.com 14
 But the best answer is to failure proof the
job design …

‘Perfection’
1 1 1 1 1 1 1 1 1 1 1 1

Failure-Proofed Series Systems

Rsystem= R1 x R2 x R3 …
1x1x…1
(1)12 = 1

(1)50 = 1

… and get each task 100% right - always perfect!

www.lifetime-reliability.com 15