Predict Future Failures From

Your Maintenance Records

Presented by:

Paul Barringer, P.E.

Barringer & Associates, Inc.
Reliability, Engineering, and Manufacturing Consultants
Humble, Texas 77347, USA
Phone: 1-281-852-6810
hpaul@barringer1.com

FAX: 1-281-852-3749
http://www.barringer1.com

Barringer & Associates, Inc. 2003

Crow/AMSAA Plots
Most data fit a power law distribution which
gives a straight line on log-log paper
We had Duane plots, then Duane/AMSAA
plots, and now Crow/AMSAA plots
The methodology handles mixed failure
modes and forecast future failures
Results are simple and easy to use log-log
plots of failures vs timeeasy to explain,
easy to make, and easy to forecast failures
Barringer & Associates, Inc. 2003

How Did This Get Started?

T. P. Wright (1936) used the idea for
learning curves in manufacturing
The WWII War Production Board used
Wrights techniques to forecast manpower
James Duane of General Electric applied
the concept to failure data for MTBF
US Army Material Systems Analysis Activity
wrote MIL-HDBK-189 and published new
data in TR-652 (Growth Guide)download
a copy from http://www.barringer1.com
Barringer & Associates, Inc. 2003

Unit Cost (hrs/unit)

Figure 1

Y=100X^(-0.3219)

Y=105.3268X^0.7739

Units Produced
Cum Aveg Cost (hrs)

Note the time decrease

of 20% for each unit

Cum Unit Cost (hrs)

Units Produced

T. P. Wrights
Idea For
Learning
Curves
Which Later
Drives
James Duanes
Idea For
Reliability
Growth Plots

Y=105.3268X^(-0.2260)

Units Produced
Barringer & Associates, Inc. 2003

Table 1

Maintenance Interventions
Month
1995
1996
1997
January
35
12
8
February
32
13
3
March
28
12
15
April
30
11
5
May
41
11
10
June
30
11
9
July
16
15
8
August
9
7
18
September
21
8
7
October
14
8
9
November
12
10
7
December
11
10
8
Total =
288
130
96
TPM Began August 1995

Actual
Failure
Data
From A
Chemical
Plant In
Brazil

Barringer & Associates, Inc. 2003

Total Cumulative Maintenance Interventions

Figure 2

Reduction Of Pump Failures In A Brazilian

Monovinyl-Chloride Plant By Use Of TPM

1000
Crow/rgr

Repair Reduction
After TPM

TPM Begins Aug 95

N=77.49t^0.529 with R^2=0.999

Before TPM

Engineering Starts TPM Jun 95

100
Slight Improvement
Before TPM

N=34.65t^0.947 with R^2=0.997

Jan 95

Lambda Beta r^2

34.65 0.947 0.997
77.49 0.529 0.999

Updated
Yr 2003
M01 D07
HPB

10
1

10

Months (Total Cumulative)

Barringer & Associates, Inc. 2003

100
Big Improvement
After TPM

The Fearless Forecast Math

N = *t

N = cum number of failures

= y-intercept at t=1
< 1 improvement
t = cum time
= 1 no change
> 1 deterioration
= slope of the line

Before TPM
N = 34.65*t0.947
After TPM
N = 77.49*t0.529

Forecast at t = 40 months
N = 34.65*400.947= 1140
Forecast at t = 40 months
N = 77.49*400.529= 545
=1140 - 545 = 595 failures
avoided
7

Barringer & Associates, Inc. 2003

The Fearless Monthly Forecast

Maintenance Interventions
Month
January
February
March
April
May
June
July
August
September
October
November
December
Total =

1995
35
32
28
30
41
30
16
18
21
14
12
11
288

1996
1997
'98 Fcst
12
8
8
13
3
7
12
15
7
11
5
7
11
10
7
11
9
7
15
8
7
9
7
7
8
7
7
8
9
7
10
7
7
10
8
7
130
96
85
TPM Began August 1995
Barringer & Associates, Inc. 2003

'99 Fcst
7
7
6
6
6
6
6
6
6
6
6
6
74
8

Failures Represented By
Monthly Maintenance Costs
Table 3

Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec

Petroleum Refinery Department Maintenance Cost History For One Area

1999
2000
2001
2002
Cum Days
Cum $'s
Cum Days
Cum $'s
Cum Days
Cum $'s
Cum Days
Cum $'s
31
$ 210,097
396
$ 4,146,017
762
$ 8,805,297
1127
$ 13,627,145
59
$ 456,441
425
$ 4,450,893
790
$ 9,077,531
1155
$ 14,076,446
90
$ 756,350
456
$ 4,846,968
821
$ 9,435,355
1186
$ 14,275,526
120
$ 1,028,044
486
$ 5,129,931
851
$ 9,746,244
1216
$ 14,537,284
151
$ 1,262,368
517
$ 5,673,580
882
$ 10,135,413
1247
$ 14,937,865
181
$ 1,540,101
547
$ 6,147,311
912
$ 10,674,844
1277
$ 14,732,077
212
$ 1,815,380
578
$ 6,896,160
943
$ 10,957,464
1308
$ 15,075,166
243
$ 2,121,788
609
$ 7,537,645
974
$ 11,420,963
1339
$ 15,310,813
273
$ 2,769,953
639
$ 7,856,635
1004
$ 11,932,656
1369
$ 15,589,596
304
$ 3,047,065
670
$ 8,254,432
1035
$ 12,857,704
1400
$ 15,826,120
334
$ 3,360,486
700
$ 8,716,149
1065
$ 13,402,128
1430
$ 15,944,082
365
$ 3,748,406
731
1096
1461
$ 16,275,941
$ 8,440,050
$ 13,214,697

Circled Data Points Are Problems! The cumulative $s decreases

thus do not include this data point in the regression for & .
However note that the $ values continue into subsequent data points.
9

Barringer & Associates, Inc. 2003

Maintenance Costs (Total Cumulative US$)

Figure 3

Department Maintenance Data--1st Look

Crow/rgr

$10,000,000

Zoom On This Region

Trend Line Using All Data

$1,000,000

Lambda Beta r^2

4228 1.148 0.996
$100,000
10

100

Days (Total Cumulative)

Barringer & Associates, Inc. 2003

1000
10

Department Maintenance Data-2nd Look

Figure 4
Maintenance Cost (Total Cumulative US$)

$ 20,000,000
Crow/rgr
$ 18,000,000

11 Month
Cum Savings
$2,222,000

Dec 2002
C$ = 110839t0.696

$ 16,000,000

Feb 2002 Cusp

Announced
TPM Program

$ 14,000,000

Started TPM
Efforts Jan 2002
$ 12,000,000
Lambda Beta r^2
3361 1.192 0.998
110839 0.696 0.995

C$ = 3361t1.192
$ 10,000,000
800

900

1000

1200

1400

1600

1800

Days (Total Cumulative)

2000
11

Barringer & Associates, Inc. 2003

Table 4
Safety Record--Major Chemical Plant Incidents
Cum
Days

Cum
Incidents

Cum
Days

Cum
Incidents

Cum
Days

Cum
Incidents

Cum
Days

Cum
Incidents

1
8
23
47
53
58
65
67
72
78
94
105
106
108
124
149
226
228
248
285
288
289
310
312

1
2
3
4
5
6
7
8
9
10
12
13
14
15
16
17
18
19
20
21
22
23
24
25

367
368
429
526
553
585
598
599
600
632
635
660
677
690
719
759
773
830
878
1009
1018
1031
1040
1044

26
27
28
29
30
31
32
33
34
36
37
39
40
41
42
44
45
46
47
48
49
50
51
52

1046
1096
1184
1195
1291
1345
1397
1565
1591
1598
1624
1626
1634
1655
1670
1692
1711
1753
1759
1990
2186
2430
2472
2509

53
54
55
56
57
58
59
60
61
62
63
74
75
76
77
78
79
81
82
83
84
85
86
87

2622
2742
2754
2825
2846
2851
2888
2922
2969
2984
3099
3106

88
89
90
92
93
94
95
96
97
99
100
101

8.5 years

Barringer & Associates, Inc. 2003

Safety Incidents
Are Failures!

Long term
average MTBF
is 3106/101 =
30.7 days/incident

12

Figure 5

Safety Incidents
Jan '94-Aug '02

300

Forecast

Crow/IEC1164

Safety Incidents
(Total Cumulative)

100

2-Aug-02
19-Sep-02
7-Nov-02
26-Dec-02
13-Feb-03
4-Apr-03
24-May-03
13-Jul-03
2-Sep-03
22-Oct-03
12-Dec-03
2-Feb-04
24-Mar-04
15-May-04

Unfavorable Trends

10

Lambda
0.6177
1

10

Actual

101
102
103
104
105
106
107
108
109
110
111
112
113
114

Forecast

Beta
0.634

1000
100
Time (Total Cumulative Days)

10000
13

Barringer & Associates, Inc. 2003

Fearless Forecast Of Spills

Table 5
Raw Data

Crow/AMSAA Data

Spill Date

Days
Between
Spill

Spill
Events

Cum.
Days

Cum.
Spills

11/18/1995
1/31/1996
5/8/1996
5/22/1996
7/29/1996
8/23/1996
8/25/1996
6/20/1997
2/22/1998
2/10/1999

35
74
98
14
68
25
2
299
247
353

1
1
2
1
1
1
1
1
1
1

35
109
207
221
289
314
316
615
862
1215

1
2
4
5
6
7
8
9
10
11

Forecasts
Failures
New
Predicted
Method
By Old
Savings
Method

8 Failures In
316 Days
This Is The
Datum
18
27
41

9
17
30

3 Failures In 899 DaysThis Is Good News!

Barringer & Associates, Inc. 2003

14

Figure 6
100

Spills

Crow/IEC1164

Spills (Total Cumulative)

Old Way

Spill Reduction

10
Improvement

Lambda Beta Fit-P%

0.0074011 1.214 <10!
1.989
0.241 >=10
1

10

100
1000
Time
(Total
Cumulative
Days)
Barringer & Associates, Inc. 2003

10000
15

Table 6
Raw Data
Spill Date
11/18/1995
1/31/1996
5/8/1996
5/22/1996
7/29/1996
8/23/1996
8/25/1996
6/20/1997
2/22/1998
2/10/1999
8/16/1999
11/7/1999
2/12/2000
4/29/2000
11/16/2000
12/25/2000
3/25/2001
8/1/2001
10/28/2001
7/10/2002
7/25/2002
9/6/2002
2/18/2003

Crow/AMSAA Data

Days
Between
Spill

Spill
Events

Cum.
Days

Cum.
Spills

35
74
98
14
68
25
2
299
247
353
187
83
97
77
201
39
90
129
88
255
15
43
165

1
1
2
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

35
109
207
221
289
314
316
615
862
1215
1402
1485
1582
1659
1860
1899
1989
2118
2206
2461
2476
2519
2684

1
2
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24

Forecasted Failures
Failures
Missed
New
Predicted
Opportunities
Method
By Old
From
Savings
Method
Relapse

18
27
41

9
17
30

Bad News!
Failures
Coming
Faster!

Barringer & Associates, Inc. 2003

Good
News!
1
2
2
3
4
5
5
6
7
9
9
9
11

16

Figure 7

Spills

100
Crow/IEC1164
Relapse

Re
tre
nc
h

ct
io
n
Re
du
ill
Sp

Spills (Total Cumulative)

Old Way

Weve
Lost
The
Way To
Success!

10
Improvement
Lambda
0.0074011
1.989
0.0091392
1

10

Beta
1.214
0.241
0.997

Fit-P%
<10!
>=10
>=10

100
1000
Time (Total Cumulative Days)

10000
17

Barringer & Associates, Inc. 2003

Table 7

All Outages

Date

Event
Outage

Event Description

2/1/1999
2/20/1999
2/24/1999
5/22/1999
7/9/1999
8/9/1999
9/13/1999
10/13/1999
11/3/1999
11/6/1999
11/10/1999
1/3/2000
6/12/2000
6/21/2000
9/11/2000
11/7/2000
12/2/2000
12/12/2000
4/11/2001
4/12/2001
4/19/2001
6/7/2001
8/22/2001
9/13/2001
9/16/2001
10/6/2001
10/12/2001
10/31/2001
12/1/2001
1/1/2002
4/15/2002
4/18/2002
9/27/2002
12/6/2002
1/3/2003

Planned
Planned
Forced
Forced
Planned
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Planned
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Forced

Tie In
Tie In
Gas Line Outage
Animal Contact
Interconnect Energized
Switching Error
Tie Wrap Failure
Lightning Strike
Static Wire Short
Switch Failed
Not Logged
Cable Bond Fault
Underground Cable Fault
Bird Contact
Lightning Strike
Animal Contact
Animal Contact
High Winds
Not Logged
Not Logged
Tie In
Not Logged
Pole Damage
Interconnect Opened
Supplemental Power Out
Power Dip
Control Tripped
Power Dip
Power Dip
Steam Outage
Switching Error
Load Shedding Error
Water In Switch Gear
Generator Air Intake Frozen
UPS Failure

Days
Between
Event
0
19
4
87
48
31
35
30
21
3
4
54
161
9
82
57
25
10
120
1
7
49
76
22
3
20
6
19
31
31
104
3
162
70
28

Cum.
Days

Cum
Failures

0
19
23
110
158
189
224
254
275
278
282
336
497
506
588
645
670
680
800
801
808
857
933
955
958
978
984
1003
1034
1065
1169
1172
1334
1404
1432

0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34

Barringer & Associates, Inc. 2003

Forced Outages
Cum.
Days

Cum
Failures

23
110

1
2

189
224
254
275
278
282
336
497
506
588
645
670
680
800
801

3
4
5
6
7
8
9
10
11
12
13
14
15
16
17

857
933
955
958
978
984
1003
1034
1065
1169
1172
1334
1404
1432

18
19
20
21
22
23
24
25
26
27
28
29
30
31

Co-gen
Failure Log:
31 Forced
Outages In
1432 Days
= 46.5 days
per failure

18

Combined Cycle Co-Gen Plant

Figure 8
100

Forced Outages (Total Cumulative)

Crow/IEC1164

Forced Outages
10

Lambda Beta Fit-P%

0.02221 0.996 >=10
1

1000
100
Time (Total Cumulative Days)

10

Barringer & Associates, Inc. 2003

Table 7 With Failure Forecast

Date

Event
Outage

2/1/1999
2/20/1999
2/24/1999
5/22/1999
7/9/1999
8/9/1999
9/13/1999
10/13/1999
11/3/1999
11/6/1999
9/27/2002
12/6/2002
1/3/2003
2/23/2003
4/11/2003
5/27/2003
7/13/2003
8/28/2003
10/14/2003
11/29/2003
1/15/2004
3/2/2004
4/17/2004
6/3/2004

Planned
Planned
Forced
Forced
Planned
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Forced
Forced

All Outages
Event Description
Tie In
Tie In
Gas Line Outage
Animal Contact
Interconnect Energized
Switching Error
Tie Wrap Failure
Lightning Strike
Static Wire Short
Switch Failed
Water In Switch Gear
Generator Air Intake Frozen
UPS Failure
Predicted
Predicted
Predicted
Predicted
Predicted
Predicted
Predicted
Predicted
Predicted
Predicted
Predicted

Days
Between
Event
0
19
4
87
48
31
35
30
21
3
162
70
28

19

Forced Outages

Cum.
Days

Cum
Failures

Cum.
Days

Cum
Failures

0
19
23
110
158
189
224
254
275
278
1334
1404
1432

0
1
2
3
4
5
6
7
8
9
32
33
34

23
110

1
2

189
224
254
275
278
1334
1404
1432
1483
1530
1577
1623
1670
1716
1763
1809
1856
1903
1949

3
4
5
6
7
29
30
31
32
33
34
35
36
37
38
39
40
41
42

More Failures
Predicted Into
The Future From
Lack Of An
Improvement
Program ( 1)!

Barringer & Associates, Inc. 2003

5000

No
Change

20

10

Summary
Five actual datasets demonstrate straight
lines of cum failures vs cum time
Crow/AMSAA technology allows failure
forecast of the mixed failure modes
The forecast objective is to start decisive
corrective action and prevent predicted
failures
Preventing failures requires proactive effort
Preventing future failures saves $!
Barringer & Associates, Inc. 2003

21

11