WEBVTT

1
00:00:00.000 --> 00:00:08.490
This video asks two questions about repairable equipment in ReliaEn: is a machine wearing out, and how good are its repairs?

2
00:00:08.625 --> 00:00:20.335
First, a conveyor gearbox: it was repaired at 50, 80, 95, 105 and 110 hours of operation, and watched to 112 hours.

3
00:00:20.750 --> 00:00:25.490
In column A, type the machine's name, P, on each of the 5 rows.

4
00:00:25.490 --> 00:00:32.110
In column B, type the 5 repair ages, and in column C the observation end, 112, on every row.

5
00:00:33.542 --> 00:00:37.002
Open Life Data and choose Repairable Fleet…

6
00:00:37.208 --> 00:00:42.228
In the form Repairable Fleet, Machine ID column stays at A.

7
00:00:42.228 --> 00:00:44.278
Set Event age column to B,

8
00:00:44.333 --> 00:00:47.613
and Observation-end column to C.

9
00:00:48.417 --> 00:00:55.027
Model stays at MCF + power law: the mean cumulative function, with a power-law fit.

10
00:00:55.027 --> 00:00:56.617
Press OK — Calculate.

11
00:00:57.083 --> 00:01:05.843
In RESULTS, the pooled shape β is 3.5968, well above 1: the repairs are bunching up as the gearbox ages.

12
00:01:06.208 --> 00:01:11.888
The Trend verdict reads deteriorating (growth test p = 0.028).

13
00:01:12.000 --> 00:01:27.890
Two tests that assume no model agree: the Laplace test gives u = 2.2131 with p = 0.0269, and the Anderson-Darling test gives p = 0.0337.

14
00:01:28.167 --> 00:01:34.387
The instantaneous mean time between failures, MTBF, is now only 6.2 hours.

15
00:01:34.833 --> 00:01:43.063
The staircase counts the repairs as the gearbox ages, and the dashed power-law curve bends upward: repairs come faster and faster.

16
00:01:43.542 --> 00:01:48.612
On the total-time-on-test plot, the curve lies above the diagonal.

17
00:01:48.612 --> 00:01:51.022
That too says the gearbox is wearing out.

18
00:01:51.167 --> 00:01:54.727
Now the second question, how good are the repairs?

19
00:01:54.727 --> 00:01:59.437
Four boiler feed pumps were each watched for 5,000 operating hours.

20
00:01:59.542 --> 00:02:05.652
Their failure log has one row per failure: the pump, and the hours since its previous failure.

21
00:02:05.652 --> 00:02:08.452
Open File and choose Import Data (Excel / CSV / JSON)…

22
00:02:08.917 --> 00:02:13.697
In Documents, choose Example-1.csv and press Open.

23
00:02:15.000 --> 00:02:18.710
The wizard warns: The current worksheet will be replaced.

24
00:02:18.710 --> 00:02:24.190
Set Unit to Text / ID, because a pump's number is its name, not a quantity.

25
00:02:24.625 --> 00:02:25.935
Press OK.

26
00:02:26.708 --> 00:02:30.378
The worksheet now holds the 40 failures of the 4 pumps.

27
00:02:30.792 --> 00:02:35.762
Open Life Data, then Repairable Systems, and choose GRP / Kijima…

28
00:02:36.208 --> 00:02:45.418
GRP stands for generalised renewal process: a Weibull hazard on a virtual age that each repair resets in part.

29
00:02:45.418 --> 00:02:53.518
The restoration factor q says how much of the age a repair leaves: 0 is as good as new, and 1 is as bad as before.

30
00:02:53.917 --> 00:02:55.817
Set Event time column to B,

31
00:02:56.625 --> 00:02:59.115
Entry to Gap since the previous event,

32
00:03:00.292 --> 00:03:04.412
and type 5000 into Observation end (fleet).

33
00:03:05.333 --> 00:03:11.413
Kijima type stays at Type I (V ← V + q·X): a repair takes back part of the age gained since the last repair.

34
00:03:11.413 --> 00:03:13.213
Press OK — Calculate.

35
00:03:13.667 --> 00:03:23.357
In RESULTS, the shape β is 1.8975 and the scale η is 785.56 hours.

36
00:03:23.500 --> 00:03:34.860
The restoration q̂ is 0.1362, with a 95% profile interval from 0.015 to 0.593.

37
00:03:35.042 --> 00:03:47.182
Both extremes are rejected: against full renewal the likelihood-ratio test gives p 0.0209, and against minimal repair p 0.0255.

38
00:03:47.542 --> 00:03:53.332
So a repair takes back most of the age a pump gained since its last failure.

39
00:03:53.332 --> 00:03:58.742
ReliaEn says it in words: Repairs preserve about 86 % of the unit’s health.

40
00:03:59.208 --> 00:04:02.658
The profile plot shows the likelihood over q.

41
00:04:02.658 --> 00:04:06.868
The shaded band is the interval, and q̂ sits near its left end.

42
00:04:07.333 --> 00:04:15.683
Below it, the fit check: the fitted model follows the 40 observed failures, while renewal falls short and minimal repair overshoots.

43
00:04:16.125 --> 00:04:27.515
In numbers, the fitted model expects 40.28 failures against 40 observed, renewal expects 27.29, and minimal repair 134.04.

44
00:04:27.958 --> 00:04:42.868
The forward card is for pumps starting new: a pump starting new is expected to fail 22.93 times in the next 8,760 hours, one year, and the 4 new pumps 91.71 times.

45
00:04:43.333 --> 00:04:48.463
So the gearbox is wearing out, and the pumps' repairs are good but not perfect.

46
00:04:48.463 --> 00:04:53.123
The next video builds a reliability block diagram and evaluates system reliability.