Reliability Data Analysis

1 Dr. Monalisa Sarma

Subir Chowdhury School of Quality and Reliability

IIT Kharagpur
2 Introduction
 Reliability analysis requires selection of appropriate reliability model.

 Selection of appropriate reliability model needs collection and analysis of failure

data.

 Failure data can be collected either from the field, that is field data, or from
laboratory, that is test data.

 Best approach for reliability evaluation is from field data, however it has its own
limitations.

 The most practical and acceptable alternative is data generation in laboratories by

conducting life test.
3 Field Data vs Test Data
Field data Test data
Time to failure of each unit cannot be obtained Possible
precisely
Data like number of failures in 1st month, 2nd month, Data is exact
etc. is possible. But this grouping results in loss of More information is obtained
information
Failure mechanism is not recorded in most cares Failure mechanism can be recorded
Takes lot of time. Hence design modification is Faster
delayed
Larger sample size Smaller sample size

Exact field environment gives correct picture Simulation of exact field environment is very difficult,
in some cases impossible
Quality of maintenance and operation is poor Better quality of maintenance and operation
(customers and field crews)
4 Objective of Life Test
1. To study the exact behavior of the item under the normal working environment

2. To evaluate MTBF, reliability, hazard rate, etc. of the system/components

3. To generate early failures so that faulty items (due to bad quality control, material,
manufacturing, etc.) are removed before actual usage.

4. To generate data and evaluate life.

5. To study the behavior of items under various types of loads—thermal, shock, fatigue, bumps
etc.

6. To compare items supplied by different vendors.

7. To generate data on new components

8. To study and find out various failure modes.

5 Assumptions of Life Testing

Life tests are conducted under certain assumptions such as:

• Simulated environments represents the actual field environment.

• Test samples are randomly selected

• Failures occur randomly.

• All failures have some associated reason.

• The test item correctly represents the lot produced.

6 Life Testing
 Test data can be obtained only if we perform life tests on several identical
units under identical environmental conditions and keep observing the
failures of the units at specified points of time.

 As the life test progresses under these controlled conditions, the number of
survivals goes on decreasing and eventually we would be left with no
survivors at all, if we continue this test for a very long time.

 Failure data along with the conditions in which a failure data has been
collected must be fully known before we can predict reliability of a unit.
7 Life Testing
 The general approach of reliability analysis is to fit failure data to a theoretical distribution. This
is also known as parametric data analysis

 Nonparametric analysis allows the analyst to characterize life data without assuming an
underlying distribution.

 In nonparametric analysis, failure distribution, reliability distribution, hazard rate is derived

directly from failure data.

 This can be advantageous because it avoids some of the dangers inherent in an analysis where an
inappropriate distribution is assumed.

 However, the analysis is limited to reliability estimates only for the failure times in the data set,
thus making it impossible to make reliability predictions outside the range of data values.
8 Data Collection
 Failure data is likely to be collected under variety of differing circumstances varying from
carefully controlled lab experiments to data from field.

 Lab Data: More precise, more informative per sample (physics of failure, mechanism of
failure occurrence etc…)

 Field data: represents actual environmental loading conditions, i.e., temperature, humidity,
dirt etc., which might be difficult to simulate in labs.
9 Failure Data Classification
 Based on the source of data and the condition of data collection failure data can be categorized as
follows:

 Operational versus test-generated data

 Grouped versus ungrouped data
 Large samples versus small samples
 Complete versus censored data
10 Data Classification

• Complete
• Singly Censored data
 Censored on left Complete Data
 Censored on right
o Type I censoring (time)
o Type II censoring (Failure)
• Multiply censored data
11 Singly Censored-Left (Left Censored Data)
Singly Censored Data:
• All the units have the same test time, and the test
is concluded before all the units have failed

Censored on Left:
• A failure time is only known to be before a certain time.
• Exact time to failure is not known but it falls within a
known time range
• This is identical to interval censored data in which the
starting time for the interval is zero.
• We may know that a certain unit failed sometime
before 100 hours but not exactly when. It could have
failed any time between 0 and 100 hours)
12 Singly Censored-Right (Right Censored Data)
Censored on Right:

• The term "right censored" implies that the event of

interest, i.e., the time-to-failure, is to the right of our data
point.
• Data sets are composed of units that did not fail.
• Failure times of some of the units are not known or test
may be discontinued
• In other words, if the units were to keep on operating,
the failure would occur at some time after our data point
(or to the right on the time scale).
13 Censoring Type-I and Type-II
Type-I:
• Testing is terminated after a fixed length of time has elapsed.

Type-II:
• Testing is terminated after a fixed number of failures, r, has
occurred.
• The test time is then giver tr , the failure time of the rth failure.
Multiply Censored Data:
• Test times or operating time differ among the
censored (removed but operating) units.
• Censored units are removed at various times from
the sample, or units have gone into service at
different times.
 Empirical Methods for Ungrouped Complete Data
14

Consider n ordered failure times t1, t2, … , tn, where ti ti +1 , in a random sample.
• A possible estimate of cumulative failure distribution is
... (1)
 This however tends to underestimate the component reliability.
• Second estimate of cumulative failure distribution is
... (2)
 This is considering a mean estimation.
• Third estimate of cumulative failure distribution is
... (3)
 This is used as an approximation for the median position.
15 Assessment of CDF
Determine the cumulative failure distribution estimate for the given sample of eight failures:

Ranking Time to fail

1 51 0.125 0.111 0.083

2 63 0.250 0.222 0.202
3 70 0.375 0.333 0.321
4 82 0.500 0.444 0.440
5 90 0.625 0.555 0.560
6 96 0.750 0.666 0.679
7 105 0.875 0.777 0.798
8 112 1.000 0.888 0.917
16 Calculation of Descriptive Statistics

, for ti<t<ti+1
17
18 Tutorial

Given the following 10 failure times in hours, estimate R(t), F(t), f(t) and .
15.4, 18.9, 20.1, 24.5 29.3, 33.9, 48.2, 54.7, 72.0, 86.1
19 Grouped Complete Data
• Failure times are placed into time intervals

• Original values are no longer retained

• Sample size is usually larger for grouped data

• Consider n1,n2,… nk be the number of units having survived at ordered time t1,t2,…tk respectively

• Estimate of
(ti) = for i=1,2,…k
• Where n is the number of units at risk at the start of the test
20 Calculation of Descriptive Statistics
(ti) = - for ti < t < ti+1

(ti) =

= for ti < t < ti+1

21 MTTF
• MTTF is estimated on the basis of the midpoint of each interval

• =
where = , =n
• Similarly, for repair time data,

(ti)= 1-
• Where ni is the number of observations exceeding ti

• (t) is the repair cumulative distribution function

22 Tutorial
Seventy compressors are observed at 5-month interval with the following number of failures: 3, 7,
8,9,13,18,12. Estimate , , , and and determine the sample mean time to failure and sample standard
deviation.
23