Scribd
Upload
4 views

Week 8 Answer Sheet

The document discusses an engineering assignment focused on analyzing testing data for reliability growth models. It includes calculations for years of test data, identifies trends in failure rates, and evaluates various reliability models, concluding that the Jelinski-Moranda and Goel Okumoto models are the best fits. The analysis emphasizes the importance of data subsets in reliability projections and highlights differences in predictions between full and subset data.

Uploaded by

Moe Aung
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
4 views

Week 8 Answer Sheet

The document discusses an engineering assignment focused on analyzing testing data for reliability growth models. It includes calculations for years of test data, identifies trends in failure rates, and evaluates various reliability models, concluding that the Jelinski-Moranda and Goel Okumoto models are the best fits. The analysis emphasizes the importance of data subsets in reliability projections and highlights differences in predictions between full and subset data.

Uploaded by

Moe Aung
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 6

Week 8 Assignment

Engineering 202
May 2020
ANSWER SHEET

a. Assuming the time units are in hours, about how many years of testing data is included in
the data set?

Answer: From the “Data and Trend Test table” sub-tab in the first tab, the last data point
(207) has a failure time of 16656 hours. There are 8760 in a year (or 8766 hours account for
leap years). This is aboutg 1.90 years of test data

b. Over what time intervals does the Laplace growth curve show a decreasing trend in the
failure rate?
Answer: From approximately 801 hours (point 25) to 2268 hours (point 51) and from 5002
hours (point) 101 until the end

c. Fit the following: reliability growth models: Jelinski-Moranda, geometric, exponential,


Yamada delayed S-shaped, and Weibull, download the image (or use a screen capture tool
to get it into a file)

NOTE: inclusion of the exponential model in this list was an error. However, see the end of
this answer

Below is a screen shot of both the tool settings (left) and results display (right)

1
Notice how the Delayed S Shape (Yamada) model diverges from the others. This is even
more apparent using the “reliability growth” display where reliability is calculated over a
defined interval based on the mean value

COMMENT: Although my inclusion of the exponential (i.e., constant failure rate) was an
error, it is interesting to analyze using a linear regression. Here is the result using Excel
(where x is actual time). Notice the high correlation coefficient. Using LINEST function, the
slope is the constant failure rate of 0.155 and the mean time to failure (i.e., the time to the
next failure) is 64.4 hours

2
Cum Failures
300
y = 0.0155x
250 R² = 0.9759

200

150

100

50

0
0 2000 4000 6000 8000 10000 12000 14000 16000 18000

d. Which are the two best fit models? Why?


Answer: Both the Jelinski Moranda and Goel Okumoto models are the best fit, having both the
lowest PSSE and AIC values. Note how close the AIC values are relative to the PSSE values.

e. When will the next failure occur according to the two best fit models?
Answer: 206 hours (Goel Okumoto) and 211 hours (Jelinski Moranda)

3
f. Repeat steps c through e for the subset of the data when failure rate trend stays in the
decreasing trend region (i.e., no longer enters the increasing trend reason).

NOTE: The were reports of problems in running the tool in the course forum. The result in
this answer reflects the same anomalous experience – the Weibull result disappeared.

Answer:

(1) Repeat of Step c for subset of data (101 to 208)

(2) Repeat of Step d for subset of data (101 to 208)

4
(3) Repeat of step e for subset of data (101 to 208)

g. What differences do you see in the best fit models, the goodness of fit, and the prediction of
failure times

Answer:

1. The Goel Okumoto and Jelinski Moranda models still have the lowest PSSE values, but
all are much closer together.
2. However, the Delayed S Shape and Goel Okumoto models have the lowest AIC values.
Given the small (1.1%) difference between the AIC values, the PSSE value differences
dominate.
3. The Goel Okumoto and Jelinski Moranda models predict a lower time to next failure
(175 and 182 hours respectively) than in the complete data set (205 and 211 hours
respectively), but the results for the complete data set vs. the subset differ but less than
15% (for reliability prediction, this is good agreement). The Delayed S Shape prediction

5
is also lower for the subset than for the complete data set, and deviates (in this case,
with a much higher value) from the Goel Okumoto and Jelinski Moranda model results

h. What conclusions would you draw from the comparison of the full data and the subset?
Answer: Whether a full or a subset of the data are chosen in this case, the results are fairly
close. However, this is not always the case. In performing reliability growth projections,
exploration of the data by taking subsets and looking at alternative distributions is vital.

You might also like