ISE 506: Jet Engine Failure statistics assignment

The jet engine represents the leading edge of technology, advanced manufacturing, quality control, design
evaluation, and extensive testing. This machinery, with its new hardware and systems can achieve very high
standards of reliability. Since the days of deregulation, the aviation industry has considered many methods,
techniques, and procedures to reduce operational costs without hindering safety and reliability standards. A
significant portion of these operational costs is the resource allocated to jet engine repairs. The jet engine has
many failure modes that can cause the removal (failure) characteristics to have a substantial amount of
variation in the expected time-on-wing (TOM). This causes inaccuracies in the jet engine lifetime estimation.
For this reason, the industry needs to be able to accurately forecast engine removals.

The engine removal data was obtained from airline databases over a period of years with a substantial sample
size of removals. In choosing these airlines, different operating environments were selected. The database
includes key information on engine delivery and removal dates. All of the historical data were used to develop
the model, while a more recent update for the following 27 months of the airline's operation was used to
evaluate the effectiveness of this forecasting methodology. Typically, these removals are found upon
inspections before airplanes take off. The data sample consists of 25 jet engine operating that were
monitored for a period of 550 flying hours, See Excel file Assignment Maintenance in Aerospace
Management.xlsx sheet Jet Engine Failure Data. As an example engine #13 has failed for the first time at 213
flying hours then it was repaired. After returning to operation, the engine failed again at 370 flying hours. So
the second time to failure is 370-213=157 flying hours. After returning to operation, the engine did not fail
before the total experiment time 550 flying hours. So, the time till the third failure is larger than 550-370=180
flying hours. Moreover, we shall assume that after repair the engine is returned to as-good-as-new state.

Note: In Excel 2010 the linear regression tool is available in Excel under Data -> Data Analysis. Don’t forget to
enable the Analysis ToolPak Add-in. to do so Click on the File -> Options -> Add-ins. On the bottom left of the
screen select Excel Add-ins and press on Go. A new window will open where you can tick on Analysis ToolPak.
You should find Data Analysis under Data.

Table 1: Sample data of 2 jet engines operating for 550 flying hours
Engine # Cumulative Failure time total failures
13 213 370 2
14 171 332 539 3
1
a) How many time-to-failure data samples do you have in total? Justify your answer.
b) How many censored data samples do you have in total? What kind of censoring they are?
c) Find the value of the Weibull distribution parameters β and η of the jet engine using the linear
regression analysis as explained in the lecture.
( ηt ) , t in flying hours. Our objective here is
β
−
The Weibull distribution has the following form: F ( t )=1−e
to estimate the shape parameter β and scale parameter η using the data. Stepwise procedure to
follow:
1. Put the data in an increasing order of time
2. As we have right censored data we shall follow the method of adjusted rank
 Starting from the lowest failure time, we recompute the rank increment every time we
n+1−i t
encounter a censored data using: rank adjustment = i−1
,
1+n−number of preceedingitems
where n is number of samples, i t is the rank of latest time to failure. Note, if this is the
i−1

first time to failure then i t =0 i−1

 The rank of time to failure t i is : i t =i t +rank increment

i i−1

i t −0.375
 Compute the empirical CDF at time to failure t i as ^
F ( t i )= i

n+ 0.25
 As we are dealing with Weibull distribution, compute ln (t i ) and ln ¿ for all time to failure t i
 Using linear regression with x-values ln (t i ) and y-values ln ¿ and determine the slope (slope
is equal to shape parameter β of Weibull distribution) and y-intercept (

(
η=exp − y−
y−interecept
β
. )
d) Discuss whether the quality of fit based on R square of the regression.
e) Is it effective to introduce a preventive maintenance program for the jet engine and why?
f) Find the reliability of the jet engine at 100 flying hours, MTTF, and B-10 life using the β and η of the
linear regression.
g) Find the value of the Weibull distribution parameters β and η using maximum likelihood estimation.

Stepwise procedure to follow is:

 Set an initial non zero values of the parameters β and η .

 Compute the likelihood of each data point including the censored data:
o If it is right censored data then the likelihood is equal to reliability function at point t i
: R ( t i )=1−F ( t i ) =1−Weibull . dist (t i , β , η, TRUE)
o If it is a time to failure then the likelihood is equal to pdf function at point t i:
f ( t i )=Weibull . dist (t i , β , η , FALSE )
 Compute ln ( likelihood of t i ) for all t i
 Compute the sum of the likelihoods for all t i

2
  Using the solver find the values of β and η that maximizes the sum of the likelihood

h) Compare the findings in d and c. Discuss what method we should use to choose the values of β and η .
i) Find the mean time-to-failure (MTTF) and B-10 life.
j) Find the reliability of the jet engine after an operation of 50, 75, and 100 flying hours.