You are on page 1of 10

School of Engineering and Information Technology

ASSESSMENT COVER SHEET


Student Name

Mark Cruickshank

Student ID

S232953

Assessment Title

Lab 3

Unit Number and Title

ENG432

Lecturer/Tutor

John Montagner

Date Submitted

16/5/2015

Date Received
Office use only
KEEP A COPY
Please be sure to make a copy of your work. If you have submitted assessment work electronically make
sure you have a backup copy.
PLAGIARISM
Plagiarism is the presentation of the work of another without acknowledgement. Students may use a limited
amount of information and ideas expressed by others but this use must be identified by appropriate
referencing.
CONSEQUENCES OF PLAGIARISM
Plagiarism is misconduct as defined under the Student Conduct By-Laws. The penalties associated with
plagiarism are designed to impose sanctions on offenders that reflect the seriousness of the Universitys
commitment to academic integrity.

I declare that all material in this assessment is my own work except where there is a clear
acknowledgement and reference to the work of others. I have read the Universitys
Academic and Scientific Misconduct Policy and understand its implications.*
http://www.cdu.edu.au/governance/documents/3.3academicandscientificmisconduct.doc.

Signed: Mark Cruickshank

Date: 14/5/2015

* By submitting this assignment and cover sheet electronically, in whatever form you are deemed to have made the declaration set out
above.

Mark Cruickshank S232953 Charles Darwin University 2015


ENG435 Laboratory 3: Machine Condition Monitoring

ENG432 Dynamics of Engineering 2015

Laboratory 3: Machine Condition Monitoring


Lecturer: John Montagner

Mark Cruickshank
S232953 16/05/2015

Mark Cruickshank S232953 Charles Darwin University 2015


ENG435 Laboratory 3: Machine Condition Monitoring

Table of Contents
1. Introduction ......................................................................................................................... 4
2. Objective.............................................................................................................................. 4
3. Resources............................................................................................................................. 4
4. Methods .............................................................................................................................. 4
5. Data ..................................................................................................................................... 6
5.1 Results - bearing no. 1 ........................................................................................................................................................ 6
5.2 Results bearing no. 2 ....................................................................................................................................................... 7
6. Calculations.......................................................................................................................... 7
7. Discussion ............................................................................................................................ 7
7.1 Elements of the data acquisition ................................................................................................................................... 7
7.2 NNFT.......................................................................................................................................................................................... 7
7.3 High value of NNFT ............................................................................................................................................................. 8
7.4 Increasing the sampling rate .......................................................................................................................................... 8
7.5 Shorter sample time ........................................................................................................................................................... 8
7.6 Reasons for why inner face faults are harder to detect then outer face faults ......................................... 9
8. Conclusions .......................................................................................................................... 9
10. Figures & graphs .................................................................................................. Error! Bookmark not defined.

Mark Cruickshank S232953 Charles Darwin University 2015


ENG435 Laboratory 3: Machine Condition Monitoring

1. Introduction
Rolling bearings are a fundamental component that is used in most modern machinery,
which have moving parts. They incorporate an inner race and an outer race that is
separated by a number of rollers or balls. These types of bearings are most suited to
applications with high static loads as the rolling friction is eliminated and replaced with
sliding friction. Over the life cycle of the bearing, they are commonly affected by the
following conditions:

Inner and outer race damage


Ball/roller damage
Cage damage

2. Objective
The objective of this laboratory is to examine the data provide by the lecturer for two
bearings that have been tested under laboratory conditions, these are defined as
Bearing 1 and Bearing 2. We need to compute the given data into the frequency domain
using a Fourier Transform before examining the graphical displays to establish as to
whether the damaged bearings are on the inner or outer races.

3. Resources
Resources for this laboratory have been given; these include the data from the
experiment and the Matlab code.

4. Methods
As the Matlab code was given to every student for the laboratory, we will explain in detail
the procedure from loading the data to obtaining our graphs in the frequency domain.
We were given the data file that contained the acceleration of the two bearings over a
period of time. The first column represented bearing number 1 and the second column
was bearing number 2. The time was a column vector, and finally the data contained the
shaft speed of each bearing.
Step 1
Involved loading the data into Matlab.
Step 2
Determining the shaft speed in rad/sec

Mark Cruickshank S232953 Charles Darwin University 2015


ENG435 Laboratory 3: Machine Condition Monitoring

Step 3
Determining the sample frequency, which is inversely proportional to the time interval.
Step 4
Plotting the acceleration of both bearings responses against time.
Step 5
Taking the shaft RPM of each bearing and dividing this by 60 seconds to obtain the shaft
speed in rad/sec for each bearing.
Step 6
Using the table provided in the practical sheet, we entered the ball pass frequencies for
both inner and outer races for both bearings. Bearing 1 was UC205 and bearing 2 was
NU205E.
Step 7
Creating row vectors called B1H and B2H, these values started from their respective
values as entered in task 6 and increased by these values until a value of 100 for both
bearings for both inner and outer.
Step 8
This was creating the magnitude called B1M and B2M for both bearings for both inner
and outer these were set to a magnitude of 600. The magnitude has no relevance for us
in the part as we are looking at the frequency these inner and outer race vibrations
occur, thus this could have been set to any value.
Step 9
This involved us determining the number of samples that will be used for the FFT. The
Nnextpow2 function is to find the nearest power of the two-sequence length, this enables
the fastest computation, as this needs to be a prime number with the smallest base IE 2.
Step 10
Involved taking the FFT of the acceleration over the length of NFFT number of
samples.
Step 11
We calculated the sample rate, as the sample frequency and the number of samples
was previously determined.

Mark Cruickshank S232953 Charles Darwin University 2015


ENG435 Laboratory 3: Machine Condition Monitoring

Step 12
As we knew the shaft speeds of both bearings and the sample rate (frequency
increment) we were able to obtain the orders of both bearings. We have essentially
divided everything by the shaft speed over the length per bearing.
Step 13
The results for both bearings have been plotted on separate plots. The inner bearing is
represented by x and the outer bearing is represented by +. The magnitudes are
plotted against the order for the respective bearing.

5. Data
Plots were determined as previously stated, these can been seen as per the following:

5.1 Results - bearing no. 1

Mark Cruickshank S232953 Charles Darwin University 2015


ENG435 Laboratory 3: Machine Condition Monitoring

5.2 Results bearing no. 2

As we can see in the above plots, bearing number 1 appears to be graphically well
represented, however the graphical display of bearing number 2 appears to behave
considerably different.

6. Calculations
Calculations can be seen in the attached Matlab m-file.

7. Discussion
7.1 Elements of the data acquisition
Please refer to 4. Methods.

7.2 NNFT
I believe it would be more appropriate to have the NNFT greater then the number of
samples when the number of samples is an odd number, as the FFT needs to be of a
prime number to ensure computation time is limited. This is because with an odd number
of samples this will skip over the Nyquist frequency, which is 50% of the sample
frequency, and this will also skip to zero as well.

Mark Cruickshank S232953 Charles Darwin University 2015


ENG435 Laboratory 3: Machine Condition Monitoring

7.3 High value of NNFT


An NFFT can have a higher value then 2^n, however this will affect the computation time
of the discrete Fourier Transform, as this will result in an increased time to compute. In
Matlab the FFT (Fast Fourier Transform function) is optimized when the sizes are of real
prime numbers, therefore we can increase 2^n where n = 2,4,6,ni+1 (where ni is equal
to one) where this would be ideal for this function to compute, however the computation
time to perform this calculation would dramatically increase, thus by choosing the lowest
prime number (IE the lowest base) we would eliminate computation time and the number
would be ideal for the FFT function.

7.4 Increasing the sampling rate


Increasing the sample rate, we are effectively increasing the number of samples over a
total sample length. Provided the sample frequency remains constant, we are effectively
increasing the sample length.
This can be seen from the below equation:
tn =

n
Fs

Where n is the number of samples, Fs is the sample frequency and the total sample
length is tn, which is expressed in seconds. Therefore by increasing the total sample
length this would effectively increase the computation time as the sample length would
increase as the number of samples increase. Also as the number of samples increase
this would result in more accurate data, the data between points would be of a lesser
length/distance.
In some instances this could have a negative effect, especially if there is noise in the
system and we are trying to analyze the data, this would result in a too high resolution
and the data may be unusable, from a graphical perspective, thus if this were the case,
we would want to reduce the number of samples taken to enable a more readable
resolution.

7.5 Shorter sample time


Decreasing the sample time would result in a much higher sample frequency, as the
sample time is inversely proportional to the sample frequency. Provided the number of
samples remains constant, the frequency increment would also increase as well as the
total sample length (sec.). Computation time would also increase, as this would involve
more computations, essentially the sampling rate would increase provided the number of
samples remains constant.

Mark Cruickshank S232953 Charles Darwin University 2015


ENG435 Laboratory 3: Machine Condition Monitoring


7.6 Reasons for why inner face faults are harder to detect then outer face faults
Being an external student of this unit, it is quite hard in trying to examine the set-up of
this laboratory, thus I can only assume the instruments used based on set up from other
universities pervious experiments.
I believe there are two main reasons why the inner race defect is considerably harder to
detect then the outer race:

This could be due how the testing apparatus is connected the inner race, the
probes used whilst conducting the experiment. With the inner race the probes
have the tendency to make contact with the rollers, this would cause a significant
amount of vibration to appear in the graphical representation, making it harder to
interpret.

The second reason is that when a bearing wears/breaks down, this generally
occurs on the outer race, therefore the life of the inner race is considerably longer.
Thus, when the inner race has a defect the outer race will be considerably worn
down and the graphical representation would look something like the graph of
bearing number 2, where it is clear that both the inner and outer races are
defective.

8. Conclusions
The results of this experiment for bearing number 1 appear to be relatively satisfactory.
As we can see in the graphical representation, the outer race appears to be damaged.
We can see that on every harmonic the peak of the vibration is precisely where the
frequency for the damaged outer race should be.
On the other hand the results of bearing number 2 appear to be less satisfactory, we can
see there is a large amount of peaks halfway through this experiment this could be due
to a number of reasons. Without physically conducting this experiment it is hard to say
as it may have been contributed by an external source such as noise or vibrations from
someone moving the instruments.
Therefore for bearing number 2 we would conclude that this is damaged due to the outer
race and inner race as the peaks go through nearly all of the relevant areas. Also the
outer race is generally damaged before the inner race, thus if the inner race was
damaged we would expect the outer race to be damaged as well, as it appears in this
case, hence the vibration peaks in shorter frequency increments then in comparison to
bearing number 1.
Overall, this experiment has allowed us to examine bearings solely based on data, by
converting the response in the time domain, using a Fourier Transform to convert this
data into the frequency domain and then finally dividing all the frequencies by the
frequency of the shaft rotation to be able to convert to an Area analysis. The calculations
Mark Cruickshank S232953 Charles Darwin University 2015
ENG435 Laboratory 3: Machine Condition Monitoring

and methods to use this analysis is fairly simple, we can see in bearing number 1 how
accurate this method is in determining the condition of bearings.
Ideally if this experiment were to be conducted again at a latter stage, the following
should be implemented to enable a better understanding and possibly more improved
results:

Having a video of the experiment, we would then be able to determine if any


external sources/events has any impact on data results.
Recording more data over longer periods of time, this would have helped in
bearing number 2 as the intervals of vibration peaks could of been an external
source, by using additional data we could have distinguished if this was the case.

Mark Cruickshank S232953 Charles Darwin University 2015


ENG435 Laboratory 3: Machine Condition Monitoring

10

You might also like