105 views

Uploaded by Paul V. Pabillon

Attribution Non-Commercial (BY-NC)

- Postresp Um
- User Manual and System Documentation of WAVEWATCH III Version 4.18
- Advanced Processing and Interpretation
- Fan Vibration Specifications
- Solutions Manual Keeler
- Fault%20Analysis%20With%20Vibration[1]
- DMIP Spatial Rain Variability
- Generation of Random Road Profiles.pdf
- Critical Gust Pressures
- Slide - Spectral Analysis
- VAPEPS
- 09-Mounting Resonance Case Study - Vibration Monitoring at Weirton Steel
- fourier transform dr sek frequency analysis - fft
- Directional Correlations of Seismic Response of an Asymmetric Structure 1992
- POSTER.pdf
- Frequently Asked Questions About Enveloping
- Iris Chapter03
- 10_okada_pg73-85
- SSPPTFourierSeriesandFourierTransform.pdf
- Full Text

You are on page 1of 5

fuzzy logic

T. Lindh, J. Ahola, P. Spatenka, A-L Rautiainen

Tuomo.Lindh@lut.fi

Lappeenranta University of Technology

Lappeenranta, FINLAND

Abstract - In this paper, a new automatic analysis method for purpose is not to tune parameters in order to maximise the

the detection of cyclic bearing faults is introduced. The reliability of the system. The fuzzy logic has been used in

method uses a multivariate statistical fault classification machinery vibration analysis by, for example, Mechefske [1] and

and fuzzy logic. Features are extracted from an envelope Lindh [2].

spectrum of the frame acceleration of a motor frame.

Qualitative and quantitative measures of the features are

utilized. II. FEATURE EXTRACTION AND CLASSIFICATION

The features are created from the coefficients of the envelope

Keywords: bearing fault classification spectra calculated from the motor frame acceleration signal. The

expected bearing pass frequencies (the characteristic fault

frequencies) [3] for different fault types are calculated and the

I. INTRODUCTION peaks near to these calculated values are selected as features. The

envelope spectra with selected components are presented in

The research of condition monitoring of electric motors has

Figures 4, 5 and 6. A bearing fault feature vector consists of 16

been wide for several decades. The research and development at

PSD components of a signal envelope at selected frequencies. The

universities and in industry has provided means for the predictive

selected components represent the characteristic fault frequencies

condition monitoring. Many different devices and systems are

and their three nearest harmonic frequencies for all the following

being developed and are widely used in industry, transportation

faults: outer race fault o, inner race fault i, rolling element fault b

and in civil engineering. Recently, in order to achieve the

and cage fault c. These PSD components X form a 16-

automatic analysis of faults, many methods have been developed

dimensional feature vector as follows

and reported in scientific arenas. In this paper, the guidelines for

the statistical fault classification of the envelope spectrum x = [ X ( f o ) X ( 2 f o ) X (3 f o ) X ( 4 f o ) X ( f i ) X ( 2 f i ) X ( 3 f i ) X ( 4 f i ) X ( f b ) X ( 2 f b )

components using the Mahalanobis distance calculation and X ( 3 f b ) X ( 4 f b ) X ( f c ) X ( 2 f c ) X ( 3 f c ) X (4 f c )]

deterministic classification with multi- valued fuzzy logic are

(1)

presented. The main idea is the following: As humans, the

The common measure for the statistical distance between

algorithm searches patterns in the spectrum, analyses the

samples is the Mahalanobis distance

magnitude of peaks found and estimates the probability of a false

triggering with qualitative factors of extracted features of the

r 2 = (x − m x ) ' C x−1 (x − m x ) , (2)

spectrum.

The Mahalanobis metrics is superior in the cases where the where r is the Mahalanobis distance between feature vector x, Cx

features are dependent on each other as is the case with the is a covarariance matrix containing variances between features and

spectrum components of an acceleration signal. Based on the mx is a mean vector of features [4].

distance calculations and a minimum distance classifier, possibly

with a trigger level for the distance from the healthy case, a fully The proposed system consists of the following training phase

automatic system can be constructed. The main problems are, actions

firstly, how the prototypes of faulty spectra are created and, • The creation of artificial spectra for the cases of an

secondly, how the trigger levels from a healthy case representing outer race defect, an inner race defect, a cage defect and

the abnormal situation are selected. In addition, the rotational a rolling element defect.

speed of the motor must be taken into account in the system by • The creation of an over all acceleration level feature for

assuming known operation conditions or by rotational speed low, normal, high and very high levels of vibration.

adaptation of the spectrum. Being so that only the healthy case • Calculation of a normal case prototype vector using

spectrum is available, a-priori-knowledge of the faulty case several measurements from a healthy case.

patterns must be included in the system artificially. Another And of test phase actions using statistical distance calculation

possibility is to use broken case spectra obtained from another • The minimum distance classification using the

motor with a broken bearing. calculation of distance between a test vector and

The advantages of the fuzzy logic approach include the healthy and all broken case prototype vectors.

possibility to change the linguistic rules into decisions by copying • Classification of abnormal versus normal using trigger

the procedure and thinking of a human analyser. The rules that level distances from a healthy case vector.

include uncertainty and inaccuracy are changed into numbers And test phase actions using fuzzy logic including

describing the severity or the probability of a fault. The rules and • Calculation of memberships in the output sets that

membership functions can be tuned so that the sensitivity of the describe the degree of fault for every fault type under

system is good. On the other hand, when the fuzzy logic requires consideration.

tuning of parameters, it is possible that it is tuned to find faults • Calculation of memberships in the output sets that

from test data, but the logic remains case specific. In this paper, describe the correct classification probability of the

the purpose of introducing fuzzy logic is to demonstrate its fault for every fault type under consideration.

possibilities as a part of an automatic fault classification; the

• As a combination of the previous memberships, the covariance information).

selection between a healthy and faulty case with fault

type information and instructions for the action. In order to demonstrate the applicability of fuzzy logic two

The values of prototype vectors are selected to represent 40 different models are created. The first model (A) uses the 16

different levels of failure. The bearing impulse increases the dimensional feature vectors presented. Features of each fault type

vibration at characteristic frequencies of the motor frame. are tested separately. Therefore, the inputs include the

Therefore the impulse response of the motor frame to the bearing magnitudes of the envelope spectrum components at the expected

impulses is a series of resonating waves. The envelope of this ball pass frequency and their three nearest harmonic frequencies

vibration is not sinusoidal. Therefore, the harmonics of the separately for all the following faults: outer race fault o, inner race

bearing pass frequency are found in the spectrum but are fault i, rolling element fault b and cage fault c. The input sets

attenuated compared to the bearing pass frequency. The values of consist of the sets low, medium, high and very high. The

the prototype vectors are selected bearing this in mind. The membership functions are triangle- shaped. In the fuzzy rules, the

prototype vectors are done without testing their discrimination minimum method is used for and- rules and the maximum method

power using a big value for the characteristic frequency and then for the or- rules. The weakness of these methods is that they do

smaller values to the harmonics in descending order. The features not take into account all of the input memberships. These

not influenced by a certain fault are set to one, the bearing pass methods are only used because of their simplicity. The output

frequency component to five, the first harmonic to four, the sets are healthy, suspicious, broken and seriously damaged

second to three and the third to two. Then the values are altered describing the degree of the fault. The centroid of membership

to represent 40 different levels of damage and 20 % random function for the set healthy is at 0.1, for the set suspicious is at

variation is added to the values. A set of prototype vectors for 0.2, for the set broken is at 0.5. For the set seriously damaged the

inner race fault is presented in Figure 1. membership increases linearly from the value 0.4.

The second model introduces new parameters in order to have

9

more reliable results. It can be supposed that the risk of a wrong

8 decision is minimised when the degree of the fault is evaluated as

well as the probability of a specific fault. Then the output of logic

7

describes better the risk of the continuing the use of a motor

6

without service. This risk can be described with the multiplication

of the degree of the fault and the qualitative probability of a

prototype vector value

5

certain fault. The probability is calculated as an error between the

4

estimated and measured quantities. Then the following rules are

used:

3

2

• If the magnitudes of the outputs (for different faults) of

previous logic (model A) are high, the degree of the

1

fault is high.

0

• If the RMS amplitude of the vibration is high and the

0 2 4 6 8 10 12 14 16

features magnitudes of the outputs (for different faults) of

previous logic are low, the condition of the machine is

Fig.1. The prototype vectors for the inner race fault class. abnormal.

• If there are high peaks in the envelope spectrum at non-

Formation of a test vector. From the envelope spectrum of a predicted frequencies the condition is classified as other

real measurement (on running 15 kW induction motor) the twelve abnormal situation.

components in the neighbourhood of the calculated characteristic • If the distance to a certain fault is low using the

frequency (frequency resolution of PSD ∆f ˜ 0.2 Hz) were statistical classification or if the spectrum peaks are

selected for pre-processing, in which, the maximum value of these found near to predicted frequencies, the probability of

was selected as a feature. Envelope spectra for the outer race fault the fault is great.

and inner race fault were formed using actual measurements. In • If there are peaks near to the harmonic frequencies of

the case of other faults the spectra of the healthy case were the supposed bearing pass peak frequency or if there

altered artificially near characteristic frequencies. are rotational frequency side bands found at predicted

Classification. In the tests, the distances between the test frequencies, the probability of the fault is great.

vector and the vectors representing any of the fault types (broken The outputs of previous rules can be used in new rules such

classes) and of the healthy case were calculated according to as

Equation 2. The jacknified Mahalanobis distance calculation, • If the probability of a certain fault is great and the

where the vector under test is not included into calculation of degree of the fault is high, the output is act fast.

covariance and mean, was used. Then a minimum distance was • If the probability of a certain fault is great and the

selected and the test vector was classified to belong to a degree of the fault is not small, the output is act.

corresponding class. In the second test, the distances were • If the probability of a certain fault is low and the degree

calculated with reduced feature spaces so that only the features of the fault is high, the output is act.

representing a certain fault class were selected to a feature vector. • If the probability of a certain fault is low and the degree

The distances between the four test vectors and the of the fault is small, the output is wait.

corresponding fault prototypes as well as healthy class • If the non-normal situation holds then the output is

prototypes were calculated separately. If in one or more warning.

calculations the distance to a faulty class was less than to the

healthy class the corresponding fault was classified. This The features that are extracted from the envelope spectrum

procedure makes it possible to recognise multiple simultaneous are illustrated in Figure 2 and the flow chart of the

fault cases. On the other hand, the possibility of an incorrect classification system is illustrated in Figure 3.

estimate is greater than in previous test because the values not

changing due to a fault are not taken into account (weaker

-3 -3

x 10 x 10

6 6

5 4

0 100 200 300 400 500 600

f [Hz]

-3

x 10

2 5

cage

race race element

1

3

2 test

vector

0

0 100 200 300 400 500 600

f [Hz] 1

0

0 2 4 6 8 10 12 14 16

Fig.2. The features extracted from the envelope spectrum. features

Signal Envelope spectrum Feature extraction Statistical distance Fig.4. The envelope spectrum of a healthy motor is illustrated in

formation calculation

the upper figure. The spectrum components that are

selected for the test feature vector are marked with *. This

vector is illustrated in the lower figure with the prototype

Feature extraction Minimum distance

classifier vectors of the outer race fault (∧ ), the inner race fault (∨ ),

RMS, peaks, etc. the ball spin fault (o ) and the cage fault ( ). The vector of

Classification

data

healthy case is a unity vector.

-3

Fuzzy logics x 10

6

Result 4

III. RESULTS 1

0

0 100 200 300 400 500 600

Results. The envelope spectrum of a healthy motor is illustrated f [Hz]

x 10

-3

are selected for the test feature vector are marked with *. This 5 Inner rolling

Outer

vector is illustrated in the lower figure with the (mean) prototype 4

race

race element cage

the outer race fault and inner race fault are presented in Figures 5 2

vector

Table 1 presents the classification results using a 16- dimensional 0

0 2 4 6 8 10 12 14 16

feature space, covering all four fault types. Table 2 presents the features

indicate only correct classification results. The statistical distance

between healthy and broken cases is bigger when four- Fig.5. The envelope spectrum of a motor with an outer race

dimensional feature space is used. On the other hand, there is a faulted bearing is illustrated in the upper figure. The

bigger risk of misclassification if the shape of the test feature spectrum components that are selected for the test feature

vector changes. It is important to bear in mind, firstly, that the vector are marked with *. This vector is illustrated in the

correct classification was obtained without any tuning of the lower figure with the prototype vectors of the outer race

prototype vectors, and secondly, that there can be many other fault (∧), the inner race fault (∨ ), the ball spin fault (o ) and

fault modes that were not taken into account and much more the cage fault ( ). The vector of healthy case is a unity

research work should be done with various fault types and vector.

motors before jumping into conclusions that generalize the result

obtained with these tests.

The classification results using the fuzzy logic for the degree of

fault are presented in Table 3. All of the broken cases were

classified to the class broken and the healthy case was classified

to the classes healthy or suspicious. In the Table 4 the outputs of

the fuzzy logic representing the fault probability are presented.

These values are multiplied with the fault degree outputs forming

the memberships of the output act. The results clearly

demonstrate the advantages of introducing both quantitative and

qualitative features in the calculation as well as combining the

statistical classification and fuzzy logic.

x 10

-3 The problem of prototype vector creation was solved here in the

6

most straightforward way. The characteristic bearing frequency

5

peaks and their three nearest harmonics were selected as features.

4

In some cases, the harmonics can be attenuated (sinusoidal

3

envelope). Then the risk of misclassification is obvious. In this,

2

the fuzzy logic will correct this if the amplitude of the peak at the

1

0

bearing pass frequency is high.

0 100 200 300 400 500 600

f [Hz]

x 10

-3 Trigger levels are not required when minimum classifier is used.

4

Outer Inner

However, for fuzzy logic requires sets that have limits. The

rolling cage

3 race race element trigger levels, on the other hand are not as strict as with Boolean

logic and therefore a priori knowledge of the behavior of different

2 test

vector fault types can be used. However, the limits are case specific and

1 are the causes for the possible misclassification.

0

0 2 4 6 8 10 12 14 16

features The risk of a misclassification increases in situations where there

exists a high peak value in the envelope spectrum at some

Fig.6. The envelope spectrum of a motor with an inner race characteristic bearing fault frequency and is not caused by a

faulted bearing is illustrated in the upper figure and features are bearing fault. This kind of situation is found in Figure 6 where the

marked similarly as in the previous figure. peak value caused by vibration at a rotational frequency is

misinterpreted to be the one caused by a cage fault. In this case,

Table 1. Classification results using 16- dimensional feature space however, the covariance between other features of cage fault

covering all four fault types. The bold cases are selected with the remains low so that the misclassification was avoided. On the

minimum distance classifier. other hand, this case also illustrates the usefulness of using the

healthy outer race inner race ball spin cage Mahalanobis metrics in the distance calculations. However, with

healthy 0.6981 1.3607 2.7905 1.6052 2.3892 the reduced feature space of Table 2 the difference between the

outer race 5.3817 1.2018 7.0394 8.7226 7.5465

distance to inner race fault and the distance to cage fault is quite

inner race 2.9616 3.2331 1.0653 2.7675 2.7684

ball spin 6.5451 5.3318 8.2085 1.0579 6.7352 small because of the weaker covariance information than with the

cage 0.7758 1.4425 0.8656 1.7417 0.1666 classification with 16 dimensional feature space of Table 1. In

addition, this example demonstrates the usefulness of using the

Table 2. Classification results using four dimensional feature deterministic qualitative measures and fuzzy logic: In this case,

space. The distance between test vector and any fault prototypes the significance of that peak was reduced because of the missing

are calculated separately. The bold cases are selected with the harmonics and rotational side bands.

minimum distance classifier.

test data distance to outer race inner race ball spin cage V. CONCLUSIONS

healthy healthy 0.1767 0.148 0.1107 0.138

broken 0.781 4.235 1.2339 1.3203 The method where the Mahalanobis distance based statistical

outer race healthy 3.9542 0.2946 0.2454 0.2164 classification results were used as one input in fuzzy logic was

broken 0.0394 6.5133 1.4492 1.4779

introduced. This input and other qualitative features of the

inner race healthy 0.556 2.4608 0.441 2.0332

broken 1.385 0.3182 1.6345 2.656 spectrum were taken into account in the fuzzy logic which

ball spin healthy 0.2327 0.3109 4.4202 0.1176 evaluated the probability of a certain fault. The other fuzzy logic

broken 0.9328 4.1784 0.2219 1.7265 evaluated the degree of certain fault and by combining the outputs

cage healthy 0.2827 0.3915 0.2275 4.0238 of these logics, the final suggestion of the state and required action

broken 1.0563 5.0148 1.7709 0.073 were given. Based on the tests, this approach improves the

reliability of the fault analysis when the reliability of the analysis

Table 3. Classification results using simple fuzzy logic for the means the minimisation of the wrong action, continuing the use or

determination of fault degree. repairing the motor. In many researches the neural network has

Fault degree been used as a pre- processor for the fuzzy logic or as post-

test data outer race inner race ball spin cage processor for time-frequency representations of signals [5,6]. The

healthy 0.18 0.17 0.17 0.18

outer race 0.47 0.42 0.18 0.10 presented approach is alternative to that approach. The statistical

inner race 0.03 0.36 0.03 0.32 Mahalanobis distance based classifier as used in this paper may

ball spin 0.18 0.17 0.50 0.18

cage 0.18 0.16 0.17 0.50

give equal results to the neural network. However, due to the

exact input output relationship, the controllability of the

classification is better when the Mahalanobis distance classifier is

Table 4. Classification results using fuzzy logic that estimates the used.

fault degree as well as the probability of the faults. Action can be The test data was rather limited (real data from outer and

formed with fuzzy logic or multiplication of the fields of Table 3 inner race faults, simulations for other faults) and jumping to

with the probabilities of this table. The actions are selected with certain conclusions should not be made before the tests are done

trigger levels. with many motors with different bearing faults. On the other

Fault probability Action

test data outer race inner race ball spin cage outer race inner race ball spin cage hand, by discussing the possibilities of an automatic fault

healthy 0.18 0.17 0.16 0.17 0.03 0.03 0.03 0.03

outer race 0.49 0.3 0.3 0.2 0.23 0.13 0.05 0.02

analysis, the guidelines for the future work are given.

inner race 0.32 0.35 0.17 0.16 0.01 0.13 0.01 0.05 Furthermore, the presented method relies on the known

ball spin 0.32 0.17 0.17 0.16 0.06 0.03 0.09 0.03

cage 0.32 0.17 0.16 0.58 0.06 0.03 0.03 0.29 properties of envelope spectrum peaks (for example [7]), if the

peaks appear, the method would classify properly any fault

types represented in the feature vectors.

Problems.

References

fuzzy logic”. Mechanical systems and signal processing, 1998,

No.12, pp. 855-862

[2] Tuomo Lindh, On the condition monitoring of induction

machines. Dissertation, Acta Universitatis Lappeenrantaensis

174, Lappeenranta University of Technology, Finland 2003,

ISBN 951-764-841-3.

[3] Schiltz, R. L., “Forcing frequency identification of rolling

element bearings”, Sound and vibration, pp. 16-19, May 1990.

[4] P. C. Mahalanobis, Proc. Natl. Institute of Science of India, 2,

49, 1936.

[5] Staszewski W. J., Worden K., Tomlison G. R., “Time-

Frequency Analysis in Gearbox Fault Detection Using the

Wigner-Ville Distribution and Pattern Recognition”,

Mechanical systems and signal processing, 1997, No.11,

pp. 673-692

[6] Paya P. A., Esat I. I. “Artificial neural network based

fault diagnostics of rotating machinery using Wavelet-

transforms as a preprocessor”, Mechanical systems and

signal processing, 1997, No.11, pp. 751-765

[7] Wang Y. F., Harrap M. J., “Condition monitoring of ball

bearings using envelope autocorrelation technique”.

Machine Vibration, 1996, No.5, pp. 34-44

- Postresp UmUploaded byGuilherme Leal
- User Manual and System Documentation of WAVEWATCH III Version 4.18Uploaded byHectortitohidalgo
- Advanced Processing and InterpretationUploaded byShashank Sinha
- Fan Vibration SpecificationsUploaded byRomie Cubal
- Solutions Manual KeelerUploaded byvivianemnasc
- Fault%20Analysis%20With%20Vibration[1]Uploaded byAhmed Hamdy
- DMIP Spatial Rain VariabilityUploaded byJuan Carlos Rodriguez
- Generation of Random Road Profiles.pdfUploaded byLea Eleuteri
- Critical Gust PressuresUploaded byManish
- Slide - Spectral AnalysisUploaded byTongYueHern
- VAPEPSUploaded bypws111
- 09-Mounting Resonance Case Study - Vibration Monitoring at Weirton SteelUploaded bySyed Hasan
- fourier transform dr sek frequency analysis - fftUploaded byapi-301025469
- Directional Correlations of Seismic Response of an Asymmetric Structure 1992Uploaded bycisco
- POSTER.pdfUploaded bychris
- Frequently Asked Questions About EnvelopingUploaded byerikmaxramos72
- Iris Chapter03Uploaded byTON
- 10_okada_pg73-85Uploaded byOerip Nurwijayanto Prabowo
- SSPPTFourierSeriesandFourierTransform.pdfUploaded bySoundarya Svs
- Full TextUploaded byElkin Arregoces
- Stylianou_VC.pdfUploaded bysankartheraak
- Review of Random Loading Concepts IIUploaded bystefan
- 01-ProbaCondUploaded byOtman Chakkor
- Proc. R. Soc. Lond. a 1998 Huang 903 95Uploaded byCasia Nursyifa
- UnitCombustor_2012Uploaded bybkm33
- 58931.pdfUploaded bybayman
- A Simple Algorithm for Band-limited Impedance InveUploaded byghassen laouini
- ADSPUploaded bybinukiruba
- Experimental investigation into the dynamic response of two DOF tuned deck simulator for shock qualification of shipboard systems.pdfUploaded byFernando Raúl LADINO
- Implicit Analyses in LS-DYNAUploaded byTiep Tran Quang

- Single Degree of FreedomUploaded bymm0493440
- 1.docxUploaded bydhanashree
- hade kanUploaded byaseprambo
- Cisco OutputUploaded byFernando Salvador Gradaille
- CHDK QuickStartUserGuide-V1.7.5_22-12-09_Booklet_3Uploaded bydelaneyluke
- Tugas Material NanoUploaded byCHRISTOFELHERITULLY
- Cavitation GUIDEUploaded byLindsey Patrick
- ErnestUploaded byrosnet
- Continuous-Time Finance: Lecture NotesUploaded byChunrong Wang
- Cylinder HeadsUploaded bypistonbreaker
- EE6201 MAY 2016 answer key.pdfUploaded bynaveeth11
- Haulotte H 41 PXUploaded byAthox Qhurien
- 25 - Battery Sizing & DischarBattery Sizing & Discharge.pdfgeUploaded byEgyptman Jan
- Catálogo TécnicoUploaded byJulio Razeck
- MEG4 Mooring Line Base Design CertificateUploaded byvdeviv
- Chap2Uploaded byMingdreamer
- Configuracion y O&M BTS NokiaUploaded byjcardenas55
- NietzscheAndObscurutyOf Heraclutus.pdfUploaded byTornado Intempestivo
- Catalogo de Estimulacion SLBUploaded bypetantelo
- Unix ArchitectureUploaded byAnkit Kumar
- GoalProg2Uploaded bybernabarillo
- TCVN 2622-1995 Fire Prevention Standards (en)Uploaded byNguyễn Văn Thành
- Routing Protocols for Wireless Sensor Networks (WSNs)Uploaded byAnonymous uVZ8dZg
- Earley AlgorithmUploaded bytriafffii
- Manual T3DS SoftwareUploaded byEgoitz
- Chain Surveying PptUploaded bySanjeev Rao
- Geometric TransformationsUploaded bySeleneGoberdhan
- 003pltfphase1 1Uploaded byapi-340864650
- OHM_INTERNATIONAL - USA W_ohminternational_com - IndUploaded byureachanit
- BnUploaded byholdonpainends