A DIAG NO NOSTIC MAINT NTENA NANC NC E SYSTEM FOR C C - - PowerPoint PPT Presentation

SU SUPERVISO SORS: S: Associate Professor J

• nathan Binns, Professor Kiril Tenekedjiev, Dr. Rouzbeh Abbassi, Dr. Vikram G

araniya, Michael Lonsdale

A DIAG NO NOSTIC MAINT NTENA NANC NC E SYSTEM

FOR C C OMMERIC IAL A L AND N D NAVAL L VESSELS LS

J ANE C C ULLU LLUM

jane ne.cul ullum um@ut utas.edu. u.au

HMAS SI S SIRIUS C APT APT AI AIN C O O K G G R AV AVING DO C K, NS W, 2014 2014 HMAS SI S SIRIUS S AND HMAS M S MELBOURNE S O S O UT H C C HINA S E A S E A, 2017

1. 1. Periodi dic pl planned d maint ntena nanc nce a and nd RC M C M a are not

• t
• p
• ptimal but wor
• rk

2. 2. Limited dat ata a an and know

• wledge

ge of

• f how
• w

to to inte terpret i t it 3. 3. No n need eed f for innov

• vation
• n?

4. 4. Appl pplications?

C OMMERC IAL AL AN AND N NAV AVAL AL VES ESSEL EL M MAINTEN ENANC E: E: Sta tate te-of

• f-the

he-ar art

2

C HALLENG ES: C ONDI DITI TION B BASED D AND D PRED EDIC TIVE M E MAINTEN ENANC E

▪ Har ardwar are e an and I Infras astructure e – Mobile as e asset et, mar aritime e en environmen ent ▪ Usef eful d dat ata ▪ Qua uant ntity ▪ Inter erpret etat ation

3

C HALLEN ENG ES ES: DATA INTER ERPRET ETATION

▪ Meani ning ngful ul inter erpret etat ation o

• f d

dat ata ▪ Ide deni nitfying ng mai ainten enan ance e tas asks Exper ert Exper erien ence e - Man anual al Rel eliab ability C C en entred ed Mai ainten enan ance e - Man anual al Diag agnostic S System em – Automat atic (can an al also be p e par art of RC M)

4

G OALS? S?

Improve e av avai ailab ability an and red educe o e over eral all m mai ainten enan ance e cost Improve m maint ntena nanc nce s sche hedul duling ng speed eed and c nd cons nsistenc ncy

5 HMAS W WALLE LLER SYDNEY Y HARBOUR, N NSW 5

DIA IAG NOSTIC IC MAINT NTENA NANC NC E S SC HEDULING NG

▪ Diagnose machine health, risk of failure … ▪ Schedule maintenance if if and when en required

System Interval A System Interval B System Interval C PM Interval PM Interval PM Interval PM Interval PM Interval PM Interval

Sche hedul ule m maint ntena nanc nce onl nly w whe hen n req equired ed

Interval A Interval B Interval C

PREDIC IC TIO IONS REQ EQUIREMEN ENTS 6

DIAG NO NOSTIC MAINT NTENA NANC NC E S SYSTEM

FOR OR A C OM OMMERC IAL OR OR NAVAL V VESSEL C OM OMPON ONENT

7

• 2. Mai

ainten enan ance e Sche hedul uling ng -

Decision Theory

• 3. Per

erforman ance e Measurem emen ent -

Availability and O verall Maintenance C

• st
• 1. Risk

sk A Asse ssessm ssment -

C

• ndition Monitoring and

Machine Learning

NUMBER ER 2 G EN ENER ERAL SER ERVIC E E PUMP

C OM OMPONENT APPL PPLIC ATION FRAMEWORK

C OMPONENT APPLIC ATION VAL VALUE = T TRAN ANSLAT ATE + S SC AL ALE + FOREC AS AST Is it BETTE TTER TH THAN periodic PM?

VAL VALUE I IN TRAN ANSLAT ATION

FOR C C OMMERC IAL O OR NAVA AVAL VE VESSEL AP APPLIC AT ATIONS

8 1. C reate system at component level 2. Tune and re-use for similar components on sameor different vessels

• eg. Estimate system reduces maintenance cost of

pump by 10% below current PM: Per Pump : ~$80 AUD per year Total for HTAs, 6 pumps : ~$500 AUD per year Total RAN Fleet– 49 ships, boats, submarines, 10 pumps per vessel: ~$40,600 AUD per year

HTA EL ELWING HTA W WAREE

Fleet eet Ves essel el Su Sub-sy syst stem 1 1 C

• m
• mpon
• nent 1

1 C

• m
• mpon
• nent 2

2 Su Sub-sy syst stem 2 2 Ves essel el Ves essel el

VAL VALUE I IN SC AL ALE

FOR C C OMMERC IAL O OR NAVA AVAL VE VESSEL AP APPLIC AT ATIONS

9 1. C reate systems at component level for high priority components 2. Integrate systems to create higher levels using RC M or alternatives Add individual component savings

10 of 18

VALUE E IN F FOREC EC ASTING

0.2 0.4 0.6 0.8 1 1.2 5 10 15 20 25 Relia iabil ilit ity Ti Time

Reliability of

• f C
• m
• mpon
• nent vs. Time

Corrosion Wear Fatigue

Eac ach s set et o

• f dat

ata a points can an b be g e gen ener erat ated ed using system em at at t time t e t, w wher ere e R = 1 1 – F (m (mode(t (t)), )), also r rec ecommen ends an action and t ther eref efore m e mainten enance e cost 10 10

12

C OM OMPLETED W WOR ORK TO M O MARC H 2018

10 10

DA DATA TA C OLLEC TI TION

• Designed ten experiments, procured and installed hardware, completed experimental data collection and

processing

• Designed C

M data collection process, procured and installed hardware, completed 65%

• f data collection
• Wrote scripts for data processing (experimental and C

M)

• C
• mpiled equipment and maintenance data to date for Number 2 G

eneral Service Pump

• C
• mpleted survey of C

hief Engineer METHOD ODOL OLOG OG Y

• Identified novelty and strengths of methodology using literature review process
• Developed new decision modelling theory in conjunction with supervisor (focus of second paper)
• Designed and wrote scripts for methodology

WRITTE TTEN C OMMUNIC ATI TION OF F RESEARC H

• Literature review paper published in Ocean Engineering J
• urnal
• Internal Serco Hub article on research
• C
• mpleted second paper draft – currently under review by supervisor
• Drafted four chapters of Thesis

13

RE REMAINING W WORK RK

11 11 DA DATA TA C C OLLEC TI TION

• [September 2018] C
• mplete remaining 2/3 of C

M – 8 fortnightly sessions - 8 hours total time

• Process remaining CM data
• Record recommendations of Engineer and preventative maintenance alongside system

recommendations METHOD ODOL OLOG OG Y

• Tune model
• Generate recommendations from CM data using tuned model
• Graph recommendations from methodology, Engineer and PM schedule, calculate availability and

maintenance cost of the three policies WRITTE TTEN C OMMUNIC ATI TION OF F RESEARC H

• Complete second paper draft and submission
• Complete results paper draft and submission
• Complete thesis

C OM OMPON ONENT A APPLIC ATION ON: NUMBER ER 2 2 G EN ENER ERAL S SER ERVIC E E PUMP

1. 1. Risk sk A Asse ssessm ssment - C

• ndition Monitoring and Machine Learning

a. Data for Algorithm Training and C

• ndition Monitoring

13 b. Machine Learning E xamples 23 c. Applying a Machine Learning Algorithm 24

2. 2. Maint ntena nanc nce Sche hedul uling ng - Decision Theory

a. Maintenance Actions as Lotteries 25 b. Modelling Lottery Prizes: Multi-attribute Utility 26 c. Making a Decision: Maximum E xpected Utility 27

3. 3. Per erforman ance M e Meas easurem emen ent -Availability and Overall Maintenance C

• st

Availability and Maintenance C

• st, Validation

28

12 12 13 13 25 25 28 28

DAT ATA F A FOR M MAC AC HINE L LEAR ARNING AND ND C C OND NDITION M N MONI NITORING NG

T wo P

• Purpos
• ses:

1. From Experiments on Test Pump- Build model 2. From Condition Monitoring on No. 2 General Service Pump – Use model to predict condition of No. 2 G eneral service pump

C REA EATE E DATASET ETS DES ESC RIBING C OMMON C C EN ENTRIFUG AL P PUMP FAULTS: S:

1. 1. No f fau ault – Run pump under normal operational conditions al alongside 2. 2. No f fau ault – Run pump under normal operational conditions eng ngine nes r runni unning ng 3. 3. No f fau ault – Run pump under normal operational conditions at at s sea ea 4. 4. Worn I Impel eller er - Lathe impeller fluid side and polish 5. 5. Worn b bear earing – Measure pump bearing with many running hours 6. 6. Dam amag aged ed b bear earing – Grind outer race of new bearing flat and polish 7. 7. Unbal alan anced ed s shaf aft/ Stat atic Imbal alan ance e – Lathe off material from one point of shaft 8. 8. Misal aligned ed s shaf aft/ Offset et m misal alginmen ent – Misalign pump-motor coupling 9. 9. Loose p e pac acking – Loosen casing bolts 10.

• 10. Poor
• or m

mou

• unting – Loosen mounting bolt on pump foot

13 13

T wo P

• Purpos
• ses:

1. From Experiments on Test Pump- Build model 2. From Condition Monitoring on No. 2 General Service Pump – Use model to predict condition of No. 2 G eneral service pump

THE E DATASET ETS (20 min s sessions): S SAMPLE E RATE: E: 1. Vibration: Dual channel on pump Every 2 minutes 2. Temperature: Thermal imaging camera Per Minute 3. Pressure: Suction and discharge gauges Per Minute 4. Motor current: Current clamp on cord Per Minute 5. Packing drip rate: Visual inspection Per Minute 6. Shaft rotation: Tacometer Per experiment

14 14

DAT ATA F A FOR M MAC AC HINE L LEAR ARNING AND ND C C OND NDITION M N MONI NITORING NG

EL ELWING BILG E/ E/ FIRE E SYSTEM EM

TEMPOR ORARY C ON ONFIG URATION ON

Operating c g con

• ndition
• ns for
• r

all pu pumps ps:

• 0.2 bar Suction

2.1 bar Discharge 15 15

16 16

TE TEST T PUMP S SETU TUP NUMBER ER 2 G EN ENER ERAL SER ERVIC E E PUMP

17 17

TE TEST R T RIG S SETU TUP

DATA C O C OLLEC T C TION - EX EXPER ERIMEN ENTAL

T wo P

• Purpos
• ses:

1. From Experiments on Test Pump- Build model 2. From Condition Monitoring on No. 2 General Service Pump – Use model to predict condition of No. 2 G eneral service pump 18 18 1.

• 1. TE

TEST T PUMP/ AIR C C ONDI DITI TIONING PUMP – C

• nduct T

TEN EN E EXPERIMENTS

• f 2

20 min se sessi ssions. s. PhD O bjective – Build a model which can detect the following: 1. No fault, no engines, ship alongside 2. No fault, engines running, ship alongside 3. No fault, ship at sea 4. Worn Impeller 5. Loose packing 6. Damaged bearing 7. Worn bearing (Air Conditioning Pump) 8. Unbalanced shaft/Static Imbalance 9. Misaligned shaft/Offset Misalignment

• 10. Poor Mounting

19 19

6 1 2 3 7 9 8 10

Po Point 1 2 3 4 5 6 7 8 9 10 10 Meas easurem emen ent Vibr bration

• n

Vibr bration

• n

Vibr bration

• n

Tem emper erat ature Tem emper erat ature Tem emper erat ature Vibr bration

• n

Tem emper erat ature Tem emper erat ature Vibr bration

• n

Loc

• cation
• n

Motor, Vertical Motor, Horizontal Drive-end bearing Pump C asing Motor, Drive End Bearing C asing C

• upling

Pump Drive End Bearing C asing, Horizontal Shaft Pump, Bearing C asing Pump C asing, Horizontal

5 4

TEST P PUMP

20 20

6 1 2 3 4 5 8 7 9

AIR IR C ONDIT ITIO IONIN ING PU PUMP

10

Po Point 1 2 3 4 5 6 7 8 9 10 10 Meas easurem emen ent Vibr bration

• n

Vibr bration

• n

Vibr bration

• n

Tem emper erat ature Tem emper erat ature Tem emper erat ature Vibr bration

• n

Tem emper erat ature Tem emper erat ature Vibr bration

• n

Loc

• cation
• n

Motor, Vertical Motor, Horizontal Drive-end bearing Pump C asing Motor, Drive End Bearing C asing C

• upling

Pump Drive End Bearing C asing, Horizontal Shaft Pump, Bearing C asing Pump C asing, Horizontal

T wo P

• Purpos
• ses:

1. From Experiments on Test Pump- Build model 2. From Condition Monitoring on No. 2 General Service Pump – Use model to predict condition of No. 2 G eneral service pump

• 2. Number

er 2 G en ener eral Ser ervice e Pump – C ONDITION MONITORING f for o

• ne 2

e 20 min s ses ession, r rep epea eat for

• rtnigh

ghtly for

• r 6 m

mont nths hs.

PhD O bjective - Detect the following using C M measurement:

1. No fault, alongside 2. No fault, engines running 3. No fault, at sea 4. Worn Impeller 5. Loose packing 6. Damaged bearing 7. Worn bearing 8. Unbalanced shaft/Static Imbalance 9. Misaligned shaft/Offset misalignment 10. Loose mounting

21 21

DAT ATA A FOR M MAC AC HINE LEAR ARNING AN AND C ON ONDITION ON M MON ONITOR ORING

22 22

1 2 3 4 5 6 7 8 9 10

NUMBER ER 2 2 G EN ENER ERAL SER ERVIC E E PUMP

Po Point 1 2 3 4 5 6 7 8 9 10 10 Meas easurem emen ent Vibr bration

• n

Vibr bration

• n

Vibr bration

• n

Tem emper erat ature Tem emper erat ature Tem emper erat ature Vibr bration

• n

Tem emper erat ature Tem emper erat ature Vibr bration

• n

Loc

• cation
• n

Motor, Vertical Motor, Horizontal Drive-end bearing Pump C asing Motor, Drive End Bearing C asing C

• upling

Pump Drive End Bearing C asing, Horizontal Shaft Pump, Bearing C asing Pump C asing, Horizontal

= M MAC HINE LEARNING C LASSIF IFIC ATION

23 23

AP APPLYING A A MAC AC HINE L LEAR ARNING AL ALG ORITHM

▪ Resu sults: s: Probability that pump is in each group:

OK - No fault Impeller wear Damaged PDE bearing …

▪ Nai aive e Bay ayes es A Algorithm

S imple modelling approach G

• od performance on few data

and many features

C M C M V Vector Mac achine L e Lear earning Algor gorithm G r G roup Probabil ilit itie ies

▪ Inp nput: Set of Measurements from Number 2 General Service Pump:

Vibration Temperature Pressure …

24 24

MAINT NTENA NANC NC E A AC TIONS NS AND H HORSE SE RAC ING

25 25

MAXIMUM E EXPEC TE TED U D UTI TILITY TY

27 27

• 3. P

PER ERFORMANC E E MEA EASUREM EMEN ENT

• Avail

ilabil ilit ity vs. PM

• Over

eral all M Mai ainten enan ance C e C

• st vs. PM
• Validation against ex

exper ert rec ecommen endat ations 28 28

SUMMA MMARY

• Innovat

ation need eeded ed in mai ainten enan ance e

• f commer

ercial al an and n nav aval al v ves essel els

• Outlined

ed a d a diag agnostic m mai ainten enan ance e system em ap applicat ation to a s a shipboar ard pu pump

• Tuning

ng a and nd validation o

• f system i

is in p progress ( (TBC September 2018) 2018) 29 29 HMAS P S PERTH AU AUS T R AL ALIAN AN MAR AR INE C O MPL PLE X C O MMO N U US E R FAC AC ILIT Y , WA, 2015 2015

AC KNOWLED EDG EM EMEN ENTS

The candidate acknowledges the support of the ARC Research Training C entre for Naval Design and Manufacturing (RTC NDM) in this investigation Serco Defence Asia-Pacific and the C

• ndition Monitoring

division, Fleet Base East. The RTC NDM is a University-Industry partnership established under the Australian Research C

• uncil Industry Transformation grant scheme (ARC

IC 140100003). The candidate also acknowledges the support of Serco Defence Asia-Pacific and the C

• ndition Monitoring Division, Fleet Base East in providing guidance and resources for this research.

THANKYOU OU!

jane.c .cullum@utas.e .edu.a .au

37 of 18

WORN IMPEL ELLER ER

UNBAL ALAN ANC ED S SHAF AFT/ STAT ATIC IMBAL ALAN ANC E

38 of 18

39 of 18

DAM AMAG AG ED B BEAR ARING

40 of 18

MISA SALIG NED SH SHAFT/ OFFSE SET MISA SALIG NMENT

41 of 18

VI VIBRAT ATION D DAT ATA A QUAL ALITY

0.5 1 1.5 2 2.5 100 200 300 400 500 600 700 800 900 1000

Amplitude mms-1 Frequency Hz Misaligned Shaft/Offset Misalignment No fault alongside

25 Hz 50 Hz 75 Hz

Expect higher amplitudes at 25, 50 and 75Hz due to misaligned shaft - Mobius Institute Training Manual (2008)

0.5 1 1.5 2 2.5 3 3.5 4 100 200 300 400 500 600 700 800 900 1000

Amplitude mms-1 Frequency Hz

No fault alongside Worn Impeller

42 of 18

VI VIBRAT ATION D DAT ATA A QUAL ALITY

25 Hz 520 Hz

Expect higher amplitudes at 25 and 520Hz due to worn impeller - Mobius Institute Training Manual (2008)