sessión 7 exakt

Upload: isaacmedina1234

Post on 06-Jul-2018

223 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/17/2019 Sessión 7 Exakt

    1/32

    Inspection Decisions Part 5

    5-1

    1

    ©A.K.S. Jardine

    Optimizing Condition-basedMaintenance (CBM) Decisions

    CBM Strategies (7.1)Estimating RUL (7.2)

    The EXAKT model (7.3)Case Studies (7.4)

    Andrew K.S.Jardine

    Department of Mechanical & Industrial Engineering

    University of Toronto

    Canada, M5S 3G8

    [email protected] .ca

    October , 2002.

    2

    ©A.K.S. Jardine

    “Smart” Condition Monitoring

    1. Engineering Approaches: physics of failure2. SPC Models: Trending3. Expert Systems: Human-computer 4. Neural Networks: Knowledge discovery

    algorithms (data mining, pattern recognition)5. Optimization Models: Blending risk and economic

    considerations

    Reference: A.K.S. Jardine, “Optimizing Condition-Based Maintenance Decisions”, RAMS 2002 , January 28-31, 2002, Seattle, Washington.

  • 8/17/2019 Sessión 7 Exakt

    2/32

    Inspection Decisions Part 5

    5-2

    3

    ©A.K.S. Jardine

    Analysis of Shear PumpBearings Vibration Data

    – 21 vibration measurements(covariates) provided byaccelerometer

    Using : – 3 measurements (covariates)

    significant

    A Check: – Had model been

    applied to previous histories – Savings obtained = 35 %

    Campbell Soup Company

    Source: Jardine, AKS, Joseph, T and Banjevic, D, “ Optimizing condition-based maintenance decisions for equipment subject to vibrationmonitoring” , Journal of Quality in Maintenance Engineering, Vol. 5. No. 3, pp 192-202 , 1999.

    4

    ©A.K.S. Jardine

    APPROACH USEDHAZARD - RISK OF FAILURE(PROBABILITY OF FAILURE)

    THAT COMBINES

    AGE OF EQUIPMENT ANDCONDITION-MONITORING DATA

    USING

    PROPORTIONAL-HAZARDSMODEL (PHM)

  • 8/17/2019 Sessión 7 Exakt

    3/32

    Inspection Decisions Part 5

    5-3

    5

    ©A.K.S. Jardine

    “Early work with PHM”

    Source: M. Anderson, A.K.S. Jardine and R.T. Higgins, “The use o f concomitant variables in reliability estimation”, Modeling and Simulation , Vol. 13, pp. 73-81, 1982.

    6

    ©A.K.S. Jardine

    Number Flight Hours Fe Cr Hazard Rate

    1 11770 5 6 0.043

    Number Flight Hours Fe Cr Hazard Rate

    1 11770 5 6 0.043

    Number Flight Hours Fe Cr Hazard Rate

    1 11770 5 6 0.043

    2 11660 2 6 0.012

    3 8460 12 2.4 0.0071

    4* 12630 8 1 0.0014

    5 7710 8 0 0.00094

    6* 9240 2 3 0.000297* 5660 10 1 0.00020

    8* 7190 2 2.5 0.000073*Doubtful Removal

    Number Flight Hours Fe Cr Hazard Rate

    1 11770 5 6 0.043

    2 11660 2 6 0.012

    3 8460 12 2.4 0.0071

    4* 12630 8 1 0.0014

    5 7710 8 0 0.00094

    6* 9240 2 3 0.000297* 5660 10 1 0.00020

    8* 7190 2 2.5 0.000073*Doubtful Removal

    Estimated Hazard Rate at Failure

    The hazard rate equation is:

    4.4724100

    tr(t)= ( 24100 )3.47 (.41z +.98z )e 1 2

    where z 1 is Fe concentration and z 2 is Cr concentration

  • 8/17/2019 Sessión 7 Exakt

    4/32

    Inspection Decisions Part 5

    5-4

    7

    ©A.K.S. Jardine

    DATA PLOT

    RISK PLOT Age

    Data

    Age

    Risk

    8

    ©A.K.S. Jardine

    OPTIMAL POLICY - OPTIMAL RISK LEVEL

    Optimalrisk level

    Age

    Risk

    Risk

    Cost/unit time

    RISK PLOT

    COST PLOT

    Ignore risk

    Replace atfailure only

    minimal cost

    optimal risk

  • 8/17/2019 Sessión 7 Exakt

    5/32

    Inspection Decisions Part 5

    5-5

    9

    ©A.K.S. Jardine

    CONDITION BASED OPTIMAL REPLACEMENT

    OF A PRODUCTION SYSTEM

    V. Makis and A.K.S. Jardine

    Department of Mechanical and Industrial Engineering

    University of Toronto

    V. Makis and A.K.S. Jardine, Optimal Replacement in the ProportionalHazards Model, INFOR, Vol. 20, pp 172-183, 1992

    10

    ©A.K.S. Jardine

    We have created atheory, but in order to

    make it work inpractice we need a tool

  • 8/17/2019 Sessión 7 Exakt

    6/32

    Inspection Decisions Part 5

    5-6

    11

    ©A.K.S. Jardine

    Commenced Funding: January 1995Developed by senior researchers, statisticians, andprogrammers at the University of Toronto’s ConditionBased Maintenance Laboratory.Underwritten by the governments of Ontario and Canadaand:

    Zachry

    12

    ©A.K.S. Jardine

    RESEARCH STAFF

    PRINCIPAL INSVESTIGATORS:Prof. A.K.S. JardineProf. Viliam Makis

    RESEARCH STAFF:Dr. Dragan Banjevic, Project Director Walter Ni, Programmer AnalystMurray Wiseman, Research Associate

    Dr. Daming Lin, Post Doctoral FellowDr. Gang Li, Post Doctoral FellowJayne Beardsmore, Administrative/ Research Assistant

    RESEARCH STUDENTS:Yan Gao (M.A.Sc ) Tian Tang (M.A.Sc)Yiding Li (M.A.Sc) Jianmou Wu (Ph.D)Bing Liu (Ph.D) Yimin Zhan (Ph.D)Miao Qiang (M.A.Sc) Ali Zuashkiani (Ph.D)

  • 8/17/2019 Sessión 7 Exakt

    7/32

    Inspection Decisions Part 5

    5-7

    13

    ©A.K.S. Jardine© CBM LABORATORY, UNIVERSITY OF TORONTO

    Age DataDiagnostic Data CBM Model

    MaintenanceDecisions

    CBM OPTIMIZING SOFTWARE

    UNIFICATION OF DATA AND DECISIONS

    Ref: Banjevic D., Jardine, A.K.S., Makis, V. and M.Ennis., “T he Optimal Control Policy and the Structure of the Software forCondition-Based Maintenance” , INFOR, 39, pp 32-50 ,2001 .

    14

    ©A.K.S. Jardine

    EXAKT : The CBM Optimizer

    Two Keys:

    ? Risk

    ? Economics

  • 8/17/2019 Sessión 7 Exakt

    8/32

    Inspection Decisions Part 5

    5-8

    15

    ©A.K.S. Jardine

    Risk factors:• Cholesterol Level• Blood Pressure• Smoking• Lifestyle• Levels of Protein Constituent

    Homocysteine

    Risk factors:• Oil Analysis (Fe, Cu, Al,

    Cr, Pb…..etc.)• Vibration (Velocity and

    Acceleration)• Thermography• Visual Inspection

    ……….………….…………………..

    HEART FAILURE

    Hazard or Risk = f (Age) + f (Risk factors)

    CONDITION BASED MONITORING - AN ANALOGY

    EQUIPMENT FAILURE

    16

    ©A.K.S. Jardine

    Failures/op.hourFailure/flying hour Failures/km.Failures/tonFailures/cycle

    ECONOMIC

    C f = C+K : Total cost of failure replacement

    C p = C : Total cost of preventive replacement

    RISK

    Mg Al Feet 01183.0867.10518.0483.0

    148790148790483.1 ??

    ??

    ?

    ?

    ??

    ?

    ??

    ? ? ? ? ? ?t n z nt z et t HAZARD ???

    ??? ???

    ??

    ?

    ?

    ??

    ?

    ?? ...11

    1

    Contribution of ageto hazard

    Contribution of conditioninformation to hazard

  • 8/17/2019 Sessión 7 Exakt

    9/32

    Inspection Decisions Part 5

    5-9

    17

    ©A.K.S. Jardine

    EXAKT V 3.00 Released in December 2001

    18

    ©A.K.S. Jardine

    EXAKT Procedures Window

  • 8/17/2019 Sessión 7 Exakt

    10/32

    Inspection Decisions Part 5

    5-10

    19

    ©A.K.S. Jardine

    Oil Analysis Data

    20

    ©A.K.S. Jardine

    Oil Analysis Events Data

  • 8/17/2019 Sessión 7 Exakt

    11/32

    Inspection Decisions Part 5

    5-11

    21

    ©A.K.S. Jardine

    Summary of PHM Parameters

    22

    ©A.K.S. Jardine

    Summary of PHM Parameters

  • 8/17/2019 Sessión 7 Exakt

    12/32

    Inspection Decisions Part 5

    5-12

    23

    ©A.K.S. Jardine

    Oil Analysis Decision

    24

    ©A.K.S. Jardine

    Vibration Monitoring Data

  • 8/17/2019 Sessión 7 Exakt

    13/32

    Inspection Decisions Part 5

    5-13

    25

    ©A.K.S. Jardine

    Vibration Analysis Events Data

    11/8/2002

    ©A.K.S. Jardine

    TRANSITION PROBABILITY MATRIX ________________________________________

    Very Smooth

    SmoothRough

    Very Rough

    Failure

    Inspection Interval = 30 days

  • 8/17/2019 Sessión 7 Exakt

    14/32

    Inspection Decisions Part 5

    5-14

    27

    ©A.K.S. Jardine

    Vibration Monitoring Decision

    28

    ©A.K.S. Jardine

    Warning Limits in ‘ppm’

    Normal

    < 20

    300

    >25

    Al

    Cr

    Cu

    Fe

    Si

  • 8/17/2019 Sessión 7 Exakt

    15/32

    Inspection Decisions Part 5

    5-15

    29

    ©A.K.S. Jardine

    In Operation

    Measurements & Decision

    30

    ©A.K.S. Jardine

    – Twelve covariatesmeasured

    – Covariates used: Ironand Sediment – Estimated Saving in

    Maintenance Costs:22% for cost ratio 3:1

    Oil Analysis data from 50 Wheel Motors

    Cardinal River Coals

    Source : A.K.S. Jardine et al, “ Optimizing a mine haul truck wheel motors’condition monitoring program" , JQME , 2001 , pp. 286-301.

  • 8/17/2019 Sessión 7 Exakt

    16/32

    Inspection Decisions Part 5

    5-16

    31

    ©A.K.S. Jardine

    Getting the Data• Using ODBC, there are automated ways of

    capturing the necessary data from your CMMS andcondition monitoring records.

    32

    ©A.K.S. Jardine

    MIMOSA Compliance

    www.mimosa.org

  • 8/17/2019 Sessión 7 Exakt

    17/32

    Inspection Decisions Part 5

    5-17

    33

    ©A.K.S. Jardine

    What isMIMOSA?

    a data model? Called CRIS. (CommonRelational Information Schema)

    System BSystem A

    Mapping Mapping

    CRIS

    strctr

    M achineryInformationM anagementO penSystemsA lliance

    34

    ©A.K.S. Jardine

    CBM OPTIMIZATION

    Executive Summaries

  • 8/17/2019 Sessión 7 Exakt

    18/32

    Inspection Decisions Part 5

    5-18

    35

    ©A.K.S. Jardine

    Analysis of Shear PumpBearings Vibration Data

    – 21 vibration covariatesprovided by accelerometer

    Using : – 3 covariates significant

    A Check: – Had model been

    applied to previous histories – Savings obtained = 35 %

    Campbell Soup Company

    Source: Jardine, AKS, Joseph, T and Banjevic, D, “ Optimizing condition-based maintenance decisions for equipment subject to vibrationmonitoring” , Journal of Quality in Maintenance Engineering, Vol. 5. No. 3, pp 192-202 , 1999.

    36

    ©A.K.S. Jardine

    Failed at WorkingAge = 182 days

    Inspection atWorkingAge = 175 days

    Had we replaced at 175 days…..!!!

  • 8/17/2019 Sessión 7 Exakt

    19/32

    Inspection Decisions Part 5

    5-19

    37

    ©A.K.S. Jardine

    Analysis of Warman-PumpBearings Vibration Data

    – Total 8 pumps each with two bearings(16 bearings) analyzed

    – 12 vibration covariates identified

    Using : – 2 covariates significant – Annual replacement cost savings= 42 %

    Feedback: – Model results found realistic by Sasol

    plant – Significant vibration covariates identified

    by are agreed as a majorproblem

    Sasol Plant

    Source: P.J. Vlok, J.L. Coetzee. D. Banjevic, A.K.S. Jardine and V.Makis, “An Application of VibrationMonitoring in Proportional Hazards Models for Optimal Component Replacement Decisions”, JORS , Vol. 53,

    No. 2, pp. 193-202.

    38

    ©A.K.S. Jardine

    Open Pit Mining Operation

    – Covariates used:Iron, Aluminum,Magnesium

    – Saving inMaintenance Costs:25%

    – Averagereplacement timeincrease: 13%

    – Warranty limit couldbe increased

    CAT 793B Transmissions Oil data analyzed

  • 8/17/2019 Sessión 7 Exakt

    20/32

    Inspection Decisions Part 5

    5-20

    39

    ©A.K.S. Jardine

    Definition of Problem• When should this oil

    well pumping system(casing, sucker rod, andpump) be replaced,given inspection dataand events data?

    • Is it more profitable topreventively replace, orrun until failure?

    Oil Well Pumping System

    © CBM Lab: University of Toronto

    40

    ©A.K.S. Jardine

    The EXAKTOptimal

    CBM PolicyModel

    Types of CBM opt imization models

    $100 $30040 h 10 h

    MTTR failure

    MTTR

    planned

  • 8/17/2019 Sessión 7 Exakt

    21/32

    Inspection Decisions Part 5

    5-21

    41

    ©A.K.S. Jardine

    Case Study: Paper Mill• Data used was from a variable speed paper mill

    – At present produces carbonless till receipts

    • Data was periodically monitored and storedwithin the MIMIC 2001 CMM system – 10 Year history of the plant – Mainly vibration and running speed

    • Area called the drier gearbox section• 18 identical units• 8 failure mode parameters• 140 measurements collected over the history

    Source: R. Willetts, University of Manchester, England

    42

    ©A.K.S. Jardine

    Results

    • The EXAKT analysis

    – Of the 8 original parameters, 5 vertical and 3axially, only 2 parameters were needed

    • Acceleration in the vertical direction• Gear mesh in the axial direction

  • 8/17/2019 Sessión 7 Exakt

    22/32

    Inspection Decisions Part 5

    5-22

    43

    ©A.K.S. Jardine

    Hong Kong Mass Transit RailwayCorporation

    – Had excessive tractionmotor ball bearingfailures

    – CBM to monitor bearinggrease colour

    – Changed inspectionsfrom every 3.5 years toannual

    – Reduced failures/yr..From 9 to 1

    – Reduced total costs by55%

    The reality :Failures reduced to 2/yr.

    44

    ©A.K.S. Jardine

    EXAKT Modeling with MWM Diesel Engine

    employed on Halifax Class ships.

    Maintenance and Diagnostic Data

  • 8/17/2019 Sessión 7 Exakt

    23/32

    Inspection Decisions Part 5

    5-23

    45

    ©A.K.S. Jardine

    Inspections Table

    46

    ©A.K.S. Jardine

    Condition-based Optimal Replacement Policy

  • 8/17/2019 Sessión 7 Exakt

    24/32

    Inspection Decisions Part 5

    5-24

    47

    ©A.K.S. Jardine

    OPG (Pickering Nuclear Station)Hydrodyne seals prevent leakage of heavy water duringfuelling operation: Seal Leak Rate data from 4 reactors

    48

    ©A.K.S. Jardine

    Integrating EXAKT&

    CMMS/EAM/ERP System

  • 8/17/2019 Sessión 7 Exakt

    25/32

    Inspection Decisions Part 5

    5-25

    49

    ©A.K.S. Jardine

    CBMdata

    E v e n t

    D a t a

    C o n d

    .

    D a t a

    Work ordersVA+OA+Othersignals

    CMMS

    I n s p e c t i o n d a t a E v e n t d a t a

    ModelingModule

    WMOD WDEC

    DecisionModule

    DecisionModels

    DMDR

    D e c i s i o n s

    50

    ©A.K.S. Jardine

    A Web Agent SiteMining

    Manufacturing

    Food and Beverage

    Pharmaceutical

    Power Environmental Pulp and Paper Web Agent Deployment Source: M.Wiseman

  • 8/17/2019 Sessión 7 Exakt

    26/32

    Inspection Decisions Part 5

    5-26

    51

    ©A.K.S. Jardine

    AAAV

    Advanced Amphibious Assault Vehicle

    52

    ©A.K.S. Jardine

    Objectives

    • Demonstrate Affordable Prognostic System – Wireless, Small, Scalable, Open

    • Provide Robust Fault Detection – Signal Processing, Identification, Isolation, Severity,

    Fidelity

    • Prognostics for Maintenance Decision Making – Prognosis, Risk, Decision Support

  • 8/17/2019 Sessión 7 Exakt

    27/32

    Inspection Decisions Part 5

    5-27

    53

    ©A.K.S. Jardine

    Prototype System Sensor Placement

    -Vibration and Temperature

    -Tachometer

    54

    ©A.K.S. Jardine

    ICHM 20/20

    Sensor

    Sensor

    Sensor

    ICHM MCU

    DSP

    TEDS

    Memory

    HCI

    App. BluetoothRadio

    ICHM®

    Sensor Module

    SignalCond .

    A/D

    Self-test

    Serial

    Interface

    •120mm x 80mm x 50mm•Supports MIMOSA/OSA CBMInformation Model•Bluetooth™ Wireless Technology

  • 8/17/2019 Sessión 7 Exakt

    28/32

    Inspection Decisions Part 5

    5-28

    55

    ©A.K.S. Jardine

    AAAV

    Advanced Amphibious Assault Vehicle

    56

    ©A.K.S. Jardine

    Model Evolution from PriorKnowledge to Prior Data

    The Continuous Improvement Cycle

  • 8/17/2019 Sessión 7 Exakt

    29/32

    Inspection Decisions Part 5

    5-29

    57

    ©A.K.S. Jardine

    Additional CBM References1. Wiseman, M. Optimizing Condition-Based Maintenance, in Maintenance Excellence:Optimizing Equipment Life Cycle Decisions, J.D. Campbell and A.K.S. Jardine(Editors), Marcel Dekker, New York, 2001.2.Jardine, A.K.S., Makis, V., Banjevic D., Bratejevic,D. and Ennis, M., “A DecisionOptimization Model for Condition-Based Maintenance”, Journal of Quality inMaintenance Engineering, Vol. 4, No.2, pp 115-121, 19983. Jardine, AKS, Joseph, T and Banjevic, D, “ Optimizing condition-based maintenancedecisions for equipment subject to vibration monitoring” , Journal of Quality inMaintenance Engineering, Vol. 5. No. 3, pp 192-202, 19995..D Banjevic, Jardine, A.K.S., Makis, V. and M. Ennis, “The Optimal Control LimitPolicy and the Structure of the Software for Condition- Based Maintenance”, INFOR,2001

    6. www.mie.utoronto.ca/cbm (For information about the CBM research activities at theUniversity of Toronto)7. www.omdec.com (For information about the EXAKT: CBM Optimization software)

    58

    ©A.K.S. Jardine

    Summary

  • 8/17/2019 Sessión 7 Exakt

    30/32

    Inspection Decisions Part 5

    5-30

    59

    ©A.K.S. Jardine

    An Overview of EXAKTEvents Data Inspection Data

    Replacement Recommendation © CBM Lab: University of Toronto

    EXAKT ModelingModule

    EXAKT DecisionModule

    CBM Model

    60

    ©A.K.S. Jardine

    Some CBM Optimization Studies

  • 8/17/2019 Sessión 7 Exakt

    31/32

    Inspection Decisions Part 5

    5-31

    61

    ©A.K.S. Jardine

    While much research and productdevelopment in the area ofcondition based maintenancefocuses on data acquisition andsignal processing, the focus of thissession has been the third and finalstep in the CBM process –optimizing the decision making step

    62

    ©A.K.S. Jardine

    THANK YOU

    www.mie.utoronto.ca/cbmwww.omdec.com

  • 8/17/2019 Sessión 7 Exakt

    32/32

    Inspection Decisions Part 5

    63

    ©A.K.S. Jardine

    Andrew JardineAndrew Jardine is a professor and principal investigator at the Condition-Based Maintenance laboratory in the Department of Mechanical and IndustrialEngineering at the University of Toronto where the EXAKT software has beendeveloped. He also serves as a Senior Associate Consultant to the GlobalLeader of PricewaterhouseCoopers’ Physical Asset Management practice. Hehas a PhD from the University of Birmingham. He is the author of theAGE/CON and PERDEC software that is licensed to organizations includingtransportat ion, mining, electrical utilities, and process industries. He wrote the

    book, “ Maintenance, Replacement and Reliability” first published in 1973. Inaddition to being a sought-after speaker he is a recognized authority in theworld of Reliability Engineering and in the optimization of maintenancedecision making. Professor Jardine was the 1993 Eminent Speaker to theMaintenance Engineering Society of Australia and in 1998 was the first

    recipient of the Sergio Guy Memorial Award from the Plant Engineering andMaintenance Association of Canada in recognition of his outstandingcontribution to the Maintenance profession. He is the co-editor with JDCampbell of the 2001 published book Maintenance Excellence: OptimizingEquipment Life Cycle Decisions.

    Tel: +1 (416) 978-2921; E.mail: [email protected]