Maintenance optimisation: 50 years of models and applications

1
Maintenance optimisation 50 years of models
and applications
Rommert Dekker Professor of Operations Research,
Quantitative Logistics and Information Technology
Erasmus Universiteit Rotterdam
2
Maintenance optimization

• An overview of history
• Maintenance Optimization
• The delay time model
• Other application areas - conflicts between
  organisations- road maintenance - service
  logistics

3
Change in maintenance techniques

• Third Generation
• Condition monitoring
• Design for reliability and maintainability
• Expert systems
• Failure mode and effect analyses
• Maint. Mgmt Info Syst.
• RCM
• Multi-skilling and team work (TPM)

• Second Generation
• Scheduled Overhauls
• Systems for planning and controlling work
• Big, slow computers
• Life-cyle costing
• Hazard studies

• First Generation
• Fix it when broke

1940 1950 1960 1970
1980 1990 2000
Source naar Moubray (1991)
4
Today (2006)

• Extensive use of Condition monitoring and it much
  better insight into failure mechanisms
• Most systems are pretty reliable, but can be
  better
• Computerised Maintenance Management Information
  Systems are widely used
• There can be extensive IT support for the
  maintenance technician
• Decision support systems are used at many places
• Maintenance knowledge can be shared through the
  WWW
• Asset management has come up as wider principle
  than MM.

5
Today maintenance problems

• Rated secondly compared to new construction, also
  at universities
• Relation budget-maintenance output is not clear
  at a high level, with the following results -
  constant pressure on budgets- effect of budget
  cuts is visible only after some years
• Government policies change rapidly and are less
  predictable, due to societal pressures
  (accidents) (e.g. railway privatisation)
• Privatisation government services creates much
  uncertainty
• Much has to be learned on effects of outsourcing
  maintenance

6
Example

• Split up of Dutch Railways (NS) into a passenger,
  a cargo operator, a Rail Infra Manager, Rail
  Traffic Control and a Rail Capacity Allocator.
  Maintenance has been outsourced to 3 process
  operators.
• At the same time, there have been large
  maintenance budget cuts in anticipation of an
  increased efficiency, although the need for
  maintenance increased because of an increase in
  traffic.
• Similar developments in the UK. First
  privitisation of rail infra management and later
  nationalisation (Railtrack - NetworkRail)
• At the moment hundreds of million euros needed to
  catch up backlog. At the same time large
  disruptions because of computer failures!

7
Maintenance Optimisation what is it?

• Model deterioration, determine risks and costs of
  failure and its consequences as well as costs of
  maintenance
• Determine best maintenance concept (which type
  of maintenance )
• Optimise within the type of maintenance (best
  frequency, inspection policy, replacement limits,
  etc).
• Coordinate maintenance execution with production
  and other maintenance (sometimes separate of 3)
• Determine priorities in execution of preventive
  maintenance and deferrable corrective, determine
  necessary work capacity

8
NB 40 million hits in Google
9
First models date from end 50s / 60s Barlow
Proschan, Jorgenson and McCall
10
Maintenance optimization

• An overview of history
• Maintenance Optimization
• The delay time model
• Other application areas - conflicts between
  organisations- road maintenance- service
  logistics

11
Delay time model (Christer)

• Objective to reduce the number of failures by
  removing faults (incipient failures) if
  discovered at inspections, which are carried out
  at regular intervals of fixed length T
  (regardless of failures)
• Model
• two stages as good as new ? fault ? failure
• faults occur with (Poisson) rate ?
• random (delay) time between fault and failure has
  Cdf F(.)
• Advanced statistical techniques are needed to
  estimate delay time distribution
• inspection costs ci, failure replacement costs cf
• Number of failures in cycle
• Average costs Optimise by applying grid
  search

12
Example delay time model application to
ventilators in the operating department
13
Delay time model

• More than 10 applications, but mainly by
  academics
• Many papers studied the estimation of the
  delay-time (as its start is unobservable). Good
  techniques are available today.
• Yet connection with RCM delay time P-F
  interval has hardly been made
• Communities do not talk with each other

14
What is needed for optimisation?

• Data - sourceshistoric dataexpert
  judgementcondition measurements and physical
  models
• Modelsgeared to the decisions to be taken and
  indicating what kind of data is needed
• Inference engines and toolsyardsticks, graphical
  aids, computers
• Much data is recorded for maintaining status quo
  rather than for finding optimum way of
  operations!

15
Issues Maintenance Optimisation in Practice

• Determination of (indirect) costs of failure is
  difficult, as many interventions may happen
• Statistical lifetimes are difficult to get and
  uninformative, because of a- lousy
  registration- modification of equipment in case
  of many failures
• Time-based preventive maintenance has been
  replaced by condition-based
• Execution of optimisation is complex and requires
  special software and expertise

16
According to Moubray (1991)

• Optimisation is like shooting with a canon on a
  fly and almost always not necessary
• Only in 5 of the parts /systems there is ageing
  and planned PM is necessary in other cases it is
  questionnable.
• Question always the same study is referred to
  (airline parts) so is the 5 true?

17
Economic optimisation vs RCM

• RCM
• general, simple approach
• moderate data requirements
• can be done on paper, or by computer
• focusses on critical components
• suitable for complex systems
• decisions are structured, but not fully supported
• no quantitative support of choices to be made
  (eg. Intervals)
• suitable at component level

• Economic optimisation
• specific, complex approach
• heavy data requirements
• needs computer
• available models for simple systems, new
  approaches needed for complex systems
• decisions are quantitatively supported, allowing
  what if questions
• economic quantification often gives more insight
• suitable for important decisions on system level

A new methodology is needed which combines the
simpleness and generality of RCM with some
economic quantification and optimisation
(Rausand Vatn 96, Horton, 92, Dekker and
Scarf 98)
18
Maintenance optimisation in DSS

• However use of the DSS MAINOPT at Shell PERNIS
  refinery saved more than 2 mln in 2 years time.
• Crucial elements combination between reliability
  and costs. Allows what-if evaluations.
  Furthermore, ensures that problems are solved in
  cooperation between maintenance and production,
  rather than in a fighting mode. This can not be
  provided by RCM.

19
Examples of decision support systems

• Design reliability selection of units,
  redundancy SPARC, RAMA, Miriam, Maros
• Strategic maintenance concepts and intervals
  for a single unitMainopt, Manco, KMOSS, MACRO
  budgets and life cycle costing
  LCC-Optcondition maintenance Jardine Makis
  97
• Tactical, operational best execution of
  maintenance -PROMPT (turbines), IVON (roads),
  PONTIS (bridges)ECOTRACK (railtrack)
• Several more are on the market

20
A comparison of application areas

• Equipment maintenancelarge diversity, complex
  hierarchy, deterioration location dependent, poor
  data collection
• Civil structure maintenance (roads,
  bridges)complex but well studied and slow
  deterioration, often much repetition, maintenance
  driven by norms
• Airplane maintenanceformalised environments,
  good data collection, little opportunity to
  change concepts, optimisation mainly in
  logistics
• Note methodological people lack domain
  knowledge, however different types of engineers
  do not talk with each other, so techniques are
  not easily spread out

21
Maintenance optimization

• An overview of history
• Maintenance Optimization
• The delay time model
• Other application areas - conflicts between
  organisations- road maintenance- service
  logistics

22
Important relations between stakeholdersneed to
be well-defined and organised to ensure overall
optimisation
Manufacturer
design
Indicates information,money,power
Owner
Maintainer
User
operations
23
Conflicts between organisations

• In design phase advanced reliability studies have
  been carried out, taking failure consequences and
  costs into account as well as lifetime
  distributions
• As a result redundancy decisions are made
• However, once system is transferred to user, he
  has to learn from scratch again. He also does not
  have the scale the OEM has, so learning goes
  slow.
• The OEM however is often not involved in the
  operations, creates uncertainty to make more
  money.

24
Questions

• Why does the OEM not only sell equipment, but
  provide also all data around it, including system
  structuring in SAP format?
• Observation
• IHC Holland fabricates dredging ships and left
  maintenance traditionally to user. Now it is
  interested in providing service, helping to
  execute maintenance and provide right spare parts
  in time. She is not the only one!
• Philips is turning to Medical systems, which are
  much more service intensive with much higher
  profit margins!

25
Whats an E-SPIR 2000 ? (Electronic - Spare Parts
Interchangeability Record)
A software program to obtain spare parts
information in projects from the equipment
suppliers in a standard format.
The tool supports the review, selection and
purchase of spare parts and provides progress and
budget control for project management.
The program consists of two parts
Suppliers program, in which the Suppliers prepare
the spare parts quotation in a SPIR format.
The program is available free of charge via
Internet.
Advisors program available against a Licence
FeeIn the advisor program all Supplier SPIRs are
collected in one file.
Look at the Website HTTP//www.e-spir.comMore
than 2100 suppliers listed and a DEMO version
available
26

A SPIR sheet as it used to be

• Complaints
• Unreadable
• Many errors

• Suppliers often do not fill in all data required

27
Maintenance optimization

• An overview of history
• Maintenance Optimization
• The delay time model
• Other application areas - conflicts between
  organisations- road maintenance- service
  logistics

28
Road Maintenance when to do it?

• Much highway road maintenance is done at night,
  while one or two lanes are closed for traffic.
  Time for maintenance is short (? inefficiency)
  and safety of road workers is not guaranteed.
• Question would it be advisable to cluster all
  maintenance and maintain large road stretches
  (about 4 km) in one go and redirect traffic to
  the other road side (in a 3-1 or 4-0 system)?
• Supposed Advantages safer, all maintenance can
  be done for 10 years, fewer traffic
  jamsDisadvantage is clustering a good idea
  (not condition-based!)

29
The three systems from top-bottom- traditional
(2-1)- 3-1 system- 4-0 system
30
(No Transcript)
31
Figure 7.2 implementation of the
sector-integration approach
32
Figure 7.3 implementation of the
junction-to-junction maintenance approach
Application of junction-to-junction Grouping of
maintenance

33
Long-run average preservation- and queuing-cost
per year, for porous asphalt roads
v/w vehicles per carriageway both carriage
ways are 12.5 meters broad.
34
Empirical distribution function of the yearly
preservation costs for dense asphalt roads and
Traditional (sector-integration) approach
35
Empirical distribution function of yearly
preservation costs for dense asphalt roads for
junction-to-junction regeneration (within a 3-1
system)
36
Maintenance optimization

• An overview of history
• Maintenance Optimization
• The delay time model
• Other application areas - service logistics

37
THE MAIN GOAL

RESEARCH
RESEARCH
To deliver innovative products that can in
principle be implemented in practise with the
purpose to improve the quality and to reduce the
costs of service logistics operations.
Examples new service logistics concepts
planning and control mechanisms The
innovative character will be demonstrated by
pilot studies and publication of the main ideas
in scientific journals.
38
RESEARCH TOPICS

RESEARCH
RESEARCH

• Design for service logistics
• Coordination of spare part inventory and
  transportation management
• Integrated Control of Service Parts Inventories
  and Repair Shops
• Demand Criticality
• Preventive maintenance and service parts
  inventory control
• Demand Forecasting and Asset management
• Design of service contracts
• Networks with lateral and emergency shipments
• Integrated control of service parts, service
  tools, and service engineers

39
Companies Involved

RESEARCH
RESEARCH

• Fokker Services B.V.
• Philips Medical Systems BV
• Stork PMT B.V.
• Thales Nederland BV
• voestalpine Railpro BV
• NedTrain B.V.
• DAF Trucks B.V.
• Ortec Consultants
• Who more?

40
Present Service Issues (1)

• Parts assortment- has expanded over the years
  (10.000 to 200.000 items)- many non-movers-
  limited standardisation
• Stocks- too many (guess 20 too much)-
  unbalanced- reflect uncertainty in the supply
  chain
• Parts get obsolete

41
Present Service Issues (2)

• Forecasting- difficult for slow movers-
  information on installed based and reliability
  lacks
• Obsolescence- suppliers suddenly stop delivering
  - final buys are difficult
• ICT- migrations to ERP are complex and take much
  capacity- use of ERP is complex, not all
  features understood- limitations of ERP are
  difficult to quantify in the boardroom

42
Present Service Issues (3)

• Cooperation in the supply chain
• Limited or not
• Historically created powerpositions
• Suppliers do not cooperate
• Installed-base information
• Often not available
• Search for right info is difficult
• A pro-active role is difficult to achieve

43
Trends and ambitions companies

• After sales supply chain management
• Centralised control
• Being the supply chain coordinator
• Control on (customer differentiated) uptime
• Outsourcing
• Not only warehousing and distribution
• Maintenance management and spare parts management
• ICT
• Whole supply chain structure and processes in one
  system
• Leave possibility open for (ERP certified) add-ons

44
Preventive Maintenance and spare parts

• Demand for spares is difficult to predict, low
  and highly erratic monthly demand 0, 0, 7, 0,
  1, 0, 0, 0, 31, 0, 0, 0, 0,
• As a result high inventories are maintained, but
  can they be lowered?
• Inventory deployment can be improved if- we do
  better forecasting by using all kind of
  information- have more insight into the demand
  process, e.g. by considering the maintenance
  planning process

45
Maintenance and Spares
46
Asset Management - questions

• Many companies / organisations struggle with
  asset management. What economic benefits are IT
  systems able to provide? In which cases is a
  detailed asset management attractive?
• Setting up an asset management is not the main
  work it is the maintenance.
• Asset management information should be provided
  from the start by the builders / manufacturers
• Asset Management is useful for maintenance, but
  also for service logistics and for recovery of
  end-of-use / life value!

47
Maintenance optimization - conclusions

• We started 50 years ago with mathematical models,
  but it is not the models, but the domain
  knowledge and IT which makes it worth
• It is certainly worthwhile, but we do not to
  fight for more applications
• Still economics pressure drive for more
• Automatic registration and cheaper IT will be big
  drivers for more application, but the main
  question is how and where

48
Maintenance optimization - Questions

• Technical systems are quite diverse civil,
  mechanic, planes, roads how can we learn from
  them
• How can we learn from different areas
• What do the users see as main issues?
• Service logistics now sees many applications
  will this also be the case for maintenance
  optimization?