Project Background

Austrian Armed Forces desire optimizing their flight and maintenance operations

• Taking into account probabilities of unexpected failures
• Considering a short, medium and long time-horizon
• For all of their aircraft types
• Data is highly confidential and requires in-depth security screening of participants
• Stakeholders

Motivation

• Scientifically interesting
  • Little prior literature wrt. military aircraft
  • Flight and Maintenance Planning (FMP) of mission aircraft vastly simplifies problem
  • Focus was on larger fleets and multiple locations
  • Highly complex combinatorial optimization problem
• Applications for neutral country
  • Optimization helps participation in UN peacekeeping
  • Increases availability for rescue missions
  • Ensures neutrality requirements
• Reduces cost for tax payers

Military Planes vs. Commercial Airliners

• Performance vs. reliability
• Short vs. long downtimes
  • Commercial airliners have much shorter downtimes
• Type of flying
• Age and high tech
• Component requirements: supersonic engines, ejection seats, ...

Methodology

• Mixed integer linear program
  • Aircraft flowchart heuristic
  • Horizon splitting heuristic
  • Metaheuristic wrapper around the above
  • Simulation
• Platform and solver agnostic (relying only on open source software)
• Developed software in agile manner
• Used structured text to tackle input data complexity
• Relied on self-generated benchmark instances

(Short-term) Model

• Maximize minimum and overall readiness, balance usage and ensure readiness beyond time horizon
  1. Ensure that all mission's patterns are assigned to a flight object.
  2. Ensure required number of objects is available.
  3. Perform up to one pattern per flight object and period and only if ready.
  4. Only capable flight objects may participate in missions.
  5. Set flight hours in short term case.
  6. Can only start maintenance in capable hangar.
  7. Cannot use hangar for more than one maintenance in a single time period.
  8. Cannot perform two maintenance operations on the same FO at once.
  9. Ensure that all required capabilities are available when needed.
  10. Ensure a spare component is available when beginning a component exchange.
  11. Maintenance sequence has to be preserved in each maintenance cycle.
  12. Each maintenance may only be performed once per object.
  13. Maintenance operations have to be completed once started.
  14. Maintenance has to stop when completed.
  15. Avoid oscillation of maintenance flag.
  16. Enforce calendric maintenance trigger.
  17. Enforce negative combined (calendric & rft) maintenance tolerances.
  18. Enforce negative calendric maintenance tolerances.
  19. Enforce negative rft maintenance tolerances.
  20. Enforce negative usage maintenance tolerances.
  21. Update the residual flight time when no maintenance happens.
  22. 'Refill' the residual flight time at the beginning of any maintenance.
  23. Update the residual usage when no matching maintenance starts.
  24. 'Refill' the residual usage at the beginning of matching maintenance.
  25. Ensure minimum residual flight time to be available for missions.
  26. Avoid oscillation of readiness flag.
  27. Never ready after lifetime is exceeded.
  28. Never ready after lifetime flight hours are exceeded.
  29. Not ready while in maintenance.
  30. Link the counter variables for consecutive cycles to regular readiness.
  31. Ensure feasible residual usage values in last period.
  32. Account for unexpected events according to statistical analysis.
  33. Count minimum number of available objects over all periods.
  34. Calculate cumulative flight hours for each flight object.
  35. Enforce priorities as planned flight hours with tolerances.
  36. Mark the flight objects' unavailable periods.
  37. The first period must not postpone the readiness counter.
  38. Mark the hangars' unavailable periods.
  39. Set the initial residual usage values for all maintenance cycles.
  40. Ensure that scheduled operations take place.
• Longer time horizons serve as input to shorter ones
• Long ⇒ medium ⇒ short term

Summary and Outlook