Скачать презентацию Modelling the airline costs of delay propagation Скачать презентацию Modelling the airline costs of delay propagation

fbb381de757d0846f862aa33f1fa826c.ppt

  • Количество слайдов: 50

Modelling the airline costs of delay propagation – a European perspective Dr Andrew Cook, Modelling the airline costs of delay propagation – a European perspective Dr Andrew Cook, Principal Research Fellow Graham Tanner, Research Fellow University of Westminster AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011

Acknowledgement The University of Westminster gratefully acknowledges the generous support of the European Organisation Acknowledgement The University of Westminster gratefully acknowledges the generous support of the European Organisation for the Safety of Air Navigation (EUROCONTROL) for enabling this research to be carried out. This presentation and the accompanying paper (to follow) have been cofinanced by EUROCONTROL under its Research Grant scheme. The content of the work does not necessarily reflect the official position of EUROCONTROL on the matter. Air France, Austrian, Britannia Airways, British Airways, Condor, CSA Czech Airlines, easy. Jet, Iberia, KLM, Lufthansa Systems: Lido/Flight AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011

Airline delay propagation § Overview and context § Modelling the primary (‘root’) costs – Airline delay propagation § Overview and context § Modelling the primary (‘root’) costs – cost elements by level – some key results § Modelling the network costs – reactionary delay – initial results § Developing new metrics AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011

Overview and context Overview and context

Overview and context § 2010 modest traffic growth, after unprecedented drop in 2009 – Overview and context § 2010 modest traffic growth, after unprecedented drop in 2009 – 9. 5 million, c. f. 9. 4 million (+ 0. 8%) § 24. 2% of arrivals > 15 mins late in 2010 (17. 9% in 2009) – worst on record (since 2001); average delay per flight (all) = 14. 8 mins – now very similar to 2006 traffic (when 22% of arrivals > 15 mins late) – still the prevalent KPI, more on which later § Arrival delay is mainly driven by departure delay – 46% reactionary; 54% primary, of which: 56% attributable to “turnaround”* (airline, airport, handling agents) 33% ANS: airport < en-route: significant increase e/r mostly due strikes ~ 11% weather (non-ANS) - increased; bad weather December 2010 – total ATFM delay mins up 82% (2010 on 2009) mainly due to en-route * Careful what this means, sometimes “ground” or “ at-gate” is better. AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011

Overview and context Variability of flight phases on intra-European flights AGIFORS Airline Operations Conference Overview and context Variability of flight phases on intra-European flights AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011

Overview and context § Cancellations - a feature of 2010 – 111 000 flights Overview and context § Cancellations - a feature of 2010 – 111 000 flights cancelled due Eyjafjallajökull ash cloud (April-May 2010) – 5 000 ‘extra’ flights, repositioning and repatriation – ash cloud alone reduced annual traffic growth by 1. 2% – 26 000 cancellations due to strikes in France and Spain – 45 000 cancellations due to bad weather § Limited effect on punctuality AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011

Overview and context § Established change – Single European Sky – SESAR technical element Overview and context § Established change – Single European Sky – SESAR technical element of SES, uses SES legislation (EC is regulator) – Definition (2005 -08), Development (2008 -13), Deployment (2014 -20) – performance-based approach: focuses on what, rather than how § Evolving change – designation (2010) of EUROCONTROL’s PRC, supported by the PRU, as the SES Performance Review Body – adoption (2010) of performance targets for 1 st ref period (2012 -2014) – Commission's new roadmap to a Single European Transport Area for 2050, plans to harmonise pax rights across all modes (EC, 2011) – specifically cites pax disruption during severe weather and ash cloud § Imminent change – EU ETS: extending to aviation (01 JAN 12) based on gate-to-gate fuel AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, all AOs – legislation currently: 2011 operating to/from EU surrender permits

Overview and context § SESAR KPAs, KPIs for 2020 – departure: – arrival: – Overview and context § SESAR KPAs, KPIs for 2020 – departure: – arrival: – fuel: (Performance Target and Target Concept ) ≥ 98% of flights departing as planned ± 3 mins other 2%: average delay ≤ 10 mins > 95% of flights arrival delay ≤ 3 mins other 5%: average delay < 10 mins > 95% of flights fuel more than planned ≤ 2. 5% other 5%: average additional consumption ≤ 5% NB. new definition of ‘on time’: ≤ |3 mins| – less variation in the actual block-to-block times for repeatedly flown routes using aircraft with comparable performance, block-to-block σ < 1. 5% of route mean – less reactionary delay & fewer reactionary cancellations (-50% 2010 -2020) – other KPAs include: capacity, flexibility, cost effectiveness, efficiency AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011

Modelling the primary (‘root’) costs Cost elements by level Modelling the primary (‘root’) costs Cost elements by level

Cost elements by level types of cost (in-house models, except fuel) fleet (fuel) crew Cost elements by level types of cost (in-house models, except fuel) fleet (fuel) crew maintenance passenger all fleet costs (depreciation, rentals & leases) Lido/Flight, € 0. 60/kg; CO 2 (2012) schemes, flight hours, on-costs, overtime extra wear & tear powerplants/airframe ‘soft’, ‘hard’ & internalised costs AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011

Cost elements by level strategic fleet (fuel) tactical = f (service hours) ≠ f Cost elements by level strategic fleet (fuel) tactical = f (service hours) ≠ f (utilisation) = 0 --- same (no hedging; carriage penalty) --- crew unit marginal (0 … full o/t) maintenance unit marginal (e. g. fixed LTOs) passenger AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011 0 dominate, non-linear

Cost elements by level ac e (r ) ry na tio al tic c Cost elements by level ac e (r ) ry na tio al tic c ta ic g te ra st time fleet (fuel) crew maintenance passenger AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011 at-gate using up hours a/c powered, small ignores transfer

Cost elements by level strategic tactical fleet fuel crew linear maintenance passenger AGIFORS Airline Cost elements by level strategic tactical fleet fuel crew linear maintenance passenger AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011 non-linear

Modelling the primary (‘root’) costs Some key results Modelling the primary (‘root’) costs Some key results

Airline passenger Kano satisfaction model, Wittmer and Laesser (2008) AGIFORS Airline Operations Conference In-house, Airline passenger Kano satisfaction model, Wittmer and Laesser (2008) AGIFORS Airline Operations Conference In-house, bespoke surveys and airline models London, United Kingdom, May 16 - 19, 2011

ch = pt q Regulation (EC) 261 (17 February 2005); airline policy AGIFORS Airline ch = pt q Regulation (EC) 261 (17 February 2005); airline policy AGIFORS Airline Operations Conference Limited airline - 19, 2011 London, United Kingdom, May 16 data & literature; care & reaccommodation model

Tactical cost distributions for 15 mins (base scenarios) B 733, B 734, B 735, Tactical cost distributions for 15 mins (base scenarios) B 733, B 734, B 735, B 738, B 752, A 319, A 320, A 321, AT 43, AT 72; B 747, B 763

Primary cost (k€) Delay (mins) AGIFORS Airline Operations Conference London, United Kingdom, May 16 Primary cost (k€) Delay (mins) AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011 B 738

12 Primary cost (k€) 9 B 738 3 6 Delay (mins) AGIFORS Airline Operations 12 Primary cost (k€) 9 B 738 3 6 Delay (mins) AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011

Modelling the network costs Reactionary delay Modelling the network costs Reactionary delay

Reactionary delay Ratio of reactionary to primary delay (intra-European) (PRR 2010) Shows the sensitivity Reactionary delay Ratio of reactionary to primary delay (intra-European) (PRR 2010) Shows the sensitivity (or robustness) of the network to delay AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011

Reactionary delay § A thought from Europe … “Despite of the large share of Reactionary delay § A thought from Europe … “Despite of the large share of almost 50% of reactionary delay, there is presently only a limited knowledge of how airline, airport and ATM management decisions affect the propagation of reactionary delay throughout the air transport network. ” (PRR 2010) § A thought from the US … “With the data on passenger trip reliability available, a public debate can occur on acceptable performance tolerances. The result would be a service standard equivalent to the 15 minute On-Time Performance standard for flight delays. Since passenger delays are derived from cancelled flights, diverted flights, missed connections and denied boarding, as well as delayed flights, service standards for airline performance in these other areas would be derived. ” (Sherry et al, 2010) AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011

Reactionary delay § Key factors – time, location, magnitude of the primary delay – Reactionary delay § Key factors – time, location, magnitude of the primary delay – schedule robustness (buffer, a/c & crew util. N, AO model, airport CDM) “Reactionary delay is driven by primary delays. By far the main share of primary delays is related to turnaround delays (airline, airport and other reasons) but there has been a significant increase in the relative share of ANS related in 2010. ” (PRR 2010) § Subtle results of (rarer) European research (Jetzki, 2009) – LCCs, on average, reactionary/primary ratios >1 – c. f. hub-and-spoke carriers <1 (partly due new primary during t/a) – even so, significant outbound delay from hubs rotates back in – high primary delay (proportionally) less likely to attract new primary – proportion of primary delay at hubs > non-hubs AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011

Modelling the network costs Initial results Modelling the network costs Initial results

Initial results § Top-down model – delay trees, by length of primary delay (Beatty Initial results § Top-down model – delay trees, by length of primary delay (Beatty et al, 1998) – weighted by delay frequencies to network ratio of 0. 8 – implicit recovery in the lower multipliers – rotational (88%), non-rotational (12%); in practice, varies by AO – all reactionary delay treated as at-gate § Rotational delay – evenly distributed over rotations, by scenario (‘ 2 x 30<60’, etc) – considered utilisation (ICAO) and curfews, with no spill into next day – narrowbody & widebody models, caps applied – passenger costs assumed to be new on each rotation – avoided double-counting of maintenance at-gate costs – different pseudo crew rosters applied by aircraft type § Non-rotational delay proportionally by traffic volumes AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011

our model: narrowbody and widebody (upper) Ahmad. Beygi et al (2008): US h&s and our model: narrowbody and widebody (upper) Ahmad. Beygi et al (2008): US h&s and p-to-p airline data AGIFORS Airline Operations Conference Caps: B 735, 4 hrs, € 17 k; B 744, 5 hrs, € 106 k London, United Kingdom, May 16 - 19, 2011

our model: narrowbody and widebody (upper) Ahmad. Beygi et al (2008): US h&s and our model: narrowbody and widebody (upper) Ahmad. Beygi et al (2008): US h&s and p-to-p airline data AGIFORS Airline Operations Conference Caps: B 735, 4 hrs, € 17 k; B 744, 5 hrs, € 106 k London, United Kingdom, May 16 - 19, 2011

Cost (k€) total primary B 738 primary delay (mins) AGIFORS Airline Operations Conference London, Cost (k€) total primary B 738 primary delay (mins) AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011

C = f (M, t, pi) C = k √MTOW C = f (M, t, pi) C = k √MTOW

Developing new metrics Developing new metrics

Developing new metrics § Emergent behaviour – emergence is a key concept of complex Developing new metrics § Emergent behaviour – emergence is a key concept of complex systems – multidisciplinary, variety of names, e. g. complexity theory – system properties cannot be explained as a sum of the parts – early example, the structure of H 2 O (John Stuart Mill, 1872) – often involves: networks, systems of systems, non-linear dynamics – still no consensus on various definitional aspects and formalisations – passenger transportation service - emergent design (Sherry et al, 2010) § Applying this to air transport – www. complexworld. eu; SESAR WPE research network (2010) – Innaxis Research Institute (Coordinator), University of Seville, DLR, NLR, AGIFORSUniversity of Palermo, University Airline Operations Conference London, United Kingdom, May 16 - 19, 2011 of Westminster

Developing new metrics k h e node arc a f b l i n Developing new metrics k h e node arc a f b l i n g c j d m ecc(a) = max {d(a, x)} = 5 x Є V(G) AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011 (Harris et al, 2008)

Developing new metrics § Existing metrics, valuable foundations – ICAO and SESAR KPIs (as Developing new metrics § Existing metrics, valuable foundations – ICAO and SESAR KPIs (as we’ve seen) – ratio of reactionary to primary (one measure of robustness) – function of primary duration and time of day (Beatty et al, 1998) – severity, depth, branching, magnitude (Cohn et al, AGIFORS 2007) – generators and sinks for propagation (especially airport nodes) – rôle of turnaround recovery and amount of schedule buffer AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011

Developing new metrics § New metrics, to be developed in ‘POEM’ (SESAR WP E): Developing new metrics § New metrics, to be developed in ‘POEM’ (SESAR WP E): – differentiating minutes from cost – embracing full OD pax centricity (c. f. flight-centric) – taking account of distributions (not just central tendencies) – from graph theory, e. g. eccentricity, centrality, betweenness – careful re. allocation/causality of new primary (Uo. W, 2001; Jetzki, 2009) – node-related (behaviour not independent of AOs; swaps, cancellations) – arc-related (prioritisation / 4 D, load factors, airborne recovery) – including opportunity costs and recovery costs of attained performance § Most important: operationally useful – airlines present today are warmly invited to participate AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011

Thank you airspace-research@westminster. ac. uk Thank you airspace-research@westminster. ac. uk

Stand-by slides Stand-by slides

Reactionary delay revenue delay 1˚ cost + react pax hrs A € 20 000 Reactionary delay revenue delay 1˚ cost + react pax hrs A € 20 000 30 mins € 2 000 € 2 200 120 B € 10 000 30 mins € 1 500 € 2 300 100 B 738 Which delay reduction is ‘fair’ to prioritise? With a focus on … primary delay minutes: A=B certain primary cost to AO: total cost to AO: societal cost: A B A less certain even less certain a ‘challenge’! (US) AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011

Prioritisation revisited revenue delay 1˚ cost + react pax hrs A € 20 000 Prioritisation revisited revenue delay 1˚ cost + react pax hrs A € 20 000 30 mins € 2 000 € 2 200 120 B € 10 000 30 mins € 1 500 € 2 300 100 B 738 § A fair target: average arrival delay per flight, per AO, same for a season? – but: uncongested airspace, average always lower: reduced leverage – average per airport-pair? variability per airport-pair? § Free market: buy a better trajectory? ‘Grandfather’ rights? § Rations? Weights (how control)? Token-based? § Network Manager responsibility? AO-AO CDM? AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011

AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011 Source: Innaxis AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011 Source: Innaxis

After primary rotation After first propagation wave Average delay of delayed flight = 40 After primary rotation After first propagation wave Average delay of delayed flight = 40 mins Average delay of delayed flight = 28 mins Average delay of delayed passenger = 40 mins Average delay of delayed passenger = 46 mins AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011 (=> 1. 6; literature: 1. 6 – 1. 7)

Vo. T Venn diagram Vo. T Venn diagram

LIS–HEL, B 738 (22 minutes delay) with emissions costs, fuel at € 0. 7 LIS–HEL, B 738 (22 minutes delay) with emissions costs, fuel at € 0. 7 / kg no emissions costs, fuel at € 0. 7 / kg AGIFORS Airline Operations Conference London, United Kingdom, Mayno- 19, 2011 16 emissions costs, fuel at € 0. 5 / kg

Annual cost implications, simple example B 738, 22 minutes delay: CIopt compared with CImax Annual cost implications, simple example B 738, 22 minutes delay: CIopt compared with CImax (i. e. recover 19 minutes); 20 such flights in network per day Case Emissions costs Fuel cost 1 Yes € 0. 7 / kg 10 mins 80 € 6. 7 million 2 No € 0. 7 / kg 11 mins 90 € 4. 5 million 3 No € 0. 5 / kg 12 mins 130 € 2. 7 million Optimum recovery Optimum CI Annual loss c. f. CI 300 (Without emissions costs: allows pre-2012 trade-offs. AGIFORS Airline Operations Conference Lido OC is ACARS-enabled: London, United Kingdom, May 16 - 19, 2011 can send CI proposal to aircraft. )

Trade-off stages § Buffers in schedules (strategic cost of delay) – large enough to Trade-off stages § Buffers in schedules (strategic cost of delay) – large enough to absorb expected levels of tactical delay – avoid over-compromising utilisation – probabilistic trade-off, expected residual delay v. ‘sunk’ cost [paper, PRU] § Slot management (pre-departure, tactical) – re-route potential – fuel uplift decision § Airborne recovery (tactical) – dynamic cost indexing AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011 weather (esp. wind, ABN) ATC / ATM cooperation

§ Reactionary effects can improve or worsen (passenger) delay § Air transport is focused § Reactionary effects can improve or worsen (passenger) delay § Air transport is focused on passengers, but the KPIs aren’t AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011

Airspace procedures § Speed control, in European context – used in TMAs (usually with Airspace procedures § Speed control, in European context – used in TMAs (usually with heading & altitude constraints) – very seldom used en route (various studies on this, although not our focus) § Evidence suggests – controllers used to +3% to – 6% (mostly ± 3%) – use of ICAO* > ± 5% rule, “inadvertent changes” = rather unclear * Rules of the Air, Annex 2 to the Convention of International Civil Aviation (1990) AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011

Emissions § No global agreement on aviation regarding emissions – regional competitive distortion; Copenhagen Emissions § No global agreement on aviation regarding emissions – regional competitive distortion; Copenhagen disappointing (Dec 2009) § Aviation contributes 3. 4% of anthropogenic GHGs in Europe – source: European Environment Agency (2006); value is increasing – long-haul flights (> 3 hours), 13% of flights & 60% of fuel burn § ATM accounts for 0. 2% of CO 2 emissions in the EU – some of this is determined by safety margins & runway capacities etc § CO 2 (warming effect; proportional to fuel burn) – EU ETS: extending to aviation (01 JAN 12) based on gate-to-gate fuel – legislation currently: all AOs operating to/from EU surrender permits § NOx (NO & NO 2: warming effect [ O 3] & cooling effect [ CH 4]) – Commission pledged aviation proposal by November 2009 AGIFORS Airline Operations Conference § Aviation will probably be London, United Kingdom, May 16 - 19, 2011 a net buyer; no net CO 2 decrease

Schedule buffer ‘Buffer’ might not be fully free choice AGIFORS Airline Operations Conference London, Schedule buffer ‘Buffer’ might not be fully free choice AGIFORS Airline Operations Conference London, United Kingdom, May 16 - 19, 2011

B 738 re-route example B 738 re-route example