Nội dung

Chapter 4 Enroute Flight Operations in Airline Networks The discussion in earlier chapters was focused on airline operations at airports. This chapter extends the discussion of airline operations to a network level with the introduction of enroute flight operations by airlines, focusing on two main areas, namely crewing and fleeting. Key issues of crewing and fleeting are discussed in Sections 4.1 and 4.2, including setting up crew bases, regulation limits of working hours, aircraft maintenance requirements and routing aircraft for maintenance needs in a network. The complexity emerging from the synchronisation between flights, fleets, crew and passenger itineraries is discussed in Section 4.3 from the perspective of airline operations. Section 4.3 aims to address the operational problems of a complex airline network, starting with exploring the complexity in planning airline networks, operational problems emerging from such networks, and how airline network complexity is measured empirically both at the stage of schedule planning and the stage of post-operation analysis as seen in Section 4.4. The issue of delay propagation in an airline network is discussed in the last section of this chapter. Mathematical models are introduced to describe the mechanism of delay propagation between flights among sub-networks of fleeting, crewing and aircraft routing. Managerial and operational implications of delay propagation are examined in depth from the viewpoint of airline scheduling and operational management, shedding some light on how to improve airline operational reliability. 4.1 Fleets and Aircraft Routing Previously in Chapters 2 and 3, the discussion was primarily focused on airline operations at airports including aircraft turnaround operations, passenger flows, goods/cargo/baggage processing, and the operational constraints imposed by airline schedules and the operating environment. In the sense of an airline network, in which airports are modelled as “nodes” and flights between nodes are modelled as “arcs” (or “links”), we are now going to examine airline operations, moving from a previous “node perspective” to a linked and dynamic “network perspective”. To demonstrate how an airline network is formed, we will introduce two major operations that are performed on a network scale, namely aircraft routing and crew   This chapter is partially based on the following publications: Wu (2005); (2006). 118 Airline Operations and Delay Management scheduling, and explain how those “arcs” in the network map of an airline are planned. 4.1.1 Fleet Assignment and Demand Uncertainties Fleet assignment plays a critical role in the overall commercial and operational performance of an airline schedule. The primary task of fleet assignment and hence the various and complex fleet assignment models is to allocate available airline resources (in particular, aircraft capacity) to best match uncertain passenger demands for flights due to airline competition and the local/global economic environment (Clark 2001). The seat capacity of an aircraft is rather inflexible from the schedule planning perspective, although aircraft configuration changes can alter the capacity marginally (but not quickly, say in days). Hence, matching the uncertainties of passenger demand with the fixed aircraft capacities of various fleets is always a challenging task for airline fleet assignment. From historical booking records, airlines can build demand “profiles” on a leg or an origin-destination (OD) basis. A leg basis views the demand of a leg (i.e. a specific flight) as the demand between two airports, while the OD basis adopts a wider view which considers the “direct” demand between an OD pair, as well as other possible forms of “indirect” demand between the same OD pair, via other transit points in a network. Based on demand forecasts and available fleets, an airline schedule is partitioned into a number of sub-networks, corresponding to different fleet types. The collation of flights in all sub-networks comprises the full schedule, while each flight is only assigned to a fleet (a sub-network) without overlapping among sub-networks of different fleets (Bazargan 2004). Fleet assignment optimisation models are usually configured to maximise the profitability of airline schedules by considering stochastic passenger demands and fleet capacity allocation such as the work by Barnhart et al. (1998), Hane et al. (1995), and Jarrah et al. (2000). 4.1.2 Aircraft Routing and Routing Constraints Following fleet assignment, those flights assigned to a specific fleet type/family are assigned to individual aircraft for operation. This assignment is called aircraft routing (also known as aircraft rotation or tail assignment by airlines), because each line of “routing” is identified by the “tail number” of an aircraft (Bazargan 2004). The objective of aircraft routing is to allocate a limited number of aircraft of the same fleet type/family to conduct those flights assigned earlier by fleet assignment. There are three main constraints to aircraft routing: (a) flight continuity, (b) flight coverage, and (c) maintenance requirements. Each line of routing is designed for operation by an individual aircraft and each line comprises a number of flights that are formed in a chronological order, so that flight continuity ensures that the same aircraft can operate those flights in the specified order. In addition, the following two requirements are needed to Enroute Flight Operations in Airline Networks 119 satisfy flight continuity: the destination airport of a predecessor flight is the origin airport of a successor flight; and, the time gap (i.e. the turnaround time) between the scheduled arrival time of a predecessor flight and the scheduled departure time of a successor flight is longer than the minimum turnaround time of a specific fleet type. Figure 4.1 illustrates flight continuity in aircraft routing. On Route X, flight SS001 departs at Airport A and arrives at Airport B, where the next flight on the same route, i.e. SS002 departs. The time gap between the arrival time of SS001 at Airport B and the departure time of SS002 is the scheduled turnaround time of SS002 and is represented by a flat line along the time-line of Airport B. This time gap should be greater than or equal to the minimum turnaround time required to turn around the aircraft at Airport B. The flight coverage constraint in aircraft routing is to ensure that any flight on the timetable is at least (and at most) operated by a fleet type and thus, is “covered” in a route in a sub-network of aircraft routing. Using the example in Figure 4.1 above, eight flights are assigned to this fleet type and form a sub-network which has two routes, i.e. Route X and Route Y. These eight flights are covered by this sub-network and are only in this sub-network. The third constraint in aircraft routing is the requirement for regular aircraft maintenance. There are four major categories of aircraft maintenance, namely A, B, C and D checks in airline jargon. A checks are carried out the most frequently and the interval between checks depends on the type of aircraft, the maintenance schedules provided by manufacturers, and the regulations of Aviation authorities. In Australia, the Civil Aviation Safety Authority (CASA) is the agency which regulates civil aviation and safety related legislations such as Civil Aviation Regulations (CAR) 1988 and Civil Aviation Safety Regulations (CASR) 1998 (CASA 2008a). Under CAR 1988, an airline (which holds an Air Operator’s Certificate) must establish a maintenance schedule that either adopts the manufacturer’s maintenance schedule, the CASA maintenance schedule, or an approved in-system of maintenance by an airline. D Route Y SS004 C SS003 B SS001 Route X SS002 SS010 A Airports Figure 4.1 Time Flight continuity and aircraft routing patterns for an example fleet 120 Airline Operations and Delay Management Aircraft maintenance systems are complex and carefully designed to minimise the risk of mechanical faults of the equipment. In essence, the backbone of a maintenance system is to schedule aircraft for proactive regular inspections, light maintenance, and major overhauls with “labels” such as A, B, C, and D checks (although most modern (post 1990) aircraft only require A and C checks). For aircraft routing, A checks are the most crucial requirement in scheduling, because of the relatively short interval before an aircraft requires an A check. For instance, the modern Airbus A320 family aircraft would need an A check for every 600 flight hours (roughly every six to eight weeks), and a C check for every 6,000 flight hours, or roughly every 18–20 months. Maintenance schedules may also vary among airlines even for the same type of aircraft, depending on airline preferences and the line maintenance schedule. (Airbus 2008). There is also a “cycle” ceiling (a cycle is a pair of take-off and landing activities) for A checks, together with flight hours and elapsed days since the last check. Whichever cycle, flight hours, or day limit comes first, an A check is required and is often done overnight (6–10 hours) at a maintenance base. Since regular maintenance checks can only be carried out at maintenance bases, an important constraint for aircraft routing is to create overnight “stopovers” for aircraft at maintenance bases, so the required checks can be conducted overnight. For our example given earlier in Figure 4.1, if the aircraft that operates Route X requires an A check by the end of the planning period, then Route X needs to connect SS002 with SS010 instead of SS003, so as to route this aircraft to Airport C, where the maintenance base is located. For this new routing arrangement, there will be a long and unavoidable ground time for this aircraft between SS002 and SS010 at Airport A, and consequently this will incur some loss of aircraft utilisation. Alternatively, this aircraft can be scheduled to operate Route Y, which ends the routing period at Airport C (just in time for an overnight A check), and then will also have high aircraft utilisation hours during the routing period. The other aircraft can then operate Route X without any changes to routing plans. However, when the fleet size and the sub-network grows, simple solutions like the aforementioned are not easily available and airlines usually rely on aircraft routing models for optimised solutions (Barnhart et al. 1998; Clarke et al. 1997; Cordeau et al. 2001; Pauley et al. 1998). 4.1.3 Aircraft Routing Network – the Backbone of an Airline Network For airline operations, the network of aircraft routing after fleet assignment is the “backbone” which converts a commercial airline timetable into an operational plan by optimally allocating limited aircraft capacity. Each fleet has a sub-network of aircraft routing such as the one shown earlier in Figure 4.1. The collation of all subnetworks of different fleets forms the aircraft routing network of an airline, which can be significantly large and complex for multiple fleet types and a large number of flights to operate. This is why most large airlines run commercial planning and Enroute Flight Operations in Airline Networks 121 operation software to assist the planning and operation of large networks such as the Sabre AirOps Suite (Sabre 2008). Since an aircraft routing network has delicate schedule synchronisation among flights, ground operations, aircraft movements, and fleets, random operational disruptions may damage this well-synchronised network. An aircraft routing network serves as the backbone of airline operations and is further synchronised with the crewing network that jointly facilitates passenger itineraries among airports. Apart from disruptions due to aircraft mechanical problems, disruptions from other sources, e.g. crew, passengers and the environment, can also disrupt aircraft routing and airline operations. In the following sections of this chapter, we will introduce another layer of the airline network, namely the crewing subnetwork and further discuss the hidden complexity of an airline network as a whole. 4.2 Crewing and Crew Scheduling 4.2.1 Crew Qualification, Safety Working Hours and Regulations Crew are required for safe operation of an aircraft for passenger charter. The enroute flight operation of an aircraft is carried out by two groups of crew, namely the flight crew and the cabin crew. The flight crew are technical crew who have received adequate flight training and have accumulated flying hours in order to safely conduct certain flight operations. Two flight crew are required for most modern passenger charter aircraft, including a Captain in command and a First Officer as a co-pilot. Cabin crew are required for providing services aboard an aircraft and more importantly, cabin crew are also responsible for facilitating safety and emergency procedures. The number of cabin crew needed for a flight depends on the workload required for the level of services provided aboard an aircraft. Often we see that there are less cabin crew for a flight operated by a low-cost carrier (LCC) than one by a network carrier (NC), due to less bundled services provided for passengers on a LCC flight. However, there is a minimum number of required cabin crew aboard an aircraft (charter or regular passenger transport), in terms of cabin crew and passenger ratio for safety reasons. Civil Aviation Orders (CAO) by CASA in Australia set the limit that one cabin attendant is required for every unit of 36 passengers for an aircraft with seat capacity higher than 36 and lower than 216, as seen in section 20.11 and section 20.16.3 of CAO-20 (CASA 2008b). For safety consideration and fatigue management, the working hours of flight crew are regulated by civil aviation authorities around the world. In Australia, the CAO by CASA also regulates the flight time limitations of flight crew in Part 48, which is also known as CAO-48 in the industry. For instance, a flight crew cannot be rostered in excess of 900 flight hours in 365 consecutive days, or 100 hours in 30 consecutive days, or 30 hours in seven consecutive days. However, CAO-48 122 Airline Operations and Delay Management does not regulate the flight time limitation of cabin crew. Instead, cabin crew flight time is constrained mostly by Enterprise Bargain Agreements (EBAs) negotiated between cabin crew unions and airlines. Some example EBAs between Australian carriers and crew unions can be obtained online from the Workplace Authority of Australian Government at http://www.workplaceauthority.gov.au. For instance, Australian Airlines’ (a former subsidiary of the Qantas Group until 2006) cabin crew EBA enforces a limit that the total duty hours for any cabin crew shall not exceed 1,365 duty hours per annum. For a roster period of 28 days for Australian Airlines, the maximum duty hours was 125 hours, and 43 duty hours was the upper limit of duty hours in any consecutive seven-day period. EBAs between airlines and crew unions may be based on flying time limits or duty hour (also known as “credit hour”) limits. For the example of Australia Airlines, flight crew were rostered and paid on the basis of flying hours, but cabin crew were rostered and paid fully on the basis of credit hours. The “credit hours” of a duty for a cabin crew include sign-on/sign-off time, on-duty rest time, flying time and transit time between flights. Thus, crew rostering becomes a complex exercise due to the calculation of credit hours and the limitations imposed by various EBAs. This complexity also introduces some substantial uncertainties for long-term crew resources planning, which is an essential part of airline business, given that crew cost is usually one of the most expensive items (second only to fuel cost) for an airline. 4.2.2 Crew Base and Crew Scheduling Apart from flight time limitations, the location of a crew base has important economical and operational implications for crewing and crew scheduling. A tour of duty (TOD) for a crew member is defined as: starting from the sign-on time at the crew base until the sign-off time at the same crew base. Hence, a TOD may last from one or two days for a domestic crew, to a number of days for an international crew, depending on flight timetables and airline networks. If a TOD involves overnights at ports other than the base of a crew, airlines will incur other crewing expenses such as accommodation, ground transport and meal allowances. Moreover, if crew are paid based on credit hours, then the “length” of a TOD would significantly influence the “cost” of crewing. Therefore, the common objective of crew scheduling is to minimise the total cost of crewing, which includes crew salaries and other crewing expenses. The location of a crew base depends on an airline network, the shape and size of the network (in terms of airport numbers and locations), crew categories (reflecting fleet types) and cost considerations, e.g. crew salaries at different bases. The common practice of establishing a crew base for an airline is to base most crew at its hub airport (or at the headquarter port of an airline) with other crew based at strategic locations, with operational or economical benefits in a network. We will discuss the operational benefits of a crew base later when we discuss crew scheduling. Enroute Flight Operations in Airline Networks 123 Regarding the economical benefits of setting up a crew base, this is often realised from setting up an overseas crew base, where the cost (mostly salary rates) of employing a crew member is significantly lower than the home country of an airline. For instance, the international operation of Qantas Group has a crew base at Sydney (where Qantas is based) and a crew base at Bangkok. Pay rates for locally employed crew at Bangkok are lower than those in Australia, providing significant cost saving for Qantas international operations. In addition, crew based at Bangkok can be assigned to those TODs connecting Southeast Asia and European destinations in the Qantas network, thus providing shorter TODs for crew and a cheaper cost base for crewing. The operation between Southeast Asia and Australia can be both operated by crew in Thailand or local crew in Sydney. There are two major steps for crew scheduling, namely crew pairing and crew rostering. Crew pairing (or crew pattern generation), similar to aircraft routing, is to create a number of crew pairings that are economically efficient (the lowest cost possible) and comply with workplace EBAs and relevant government regulations. Flight crew can only operate one type of aircraft at a given point of time, due to “type endorsement” regulations. Thus an A320-endorsed pilot cannot operate a B737 aircraft, even though those two types of aircraft are similar in size. Cabin crew, on the other hand, are more flexible and can be assigned to various types of fleets, as long as a cabin crew has received the required safety training for certain aircraft types. Following the previous aircraft routing example provided in the previous section, a corresponding crew pairing solution is presented in Figure 4.2, in conjunction with the planned aircraft routing plan. Two aircraft of the same fleet type are needed to operate eight flights. Crew pairing results show that there are three crew pairings that comply with working hour limitations and other constraints specified in EBAs. Pairing No.1 connects SS007, SS008 and SS004, in which an aircraft change takes place between SS008 and SS004 (from Route Y to Route X) and a on-duty rest period is also involved. Pairing No. 2 includes SS001, SS002 and SS003. Due to working hour limits, SS004 on Route X is operated by Pairing No. 1. Pairing No. 3 includes SS009 and SS010, which are from the second half of Route Y. We can see from this example pairing solution that at least three pairings are required to “cover” all flights (flight coverage), and the flight “continuity” requirement is also reserved within individual crew pairings, just like in aircraft routing. In addition, there is a close relationship between aircraft routing plans and crew pairing generation, because the construction of crew pairings is based on aircraft routing solutions. The close synchronisation of these two networks will be further discussed in the following sections of this chapter. Crew pairings, so far, are not yet assigned to individual crew members. This assignment only occurs during crew rostering, which is the process of building up detailed rosters for individual crew within a specific roster period. A TOD for a crew may contain a few pairings over a number of days. A TOD, by definition starts and finishes at the crew base and can not exceed the limit of working hours in seven consecutive days. After a TOD, a required minimum rest time (in hours Airline Operations and Delay Management 124 D Route Y C SS007 SS004 Pairing 1 SS008 SS003 B SS001 Route X A Pairing 2 Airports Figure 4.2 SS002 SS009 Pairing 3 SS010 Time Crew pairings in conjunction with aircraft routing or days) is scheduled to allow crew to recover from duty, especially if a TOD involves long-haul flights that have more than a four-hour time zone difference from the crew base. Each crew receives a detailed roster for a roster period, and this roster is also the basis on which crew salary is calculated (either on flying hours or on credit hours). For most EBAs in the airline industry, a minimum of paid flying hours or paid credit hours is often required, so a minimum wage is guaranteed for any roster periods. 4.2.3 Crewing Network and Synchronisation with Aircraft Routing Network It is seen in Figure 4.2 earlier that the crewing network for a specific fleet is closely synchronised with the corresponding aircraft routing network. Due to this synchronisation, aircraft routing and flight operations are influenced by the reliability of crewing networks, and vice versa. Delays may occur to the crewing sub-network, due to sickness, late crew connections between flights, or delays due to passenger connections at airports. For instance, if flight SS008 is delayed due to passenger check-in and baggage loading, then SS009 may be delayed because both flights are on the same route, i.e. Route Y, which is operated by the same aircraft. Crew operating for Pairing No. 1 can be late (due to late SS008) for SS004, affecting the operation of Route X by another aircraft. 4.3 Complex Airline Networks 4.3.1 Complex Network Operations Flights are connected in an airline network (as well as between the networks of different airlines for code-sharing operations) through four types of “flow networks” including: the passenger itinerary network (including passenger baggage transport), aircraft routing network, cargo/goods shipping network, Enroute Flight Operations in Airline Networks 125 and the crewing network. While an airline designs a network and accordingly the aircraft routing and crewing in the network, passenger itineraries and cargo/ goods shipping are influenced and driven by stochastic market demands among the airports in the network. Earlier in Chapters 2 and 3, we briefly discussed the complexity of airline operations at airports. This complexity mainly comes from the time pressure of conducting aircraft turnaround operations, passenger (and baggage) transfer among flights, and crew connections among flights at an airport. On a network scale, the complexity of airline operations extends from operations at individual airports to the operations of flights between airports in a network. Unlike the operations at airports, operations in a network cannot be fully managed by focusing only on individual airports, because operations at an airport may cause disruptions to operations at other airports in the same network, due to “flows” in the network. This complexity is best illustrated by Figure 4.3. A/C 1 A/C 2 A (SS020) B B (SS030) C C D (SS100) B B (SS070) E E F (SS210) B A/C 3 A/C 4 H (SS460) B (SS080) B G B (SS090) G J J Aircraft Routing Layer A/C 1 A/C 2 A (SS020) B B (SS030) C C D (SS100) B B (SS070) E E F (SS210) B A/C 3 H A/C 4 (SS460) B (SS080) B G B (SS090) G J J Passenger Itinerary Layer A/C 1 A/C 2 A/C 3 A/C 4 A (SS020) B B (SS030) C C D (SS100) B B (SS070) E E F (SS210) B H (SS460) B B Crew Connection Layer Figure 4.3 Layers of an airline network (SS080) B (SS090) G G J J 126 Airline Operations and Delay Management On the “layer” of aircraft routing in Figure 4.3, flights on a timetable of an airline are operated and connected by individual aircraft of different types, depending on the result of fleet assignment and aircraft routing. In operations, delays to a flight may cause delays to other flights, if those flights are operated by the same aircraft. For example, aircraft No. 1 (denoted by “A/C 1”) operates flight SS020 from airport A to airport B (which is the hub airport in this example), then SS030 from airport B to airport C, and so forth. In the example network illustrated in Figure 4.3, there are four aircraft in the routing layer of the network and there are four flights from spoke airports (A, D, F and H) to the hub airport (B) during the inbound “wave” of the schedule. On the layer of crew connection, the crew (both flight crew and cabin crew) follow aircraft routing and airline schedule, as well as working under the safety regulations of aviation authorities and the limits of individual airline EBAs. Working conditions for crew are stricter than for aircraft routing for safety and fatigue management reasons. For example, we see from Figure 4.3 that flight crew transfer from flight SS100 to flight SS080, while cabin crew connect from SS460 to SS080 during aircraft turnaround time. Since crew arrive with the aircraft and need to connect to other flights (according to crewing plans) during aircraft turnaround time, it is possible that outbound flights, e.g. SS080 may be delayed due to late connecting crew from SS100 and SS460. The “sit time” required for a crew connection at an airport is usually longer than the minimum turnaround time of an aircraft; however, it is still possible that a flight incurs departure delays due to late inbound crew connections. The passenger itinerary layer, as shown in Figure 4.3 is the most complex one in an airline network. In a strong hub-and-spoke network, an inbound flight may bring passengers who will connect to a number of outbound flights at a major hub airport. Similarly, an outbound flight, e.g. SS070 as illustrated in the example, may receive connecting passengers from a number of inbound feeding flights such as SS020, SS100, SS210 and SS460. In a typical “wave” of inbound flights in a hubbing schedule, there could be ten to twenty flights, depending on the intensity of hubbing by an airline (Burghouwt 2007). Inbound flights carry passengers with different itineraries to a hub airport where these passengers can then connect to different outbound flights according to their planned travel itineraries. The higher the hubbing intensity of a schedule at a hub airport (i.e. more inbound/outbound flights in a wave), the more connection opportunities a schedule can create at the hub. Furthermore, the checked-in baggage travelling with passengers also need to “connect” between flights. Depending on how transfer baggage is handled at an airport, the minimum connection time between flights also needs to consider the time required for baggage processing. Hence, it is likely that passengers may be able to transit to an outbound flight even if an inbound flight is delayed, but the passengers’ baggage is left behind waiting for the next connecting flight to the same destination. This will surely cause travel inconvenience to passengers and extra operating costs to an airline.