Article Contents
Article Contents

# A note on optimization modelling of piecewise linear delay costing in the airline industry

The author was supported by Grant No. GJJ161113 (2017-2019) from the Education Department of Jiangxi Province, P.R. China

• We present a mathematical model in an integer programming (I.P.) framework for non-linear delay costing in the airline industry. We prove the correctness of the model mathematically. Time is discretized into intervals of, for example, 15 minutes. We assume that the cost increases with increase in the number of intervals of delay in a piecewise linear manner. Computational results with data obtained from Sydney airport (Australia) show that the integer programming non-linear cost model runs much slower than the linear cost model; hence fast heuristics need to be developed to implement non-linear costing, which is more accurate than linear costing. We present a greedy heuristic that produces a solution only slightly worse than the ones produced by the I.P. models, but in much shorter time.

Mathematics Subject Classification: Primary: 90C10; Secondary: 90B06.

 Citation:

• Figure 1.  $S$-shaped piecewise linear curve for Cost vs Delay; the curve is convex first and concave later

Table 2.  Delay ranges and corresponding costs. (The letters (A) to (F) identify the various columns.)

 Range(A) $\Delta_f$ range(B) Cost (C)coefficient Delay cost(D) 1 $0 \le \Delta_f \le d_{1,f}$ $c_{1,f}$ $c_{1,f}\Delta_f$ 2 $d_{1,f}< \Delta_f \le d_{2,f}$ $c_{2,f}$ $c_{1,f}d_{1,f} + c_{2,f}(\Delta_f - d_{1,f})$ 3 $\Delta_f> d_{2,f}$ $c_{3,f}$ $c_{1,f}d_{1,f} + c_{2,f}(d_{2,f} - d_{1,f})$+ $c_{3,f}(\Delta_f - d_{2,f})$ (A) $\delta$ values (E) Requirements (F) 1 $\delta_{1,f} = 1$ and $\delta_{2,f} = \delta_{3,f}$ = 0 $\mu_{2,f} = \mu_{3,f} = 0$ 2 $\delta_{1,f} = \delta_{2,f} = 1$ and $\delta_{3,f}$ = 0 $\mu_{2,f} = (\Delta_f - d_{1,f})$ and $\mu_{3,f} = 0$ 3 $\delta_{1,f} = \delta_{2,f} = \delta_{3,f}$ = 1 $\mu_{2,f} = (\Delta_f - d_{1,f})$ and $\mu_{3,f} = (\Delta_f - d_{2,f})$

Table 3.  Parameters used in the testing of the non-linear cost model

 Number of flights 261 arrivals, 256 departures Max. delay allowed Eight periods (four hours) Perfect capacities 20 arrivals, 20 departures (per 30-minute interval) Time intervals 30 minutes each Cost break points $d_1$ = 2 units of delay, $d_2$ = 6 units of delay Cost coefficients $c_{1,f} = c_f$, $c_{2,f} = (1.5)c_f$ and $c_{3,f} = 0.5c_f$.($c_f$ is the coefficient in the single-range model)

Table 4.  Results of comparing single-range versus three-range cost model (Note. The "＄" used in Rows 7 and 8 refers to a generic monetary unit, not actual amount in US dollars or Australian dollars or any other currency.)

 Model $\to$ Single-range 3-range A 3-Range B 3-Range C (1) Solution how far from optimal ($A_{\rm{S}}$ value) 1.0 $\le$ 1.36 $\le$ 1.33 $\le$ 1.32 (2) Number of delayed flights 85 117 122 122 (3) Time taken to find solution 0.18 seconds 97 mins 6.5 hours 38 hours (4) Sum of the delays of all flights (periods) 518 518 518 518 (5) Average delay (periods) over all flights 1.002 1.002 1.002 1.002 (6) Average delay (periods) only over delayed flights 6.094 4.427 4.2459 4.2459 (7) Objective function value (＄) 32960 38975 37830 37830 (8) Optimal value of the Linear Programming relaxation (＄) 32960 7762

Table 5.  Results of comparing the Integer Programming solution and the Heuristic solution

 Model/Solver $\to$ 3-range C (from Table 4) Algorithm 1 (Pages 10-11) Number of delayed flights 122 121 Time taken to find solution 38 hours 60 seconds Sum of the delays of all flights (periods) 518 518 Ave. delay (periods) over all flights 1.002 1.002 Ave. delay (periods) only over delayed flights 4.246 4.281 Objective function value (＄) 37830 39660

Table 6.  Results of the Heuristic solution with 791 flights and flight connections

 Solver → Algorithm 1 (Pages 10-11) 1 Number of delayed flights 130 2 Time taken to find solution 35 seconds 3 Sum of the delays of all flights (periods) 690 4 Ave. delay (periods) over all flights 0.872 5 Ave. delay (periods) only over delayed flights 5.308
•  [1] N. Boland, A. Ernst, C. Goh and A. Mees, Optimal two-commodity flows with non-linear cost functions, Journal of the Operational Research Society, 46 (1995), 1192-1207. [2] L. Brunetta, G. Guastalla and L. Navazio, Solving the multi airport ground holding problem, Annals of Operations Research, 81 (1998), 271-287.  doi: 10.1023/A:1018909224543. [3] A. Cook, G. Tanner, V. Williams and G. Meise, Dynamic cost indexing - Managing airline delay costs, Journal of Air Transport Management, 15 (2009), 26-35.  doi: 10.1016/j.jairtraman.2008.07.001. [4] A. Cook and G. Tanner, European Airline Delay Cost Reference Values, Report from the University of Westminster UK, Eurocontrol, 2015. [5] A. Cook, G. Tanner and S. Anderson, Evaluating the True Cost to Airlines of One Minute of Airborne or Ground Delay: Final Report, Eurocontrol, 2004. [6] A. Cook, G. Tanner and A. Lawes, The Hidden cost of airline unpunctuality, Journal of Transport Economics and Policy, 46 (2012), 157-173. [7] J. Ferguson, A. Q. Kara, K. Hoffman and L. Sherry, Estimating domestic US airline cost of delay based on European model, Transportation Research Part C: Emerging Technologies, 33 (2013), 311-323.  doi: 10.1016/j.trc.2011.10.003. [8] J. A. Filar, P. Manyem, D. M. Panton and K. White, A model for adaptive rescheduling of flights in emergencies (MARFE), Journal of Industrial and Management Optimization, 3 (2007), 335-356.  doi: 10.3934/jimo.2007.3.335. [9] J. A. Filar, P. Manyem, M. S. Visser and K. White, Air traffic management at Sydney with cancellations and curfew penalties, Optimization and Industry: New Frontiers, Appl. Optim., Kluwer Academic Publishers, 78 (2003), 113-140.  doi: 10.1007/978-1-4613-0233-9_5. [10] W. Gao, X. Xu, L. Diao and H. Ding, Simmod based simulation optimization of flight delay cost for multi-airport system, 2008 International Conference on Intelligent Computation Technology and Automation (ICICTA), 1 (2008), 698-702. [11] A. Gardi, R. Sabatini and S. Ramasamy, Multi-objective optimisation of aircraft flight trajectories in the atm and avionics context, Progress in Aerospace Sciences, 83 (2016), 1-36.  doi: 10.1016/j.paerosci.2015.11.006. [12] S. Ketabi, Network optimization with piecewise linear convex costs, Iran. J. Sci. Technol. Trans. A Sci., 30 (2006), 315–323,379. [13] P. Manyem, Disruption recovery at airports: Integer programming formulations and polynomial time algorithms, Discrete Applied Mathematics, 242 (2018), 102-117.  doi: 10.1016/j.dam.2017.11.010. [14] A. Mukherjee, Dynamic Stochastic Optimization Models for Air Traffic Flow Management, PhD thesis, University of California at Berkeley, 2004. [15] L. Navazio and G. Romanin-Jacur, The Multiple Connections Multi Airport Ground Holding Problem: Models and Algorithms, Transportation Science, 32 (1998), 268-276.  doi: 10.1287/trsc.32.3.268. [16] H. Zolfagharinia and M. Haughton, The importance of considering non-linear layover and delay costs for local truckers, Transportation Research Part E: Logistics and Transportation Review, 109 (2018), 331-355.  doi: 10.1016/j.tre.2017.10.007.

Figures(1)

Tables(5)