Here is an overview of the conference program:
Abstracts in alphabetical order of author. Maintenance presentation abstracts behind operations.
Assessment of flight delay propagation
Oleksandr Basanets, Jeppesen
Propagated flights delay is one of the main sources of aircraft schedule disruption. Therefore, it might increase: utilization of the reserve crew, passengers’ misconnections, maintenance schedule disruption, and so on. We analyze known qualitative and quantitative methods for flights delay propagation assessment. In addition, we propose a new quantitative method, based on Markov chain approach.
How the pandemic has revolutionized the approach of Airlines to Manpower Planning and Flight Scheduling
Franziska Burmester, WePlan
Old traditional approaches to assessing demand and managing crew supply are reaching their limits. In order to keep up with the new speed of decision-making and complexity, new technologies and fast optimization methods are needed to evaluate the various scenarios at place. Fast movers will be the winners of the pandemic’s aftermath. This new speed will most likely stay and legacy processes have to be adapted. We show where the process across airline ops planning teams needs to adapt as a result of the pandemic and which solutions help to manage the ramp-up.
Prediction of reactionary delay and cost using machine learning
Paolino De Falco, University of Westminster
Estimating the potential propagation of delay (and its associated cost) is crucial for the management of the fleet the day of operations. As the flight departure approaches, these estimations need to be as accurate as possible to support pre-tactical and tactical interventions aiming at minimising the network disruption. Dispatcher3 project aims at the use of machine learning models for the prediction and analysis of differences between planned and executed flight plans. Models corresponding to different time-horizons (from departure time to 15 hours before) are developed to predict the flying and turnaround times. These models are combined (with other operational parameters) to estimate reactionary delay and associated expected costs (including potential breaching of curfews). The outcome (delay and costs) is presented as probability distributions. This information could be used to support the flight planning process and the management of the fleet.
Optimizing Tail Assignments Across Aircraft Families
Kadir Göcer, Swiss International Air Lines Ltd.
The tail assignment problem typically focuses on assigning tails to aircraft rotations within a predetermined aircraft type or family. This family is chosen to satisfy passenger demand at the minimum operating cost. However, Covid-19 has made passenger demand exceptionally unpredictable, and the ideal family of aircraft to assign to a rotation may not be known until the day of operations. Changing the family of aircraft assigned to a rotation means crew assignments must also change, and these changes are subject to complex and varied contractual and operational requirements. This talk proposes a set of tail assignment approaches that use a black-box crew feasibility checker to allow swapping tails across different aircraft families. The talk will also focus on the modeling and engineering techniques used in the tail assignment solver to improve solve times and system robustness required by these algorithms when repeatedly solving tail assignments to find crew-feasible swaps.
Credit-based mechanisms for user-driven prioritisation during ATFM regulations
Gérald Gurtner, University of Westminster
Air traffic regulations are a major source of delay in Europe. Due to airspace congestion, or other disturbances, they lead to flights being delayed at departure airports, by assigning explicit time periods (slots) within which to depart. The current algorithm to assign slots is called CASA and is agnostic to flight characteristic. For instance, potentially costly flights with many connecting passengers are treated in the same way as ‘cheap’ secondary flights. Hence, there is a need for a slot assignment mechanism taking airlines’ cost into account. We would like to present some concepts developed during the SESAR exploratory research project BEACON that go beyond the CASA algorithm and its successors (the User Defined Prioritisation Processes) to take airspace users’ needs into account. We will present credit-based mechanisms, including an auction, and discuss their economic efficiency as well as their operational feasibility.
COVID 19 Impact on Aircraft Ground Time at Congonhas Airport- Sao Paulo Brazil (CGH)
Leila A. Halawi, Embry-Riddle Aeronautical University
This project aims to research and analyze the impact of the increase of ground time at Congonhas Airport caused by the new safety procedures in aviation to avoid the spread of the COVID-19 virus. Aviation is a highly regulated industry, with the high cost and tiny margins. It focuses on optimizing one of the most expensive assets, the aircraft. To keep aircraft flying is one of the airline's primary goals as aircraft on the ground means no revenue generation and cost increase. Congonhas Airport is one of the busiest airports in Brazil and already operated on its limited capacity before the crisis; higher ground time will bring operational and financial impact to the airlines, airport administrators, and suppliers. We selected the public database from ANAC and ABEAR that contains detailed data about arrivals and departures at Congonhas Airport to analyze the impact. We also used available data from INFRAERO about the capacity and operational restrictions of the airport. In the first phase of the analysis, we worked with Python to measure the ground time duration before and after implementing the new procedures and validating the ground time increase hypothesis. The second phase was performed using Arena Software to simulate the airport's movements, considering airlines' ground time before and after implementing the new procedures. The results from both analyzes were complementary and allowed to confirm the hypothesis and identify the impact of the increase in the ground time due to the new cleaning and social distancing measures adopted during the COVID-19 pandemic.
Opportunities in a global crisis: Setting the course towards the next generation of disruption management systems
Dr. Sebastian Heger, M2P Consulting GmbH
Furthering technological innovation right now is one of the key success factors to unlock post-crisis growth – despite the massive challenges inflicted on airlines and vendors by Covid-19. Especially in the area of airline operations control and disruption management there is an urgent need to adopt more efficient technologies available already today. The obsolescence of today’s disruption management tools and practices becomes particularly obvious if we compare them to technologies used in daily life: While modern smart phones mostly feature the latest software applications from a cloud-based ecosystem, airlines still use on-premise installed legacy systems with less flexible interfaces and old-fashioned architectures. Likewise, scalable business models such as “Software as a Service (SaaS)” have not yet made their way into airline disruption management – “Disruption Recovery as a Service (DRaaS)” or “Pay per disruption” could jumpstart this function to the next level. This talk will thus present our perspective on how legacy disruption management practices cannot only be improved today, but also how to proceed to a next generation for disruption management tomorrow.Protecting Crew in Recovery Management
Xianfei Jin, Sabre
The airline recovery problem is usually solved to help airlines repair the disrupted flight schedules and reassign crew to operate the modified flight schedules close to the days of operations. For most airlines, it is commonly conducted sequentially due to the company resource management constraints, planning process, and complexity of the problem. First, reschedule flights then, recover the crew. The challenge is how to optimize the flight reschedule while considering the factors from the crew assignments, passenger connections, and other operational constraints. In the first part of this presentation, we will discuss the schedule recovery approach with capabilities of considering crew qualification, connection, layover, and duty legality factors. In the second part, we will introduce the subsequential crew recovery process with comprehensive crew considerations and solutions in solving real-world disruptions. As a result, our crew-friendly schedule recovery decisions are designed to facilitate overall crew recovery efficiency.
Using Adaptive Simulation Optimization to determine an optimal runway usage strategy
Kunal Kumar, KLM Royal Dutch Airlines
COVID-19 proved again that having the right people for the right job is crucial to be successful in the airline business. The flexibility and creativity of our people enabled us to adapt quickly and decisively to the rapidly changing environment. As Operations Decision Support (ODS) department, we implemented new features, new dashboards and even completely new optimizers to answer the ever-changing questions from our network planners and operational control staff in record time. We would like to take this opportunity to talk about one example of the engineering culture within KLM ODS that enables this: our bi-annual hackathon. Even within COVID times we decided to continue this tradition and commit our entire department in a weeklong competition to solve a problem that is relevant and challenging. During our talk you will learn why and how we organize these hackathons and how the winning team leveraged Adaptive Simulation Optimization to truly dazzle the challenger (LVNL - Dutch Air Traffic Control) with their solution.
Turn Operations in the Post-pandemic era
Nikita Rana, United Airlines
Aviation is one of the key industries that has been deeply impacted due to the pandemic. Pandemic posed major challenges not just in terms of low demand but for the overall operations of airlines. The focus for the turn operations has now shifted from maximum operational efficiency to maximum safety and social distancing for customers. United Airlines’ operations planning and execution focus is to put the customer at the center of everything we do. United Airlines introduced rigorous safety and cleaning procedures focused on delivering a new level of cleanliness on the ground and in the air as part of United CleanPlus program. Various initiatives have been introduced to ensure cleaner environment at check-in, at the gate, on board (will be discussed in detail)
The additional cleaning and social distancing measures resulted in an increased aircraft turnaround time. Our MSTs (Minimum service time) were increased over the period of last year by 20-30 minutes across all our fleets. Increased MSTs also drove an increase to MGTs. We also notice increased turnaround times as the seat factor started rising towards 2020-year end. This led to further adjustment to scheduled ground times. Apart from additional cost due to increased ground times, increased fuel burn due to APU (Auxiliary Power unit) run during deplaning and boarding to ensure the air flow volume for HEPA (high-efficiency particulate air) filters, United also faced additional challenges in terms customer handling, third-party related delays. As seat factor continue to rise in 2021, these challenges present new opportunities for innovations and efficiencies in the turn operations.
Continuous Dynamic Scheduling and the Change Management required of Airline Functional Areas
Edward L Stephens, Boeing
The challenges posed by the COVID19 pandemic to the passenger and cargo markets have resulted in an unprecedented reliance on short- fused planning and execution by airlines/operators planning and operating functions. The opening of new cities within weeks instead of months, weekly shifts in capacity, and continuous re-fleeting has been unprecedented in the COVID era. This forced innovation has renewed airline confidence and stakeholder willingness to schedule dynamically in order to meet short term demand, add capacity during peak seasonal periods, and to manage major disruptions. Operators have demonstrated their nimbleness by dynamically adding, drawing down, and recovering capacity ahead of, during, and after major disruptions. These techniques are not new, but the timelines, scope, and willingness of airline and operator stakeholders to embrace these methods must become the new normal. This presentation will explore how effective business process change management will enable nimble operators to realize the benefits of dynamic scheduling or any business reason and avoid the pitfalls of short fused schedule changes.
Resilient Airline Operations
Goran Stojkovic’ Boeing
Adapting to the rapidly changing COVID-19 environment is a key survival criteria for airlines. Increasing availability of data and increasingly smarter algorithms are powering real-time analytics and resource optimization capabilities. To run resilient operations, airlines need new tools to be able to quickly identify anomalies in operations, manage unpredictable en-route events and complex turnaround processes, cope with high numbers of groundings, continuously change schedules due to demand volatility, and optimize fuel & crew cost due to surplus/lack of resources. At the same time, new tools must be easy to deploy, quickly available, and show immediate value. In this presentation we showcase several COVID-19 IROPS opportunities to integrate publicly available and Boeing proprietary data, and develop new analytical methods and tools. By exploiting enriched datasets and novel analytical methods, developed tools are helping airlines improve resilience by running their business safely, more cost-efficiently and competitively in a highly volatile and heavily regulated environment.
Mixed Signals - Predicting Airport Delays in a Year of Transition
David White, Cirium
Building an accurate and reliable airport delay prediction algorithm during the COVID-19 recovery phase requires a balancing act using training data from times of both normal and wildly abnormal traffic patterns. This presentation will review the impact of COVID on on-time performance, designing a model capable of expressing uncertainty, outline methodologies for building a system capable of identifying changing conditions and delivering consistently accurate predictions.
Using machining learning to handling connection based turn time in flight schedule recovery problem
Yichen Yang, Sabre
Time-space networks is widely used for the milp model of the flight schedule recovery problem. However, the classic Time-space network is hard to deal with connection based turn times, for example variable minimum ground times. Therefore, we need to perform network reconstructions before building the milp. Traditionally, a optimization model is built for addressing this issue per equipment per airport level. The challenge is that such process is required for hundreds or thousands of times depending on the problem size. It eventually takes quite a long time to get a solution. To speed up this process, we train a graph attention network with reinforcement learning. The experiments results shows that the machining learning based algorithm can finish the network reconstructions very fast and with equivalent or better quality.
Solving flight combining and rerouting problem with best first search algorithm
Sudhibhum Yaowiwat, Royal Thai Naval Academy
According to COVID crisis, human may not be able to solve irregular airline operation properly, several techniques can be used to solve irregular flight scheduling problem automatically. Those techniques are linear programming, network modeling, heuristic search, decision support system, Genetic Algorithms and Multi Agent System. The problem can be solved automatically by implementing those techniques into software program. This presentation aims to introduce an alternative approach based on Best First Search algorithms
Predictive maintenance for a fleet of aircraft with Remaining-Useful-Life prognostics and the management of spare components
Ingeborg de Pater, Delft University of Technology
Aircraft maintenance is key for safe and efficient airline operations, with airlines spending approximately 9% of their total operation costs on Maintenance, Repair and Overhaul, which in 2018, was estimated to be 69 billion dollars . Striving for cost savings, aircraft maintenance is currently shifting from corrective or preventive maintenance towards predictive maintenance. For predictive maintenance, sensors are continuously monitoring the health of components and systems, algorithms are generating Remaining-Useful-Life (RUL) prognostics, and maintenance is performed based on these prognostics. One of the main challenges of predictive maintenance is to obtain Remaining-Useful-Life (RUL) prognostics for systems and components. Equally challenging is to integrate RUL prognostics into the aircraft maintenance planning, while the entire complexity of this process is taken into account: the management of spare components, the availability of maintenance slots during which the aircraft can be maintained, and system reliability requirements.
In this presentation we discuss an integrated approach for predictive aircraft maintenance planning for a fleet of aircraft, each equipped with a multi-component system. First, we discuss the development of model-based Remaining-Useful-Life prognostics. Then, a rolling horizon approach is introduced for the maintenance planning of a fleet of aircraft with multi-component systems. This model integrates the Remaining-Useful-Life prognostics with the management of a limited stock of spare components, available maintenance slots and system reliability requirements.
Leveraging Analytics to Create a Dynamic Technical Operations Strategy for Storing Aircraft and Returning Them to Service
Lynn Garrett, United Airlines
In an uncertain environment, it has been critical for United to maintain a flexible approach in its return-to-service strategy. In this talk, we will discuss the ways in which analytics allowed the TechOps team to act strategically to manage tradeoffs between cost, reliability, and flexibility while an aircraft was in storage and when building a return-to-service (RTS) plan. To decide which maintenance work to clear while an aircraft was in storage, the TechOps Data Analytics team formulated a Maintenance Debt Score to strategically clear workload to reduce the labor burden when the aircraft is returned to service. To formulate a return-to-service plan, we built an optimization model that recommends which aircraft should be returned to service each month. The model minimizes costs including engine operating and maintenance costs, airframe and landing gear maintenance costs, line maintenance costs, and expected future delay or cancel costs, while ensuring that we have enough aircraft to meet the fluctuating network demand plan.
Multi-objective analysis of condition-based aircraft maintenance
Juseong Lee, Delft University of Technology
When considering condition-based maintenance for aircraft, multiple objectives need to be analyzed: the delay caused by the maintenance, the duration of aircraft-on-ground, the number of flights that do not comply with the regulations. Reducing the delays and the aircraft-on-ground time is a key goal in aircraft maintenance as these may cause high costs. Apart from these cost-related objectives, additional objectives for g such as the number of component failures, the number of incidents, the number of maintenance tasks, and the mean cycles to replacements (MCTR) are of interest.
In this presentation, a simulation-based methodology is introduced to analyze multiple objective when considering condition-based maintenance for aircraft. The methodology takes into account the detailed process of aircraft maintenance: operating the aircraft based on dynamic flight schedules, generating maintenance tasks, analyzing condition data, planning and executing the maintenance tasks.
The proposed methodology is demonstrated for the maintenance of aircraft landing gear brakes. The brakes degrade over time due to friction wearing. Traditionally, the degradation of brakes is inspected at fixed time intervals. In this study, we consider CBM strategies for the brake maintenance, where a part of the periodic inspections is replaced by sensor monitoring. Using simulation for CBM strategies, the delay caused by the maintenance, MCTR, the duration of aircraft-on-ground, the number of flights that do not comply with the regulations, are analyzed. Overall recommendations on the design of CBM strategies for the aircraft landing gear are provided.
Airline maintenance task rescheduling in a disruptive environment
Paul J. van Kessel, Bruno F. Santos, Floris C. Freeman - KLM, Delft University of Technology
Scheduling of airline maintenance tasks takes place in a disruptive environment. The stochastic arrival of corrective maintenance tasks and changes in both fleet and resource availability require schedules to be continuously adjusted. An optimal schedule ensures that all tasks are executed before their due date in both an efficient (at minimum use of ground-time) and stable (limited number of schedule changes) manner. This is the first study to address disruption management for the hangar maintenance task scheduling problem, proposing a practical and efficient modelling framework. The framework comprises a mixed integer linear programming model for airline maintenance task rescheduling in a disruptive environment, in which task scheduling is constrained by the availability of resources. The capabilities of the model include both creating and adjusting maintenance schedules in a continuous manner and dynamically reacting to new information when this becomes available. The modelling framework was tested in a case study provided by a large airline, and its performance was compared with the hangar maintenance schedule created by the airline itself. The results show that the proposed approach produces more efficient and stable results. A 2.95% ground time decrease can be achieved, while the number of schedule changes in the last days before operations is decreased by more than half.
A Knowledge and Data Driven approach for Air Cycle Machine Prognostics
Changzhou Wang, Hamid Nikjou, William Hanna and Mark Mazarek - Boeing Global Services Engineering
The Boeing 787 Air Cycle Machine (ACM) uses air bearings, which may degenerate as the ACM ages or may suffer mechanical damage. ACM build tolerances have an effect on machine performance, which leads to subtle variances in sensor data. Without detailed analysis of large amount actual flight sensor data, it is difficult, if not impossible, to identify degradation evidence solely using engineering knowledge. On the other hand, ACMs are designed to be highly reliable and we observe very few failures. As a result, a pure data-driven approach cannot predict failures due to the lack of sufficient training data containing failure cases. We have created an approach to combine both engineering knowledge and observed data to enable ACM prognostics. Our approach starts with collecting and preparing full flight sensor data and the ACM removal data. Based on engineering insights, intra-flight and inter-flight features are created. Machine learning methods are then used to explore the large feature spaces and identify critical logic. Finally, the best machine learning model is simplified into prediction rules under the guidance of engineering knowledge. Our prognostics logic has been validated on historical data and deployed for routine ACM monitoring for seven airlines, and has successfully predicted more than ten ACM performance degradations, prevented schedule interruptions and significantly reduced repair costs.
Leveraging Advanced Analytics to Enhance the Line Maintenance Manpower Modeling
Patrice Yapo, Karthik Manohar, Jose Ramirez-Hernandez, American Airlines Inc.
Aircraft Maintenance could be broadly classified as Line and Base maintenance. While Base Maintenance procedures require a hangar and aircraft to be out of service for extended periods, Line Maintenance (LM) is carried out more frequently during turnarounds through the operating day and aircraft that remain overnight on the ground. In this presentation, we would provide an overview of our LM manpower model, that forecasts the headcount requirements for each Line Maintenance station in the American Airlines network based on different drivers like Flight schedule, check volume forecasts, etc. and by leveraging advanced analytics to provide accurate headcount forecasts within short turnaround times. For instance, automated data pipelines help to reduce manual spreadsheet work, optimization techniques help to effectively utilize line maintenance footprint, and statistical models predict the headcount need to support aircraft out of service work. An important benefit of this process is moving away from using anecdotal information to a data-driven analytical approach to support better decision making.