Due to the nature of the Crew Recovery Problem, it is important to use mixed models containing the sophisticated methods because of the need to find a fast and applicable solution. A significant speed advantage will be achieved by hybridizing machine learning and classic optimization, which contain very important developments recently.
With the model that is implemented within the scope of our study, intuitive methods that transfer the experiences of people who are in the aviation sector, who work in the crew operation and who make manual interventions during the operation, into the digital environment. Subsequently, alternative mathematical models and decomposition methods is used to accelerate these methods. Thus, it is ensured that deep learning based methods search the solution space with an intelligent strategy so that they can quickly give the near-optimal solutions.
Within the scope of our study, it is seen that modeling past recovery actions with deep learning techniques will make the model faster and more effective while solving the Crew Recovery Problem.
Due to the irregular nature of flight operations, airlines need to take a range of actions to recover their aircraft and crew schedules. The limited time frames prevent airlines from using a full-scale optimization approach. Consequently, airlines typically apply recovery solutions that can be far from optimal.
This study proposes a practical method that combines machine learning and optimization to find improved recovery solutions. Our procedure is based on the idea that the most effective constraints to add to the recovery models without sacrificing the solution quality, can be determined in advance by leveraging the similarities between disruptions. Our experiments show that, this approach can reduce solution time significantly while still achieving high-quality solutions.
Fatigue from multiple sources, such as scheduling, workload, COVID-related precautions, and human factors, creates a compound safety risk. The Sleep, Activity, Fatigue, and Task Effectiveness Fatigue Avoidance Scheduling Tool (SAFTE-FAST) has recently added a feature called Insights that can identify proposed schedules with high compound fatigue risk.
Fatigue related to predictable workload factors were identified from a survey of medium-haul pilots conducted by a major European airline. SAFTE-FAST workload settings were modified to reflect the amount of fatigue associated with each workload factor, and a use case was developed to compare schedules identified as having high compound risk by Insights against fatigue reports from those rosters.
The Crew Rostering Problem (CRP), consists in determining an optimal sequencing of a given set of crew pairings to create rosters, satisfying a series of operational constraints coming from technical and legal regulations (e.g., government, company, labor union). This problem, considered to be NP-hard, has been largely studied by the operational research community and several good solution approaches (from an industrial point of view) can be found in the literature.
The line-training rostering problem (LTRP) is an extension of the classical CRP where the objective is to determine an optimal sequencing of crew pairings to create simultaneously compatible rosters for both trainees and instructors. In this context, it is imperative to consider a series of pedagogical constraints, such as checklist of requirements for each trainee and linking constraints between trainees and instructors’ schedules. The LTRP is also NP-Hard and its effective resolution using exact methods can be prohibitive in an industrial context.
In this work, the math-heuristic approach developed at Air France to handle the LTRP is presented, which employs a novel algorithm for trainee rosters generation based on a Randomized Greedy Algorithm, allowing to have personalized trainee rosters, and an extended version of the Mixed-integer-linear programming CRP mathematical model, which considers the instructor / trainee coupling constraints, among others. The proposed approach allows creating cost-effective line-training rosters in under an hour of computing time, compared to at least one week of hand-made schedules.
Boeing recently launched the 777-8 Freighter, expanding its market-leading 777X and freighter families. The 777-8 Freighter will be the world’s largest, longest-range and most capable twin-engine freighter. With payload capacity nearly identical to the 747-400 Freighter and a 25% improvement in fuel efficiency, emissions and operating costs, the 777-8 Freighter will enable a more sustainable and profitable business for operators.
Globally operating airline companies face significantly low rates of crew standby utilization during the operation when applying the conventional methods of standby planning which increases labor costs substantially. Researches and surveys have shown that airlines make use of around 50 percent of their standby capacities, especially for cabin crew. This study aims to investigate a data-driven solution by analyzing historical standby data combined with operational flight data. Based on the business rules and constraints in the airline business, it asks: Is there a machine learning approach to reveal the patterns in the historical operational data that facilitates better crew standby scheduling than the classical approach? In this context, crew standby is defined as the extra crew resources that are utilized by crew management to manage operational disruptions during the day of operation, which is a way of unforeseen risk mitigation.
Based on a review of the literature on crew standby scheduling and airline disruption management, anonymized operational data for a typical airline was created containing all flight activities combined with assigned standby activities for cabin crew. In this way, the anonymous dataset imitated the distribution of real operational data considering scheduled flight network and assigned standby activities. Then, data pre-processing and feature engineering was applied to the raw dataset by coding with python libraries for data science. The machine learning model was evaluated in the Weka data mining toolkit for training and testing the model applying a multilayer perceptron algorithm, which is a type of artificial neural network. Hyperparameter tuning and neural network structure were built in the Weka environment for the final model. Evaluations on the training and testing datasets indicate that a multilayer perceptron algorithm can predict which flights need a standby and how many cabin crews with an accuracy of 95 to 98 percent, which is far better than the classical way of standby planning. Both one and two hidden layered neural network models can reveal the patterns in the pre-processed dataset with different building time performance.
The results indicate that a typical airline can deploy a machine learning approach in its crew standby scheduling operation also by applying the risk mitigation of holding a “buffer standby” mechanism. In this way, one of the main cost items of a typical airline could be optimized by applying a data-driven approach to balance conflicting targets of productivity of crews and operational stability. On this basis, it is recommended that a typical globally operating airline should use data analytics and machine learning as a key technology to optimize their business models and differentiate itself amongst their competitors by identifying its actual cost optimum.
We would like to share our story - How Qatar Airways tackled the unprecedented operational challenges during COVID situation and our learnings from overcoming the situation and how it impacted our ways of working going forward.
Vikas Rameshan, Lead Business Analyst – Flight Ops IT
I have got overall 19 years of industry experience, with last 8 years in QR leading the functional stream of Flight Operation System Program. Prior to QR, I was working with a leading product company in the area of Airline Crew Management and has handled product implementations across various airlines such as KLM, BA etc. I completed Master Of Computer Application from Bangalore University, India in 2005.
Arunchandran S R, Lead Business Analyst – Flight Ops ITCertified Product owner with 15 years of industry experience, with last 8 years in QR leading the functional stream of Maintenance Planning, Tail assignment, Crew management etc. I have started my carrier with a leading product company in Airline product development and has handled implementations across various airlines such as KLM, TUI etc. I have completed B.Tech in computer Science and Engineering from Mahatma Gandhi University, India in 2006.
Human Resources management is the first step to be taken in order to achieve company goals and managerial success in the most economical way. Human resource management (HRM); It is the process of recruiting and training people, developing relevant policies, and developing strategies to retain them. The first and most important condition for success in Human Resources Management; It is the determination of the personnel that the company will need in the future in terms of quality and quantity and determining how and to what extent this need can be met.
Airline companies have a wide range of activities. Human resources management is quite complex due to the need for personnel who can perform different tasks in order to carry out these activities completely. Crew costs are one of the biggest expenses for airlines and therefore effective manpower planning is essential to maximize profits. One of the most important issues for all airline companies is the need for crew. It is of great importance to be able to fully plan the crew needs, both to make the flights smoothly and to keep the crew costs at a minimum.
This study proposes the scope of the engagement of doctoral studies in Turkish Airlines manpower planning procedures. This doctoral study is in the early stage of research thesis in data mining. Topic of research thesis is Decision Support System for Manpower Prediction Using Machine Learning Algorithms.
Basically, manpower planning is a supply-demand problem. The current manpower calculation work is done manually via the Microsoft Excel program. The biggest risk of the current working order is that it is prone to error due to excessive use of manpower and manual work in business units. The aim of the project is to achieve the goals by establishing a decision support system with the help of machine learning.
One of the main factors taken into account when calculating the Manpower is non-flight preoccupations. some of these preoccupations are parameters to be estimated from historical flight data with the help of machine learning.
Our experience with existing and potential clients shows that one of the biggest challenges in coping with the massive increase in the market is to provide the training capacity in a timely manner.
In this session, we would like to share our joint experiences with our client Eurowings which WePlan supported during the ramp-up process in 2022.
Nicole Süß, Head of Crew Capacity & Ops Planning, and Franziska Burmester, Founder of WePlan, will give insights into their experiences and learnings that were faced during the joint project and what measures have been taken for next year – including digitalization, but also organizational and procedural improvements.
Main focus will be on the strong dependency between capacity planning and training resources and how digitalization can help overcome the complexities of creating optimal and timely training plans.
Nicole Süß is employed as Head of Crew Capacity & Ops Planning at Eurowings aviation. Eurowings GmbH is a German airline. It is a subsidiary of Lufthansa and has been bundling its flight offerings away from its Frankfurt and Munich hubs since 2015. Nicole has been working in the airline industry for more than 30 years and has been able to build up extensive knowledge during this time. Important stations of her career are controlling, passenger processes and operations. For more than 10 years, she has been responsible for crew workforce planning - previously at Lufthansa since the beginning of 2022 at Eurowings.
Franziska Burmester is co-founder of WePlan, a company based in Frankfurt. WePlan offers a web-based SAAS solution for Manpower Planning which amongst others also includes a module for creating cost-optimal training plans.
Before founding WePlan, Franziska has broad experience in consulting in the TT&L industry where she supported clients in improving workforce productivity and digitalize their planning process for more than 8 years. Franziska holds a bachelor degree in Economics and Tourism.
This work introduces a two-stage stochastic optimization approach for the strategic airline crew planning problem with uncertain demand. The developed model provides the airline with a cockpit crew composition plan per crew position before the flight schedule and crew demand are known. This is done by estimating future crew demand and by treating it as a stochastic variable. Historical crew demand is analyzed and is assumed to follow a Beta probability distribution. Since demand at different crew positions is correlated, demand scenarios are generated from these distributions by using Latin hypercube sampling (LHS) for correlated variables. Case studies are performed for a holiday airline with both scheduled and charter flights to validate the model and test its possibilities. Results show that the model provides a cost reduction of 2.1% concerning the airline’s current practice and a further 4.5% when affluent crew downsizing policies are considered.
Jody Snowdon: A mixed integer programming approach for optimal sequencing of pilot re-training
This work introduces a mixed integer programming model created to support Crew Planning understand how to sequence pilot training associated with the return to service of an aircraft fleet that was parked due to COVID-19 border closures. The model has two parts:
The solution of this model provides when the training of each pilot move should occur to effect the return to service of each aircraft as quickly as possible while adhering to training resource constraints and maintaining required numbers of pilots in all ranks.