3 Presentations

• 
  10:30 AM - 10:55 AM

  Long-Term Planning of Preventive Maintenance using Constraint Programming: A Naval Case Study

  • Raphaël Boudreault, presenter, Thales Digital Solutions
  • Vanessa Simard, NQB.ai

  Maintenance planning is an essential element in the life-cycle management of an asset. Unplanned maintenance work can cause significant productivity and financial loss, while manually assessing compliance is complex and prone to errors. In the naval domain, ensuring mission readiness and operational availability is critical. Thus, periodic preventive maintenance tasks must be carefully allocated over a long-term horizon considering the ship availability, business rules, and workload limitations. This distribution over fixed short work periods can result in tasks being excessively advanced or deferred instead of executed when due. We propose a Constraint Programming approach to produce feasible tactical plans of preventive maintenance for ships minimizing advancements and deferrals of tasks. We validate our methodology on an industrial naval use case and demonstrate its relevance compared to a currently used manual method, greatly reducing over-maintenance and creating safer plans. The method is integrated into Maintenance Optimizer (TM), an interactive planning solution that supports decision-making.

• 
  10:55 AM - 11:20 AM

  A Mixed-Integer Programming Approach for an Extended Fixed Route Hybrid electric Aircraft Charging Problem

  • Anthony Deschênes, presenter, Student
  • Raphaël Boudreault, Thales Digital Solutions
  • Claude-Guy Quimper, Université Laval
  • Jonathan Gaudreault, Université Laval

  Air mobility is rapidly moving towards the development and usage of hybrid electric aircraft in multi-flight missions. Aircraft operators must consider numerous infrastructure and operational constraints in their planning, during which predicting energy usage is critical. In our past work, we introduced this problem as the Fixed Route Hybrid Electric Aircraft Charging Problem (FRHACP) and proposed a dynamic programming approach. Given a fixed route, this problem aims to decide how much to refuel/charge at each terminal as well as the energy types to use during each flight leg. We introduce a Mixed-integer programming (MIP) model to solve an extended version of the Fixed Route electric Hybrid Aircraft Charging Problem. This model is then compared with the dynamic programming model on a benchmark of 10 realistic instances. Results show that the MIP model obtains better quality solutions than the dynamic programming model on all instances. The dynamic programming model, however, computes its solution faster. Finally, an experiment where the size of the instances is increased is conducted to verify the scalability of the models. Results show that the dynamic programming model has the best scalability.

• 
  11:20 AM - 11:45 AM

  Developing hybrid learning models to predict the duration of maintenance tasks

  • Yun Yin, Dalhousie University
  • Jiye Li, Thales Research and Technology Canada
  • Alireza Ghasemi, presenter, Dalhousie University
  • Claver Diallo, Dalhousie University

  Accurate estimation of maintenance task duration helps manage resource usage and allows planners to decide on maintenance priorities within a limited time frame. Better estimated task durations help produce more robust resource schedules, perform more tasks in maintenance facilities, reduce resource idling time and increase operational availability. However, task duration estimations have until now been typically performed by human experts with little artificial intelligence-based forecasting for maintenance operations. The analysis of historical data is also not a common practice to complement any expert-driven forecasting. To explore opportunities for using AI in this work domain, and to improve human estimations for task duration, we propose a novel hybrid learning approach that integrates human forecasts with data-driven models. Our empirical data sourced from a Canadian Aerospace & Defence company comprises maintenance records for 12 aircraft, encompassing over 58,000 anonymized work orders spanning from 2010 to 2018. Supervised learning algorithms and bidirectional LSTM models are used to forecast the preventive maintenance task durations with and without expert estimates. Several data engineering and pre-processing techniques such as PCA, MCA, and feature importance are explored. The results demonstrate that hybrid learning models perform better than both human expert model and historical data alone.