Improving Plane Maintenance Process with AnyLogic Agent-Based Modeling

Problem:

We all take commercial air flights from time to time. However, we do not know how complex plane maintenance can be. The military aircraft maintenance turnaround process (the in-between time when the aircraft touches down, is refueled, rearmed, and inspected, in order to be released) is even more complex and, being fairly time consuming, includes multiple interactions and parallel workflows. In addition, skilled staff are needed to maintain the sustainable level of the turnaround process, which leads to associated costs.

Engineers from Lockheed Martin, one of the largest companies in the aerospace, defense, security, and technologies industry, used AnyLogic simulation modeling and tried to improve decision making in the entire military airplane turnaround process and evaluate the impact of process changes on turnaround time.

Solution:

To complete a model, the three main elements of the turnaround time process had to be considered. The elements included the aircraft inspections on the flight line, the signoff (meaning that all of the inspections and refueling have been completed), and the review and disposition of any maintenance codes that were downloaded from the aircraft.

Once these processes were clarified, a mobile application was designed to enable recording, validation, and understanding of the process at each stage of maintenance. This data collection tool was used by the observers, who monitored the maintenance staff. The application was modified several times over the course of the project.

Business Process Simulation Model Application

The main requirements for the data collection tool are listed below.

For each step in the workflow, actors, resources, dependencies, and other process definition data were identified. The data needed for the model included the start and stop times of each task. In addition to the start and stop times, it was important to provide an audio recording capability to capture activities that were not mentioned in the application. For example, observers might record the reason that a task was taking longer than expected, or record that they had accidentally pushed the wrong start button. Such information was valuable when trying to interpret the data.

Through the interface, the user could add tasks, if new tasks were identified during the data collection period. The observers could also record comments against specific tasks. All of these made the data collection application highly flexible and adaptable.

Studying the aircraft turnaround process revealed that agent-based modeling and simulation environment should include experimentation and presentation capabilities. The AnyLogic simulation modeling tool fulfilled these requirements. Additionally, process visualization in the model contributed in its presentation to all levels of developers and senior leadership.

Agent-based Simulation Model

In the next stage, agents, resources, and tasks that were identified during the process modeling step, were implemented into the process flow in AnyLogic, along with multiple visualizations. After that, baseline models of the “as-implemented” processes were built. They were iteratively run in a deterministic mode for debugging purposes, as well as single and multi-run Monte Carlo modes. The outcomes were compared to what was experienced on locations.

After validating and updating, a stochastic agent-based model was able to capture the dynamic and interacting processes that comprised the turnaround process. To make the process more efficient, the experiments were performed to quantify the impact of process changes, whether through deletion of process steps, a reduction in the amount of time needed to execute a process step, or redefinition of process’ portions.

The experiments with the model helped:

Outcome:

Various experiments with the model, including Monte Carlo method, resulted in suggestions, showing which modifications to the process would make the most difference, and the potential range of that difference. Not only the modeling and simulation approaches of AnyLogic helped model people/machines/workstations interaction, but they also tracked workflows’ peculiarities, which were unknown before the experiment.

All of this let engineers propose modifications to the workflow and show quite sizable improvements in terms of the airplane turnaround process. Using AnyLogic process modeling allowed them to capture the bottlenecks of the turnaround process, and also allowed tasks under non-deterministic data to be added, removed, or reordered.

Download full case study (PDF)

Project presentation by Nadya Belov, Senior Researcher at Lockheed Martin

Similar case studies

More case studies