Transforming Aviation Maintenance

The "FIVA" Project

Introduction

In the demanding field of aviation maintenance, efficiency and accuracy in fault diagnosis and resolution are critical. This case study examines the development and implementation of the Fault Isolation and Virtual Assistant (FIVA), a pioneering AI-driven solution designed to revolutionise maintenance processes and reduce aircraft downtime.

Situation

At Mesmerise, I led a team of ten professionals to tackle a significant challenge: enhancing the prioritisation and efficiency of aviation maintenance tasks. The complexity and urgency of these tasks necessitated a sophisticated system capable of streamlining workflows and minimising downtime, which led to the conceptualisation of FIVA.

Planning

My role involved comprehensive planning and coordination from project inception. We started by gathering a vast array of historical maintenance data and engaging with maintenance engineers to deeply understand their pain points. This phase included traveling to Istanbul to work closely with and shadow maintenance engineers, allowing us to observe their workflows firsthand and gain invaluable insights into the challenges they faced daily. This direct engagement helped us develop a detailed project roadmap, outlining key milestones, risk assessments, and resource allocation.

Execution

The FIVA project was marked by a collaborative approach involving data scientists, software engineers, and maintenance experts:

Data Collection: Extensive historical maintenance data, including time logs and engineer notes, were analysed.
Feature Engineering: We identified key factors affecting task complexity and developed a machine learning model to predict complexity scores for maintenance tasks.
Model Training: Decision trees and random forests were chosen for their effectiveness in handling the categorical and textual data found in maintenance logs.
Integration: The AI model was integrated into the FIVA system, providing a user-friendly interface that allowed engineers to prioritise tasks efficiently using AI-driven insights.

Challenges and Solutions

Technical Adaptation: The project required advanced technical expertise and iterative testing to develop a reliable machine learning model.
User Adoption: We conducted multiple user testing sessions to refine the user interface, ensuring the system was intuitive and met engineers' needs.
Data Quality: Rigorous data cleaning protocols and validation checks ensured the integrity of the data used for training the model.

Results

Initial testing of FIVA indicated improvements in operational efficiency and usability. These early results suggested a reduction in the time required for engineers to prioritise maintenance tasks and familiarise themselves with new procedures.

Impact

FIVA demonstrated the effective application of AI and machine learning in high-stakes environments to improve outcomes and safety. The system's scalable architecture allows for easy updates and the integration of new functionalities to adapt to evolving maintenance needs.

Conclusion

FIVA exemplifies the transformative potential of AI and machine learning in traditional industries by enhancing efficiency, accuracy, and knowledge sharing. The project has laid the groundwork for future advancements in predictive maintenance and AI-driven diagnostics, promising to lead further innovations in the field of aviation maintenance.