Patsy Jammal
(Advisor: Prof. Dimitri Mavris]

will propose a doctoral thesis entitled,

Digital Twin-Driven Condition Monitoring Approach for Aircraft Carbon Brakes

On

Thursday, August 31st at 09:00 a.m.
Virtually
Click here to join the meeting

Abstract
Wheels and brakes are among the largest contributors to aircraft component maintenance costs. Carbon brakes have a high initial cost, and their current time-based maintenance practice often leads to unnecessary inspections and increased aircraft downtime. Condition-based maintenance strategies, such as those enabled by prognostics and health management studies of operational data, have the potential to help reduce maintenance costs, increase aircraft availability and safety, and enhance carbon brake life by providing tailored insights to reduce wear.

Carbon brake maintenance is a significant revenue stream for manufacturers, which also grants them access to operational data from their customers. The use of big data analytics on such data enables innovative solutions to optimize brake maintenance. Traditionally, condition monitoring techniques lacked real-time predictive capabilities.  With the advent of Digital Twin (DT) technology and the abundance of multidomain data availability, a new approach to condition monitoring using data-driven methods such as Artificial Intelligence (AI) and Machine Learning (ML) has emerged.

The existing literature on carbon brake wear highlights a number of gaps, including the lack of understanding the effects that varying operational and environmental conditions have on brake wear, the limited use of advanced ML techniques with uncertainty quantification for high-dimensional datasets, and the lack of generalizability assessments of predictive models across different domains (e.g., varying aircraft types, route structures, etc.). Therefore, the objective of this research focuses on developing an optimized and generalizable data-driven methodology that predicts carbon brake life remaining based on various parameters, including aircraft-specific parameters, operational conditions, and environmental factors, while quantifying the uncertainty in the predictions.

This research addresses the aforementioned gaps through the development of a rigorous and repeatable methodology that can be applied to various other components that experience wear. First, clustering techniques are used to identify varying ranges of aircraft, operational, and environmental parameters corresponding to different groupings of brake wear severity. Doing so allows for the identification of distinct duty cycles used to test brake performance in a lab environment (i.e., on a dynamometer). Next, supervised ML algorithms are used to develop classification models that determine the severity of carbon brake wear based on varying operational and environmental conditions. This allows flights experiencing excessive brake wear to be identified along with the influencing factors.

The problem is then tackled as a regression problem for more accurate and precise brake wear predictions using both traditional ML and advanced Deep Learning (DL) techniques, allowing for a benchmark of multiple algorithms and an assessment of their suitability for DT modeling purposes. Both regression and classification techniques rely on the availability of the "Wear Pin Value" parameter from aircraft flight data, which indicates the percentage of carbon brake disk remaining and is reported inconsistently every 10 flights or so. As such, the optimal frequency at which the wear pin value should be collected is investigated to determine how often operators would need to report the carbon disk thickness remaining on aircraft that do not have electronic wear pin sensors.

Furthermore, the generalizability of the predictive brake wear models is assessed to determine whether specific models need to be developed for different segments of the data (e.g., diverse aircraft types or route structures), or if a single model can be used with acceptable predictive accuracy across the entire dataset. Lastly, transfer learning techniques are incorporated to explore their potential to provide increased model performance across different data segments.

Committee

• Prof. Dimitri Mavris – School of Aerospace Engineering (Advisor)
• Dr. Olivia Pinon-Fischer – School of Aerospace Engineering
• Assoc. Prof. Brian German – School of Aerospace Engineering
• Prof. Daniel Schrage – School of Aerospace Engineering
• Dr. Gregory Wagner – Senior Manager, Collins Aerospace