Predictive maintenance in aviation requires high-quality, well-processed data to build accurate machine learning models. The reliability of these models depends heavily on how data is collected, cleaned, and prepared for analysis. However, aviation data is often complex, heterogeneous, and incomplete. This thesis focuses on designing a systematic framework for collecting, organizing, and preprocessing data to support predictive maintenance applications in aviation.
Objectives:
The goal of this thesis is to gather, preprocess, and analyze aviation maintenance data to identify patterns, correlations, and trends relevant to predictive maintenance. The study will create a clean, well-structured dataset that can serve as the foundation for predictive modeling in subsequent research (Master’s Thesis).
Research Questions:
- What types of data are available and relevant for predictive maintenance in aviation?
- What preprocessing steps are required to transform raw data into a suitable format for machine learning models?
- What initial patterns or correlations can be identified in the data that are potentially predictive of component failure?
Tasks:
- Data Collection: Simulate or access real-world data from publicly available repositories like NASA's C-MAPSS dataset.
- Data Preprocessing: Implement preprocessing steps, including normalization, feature scaling, outlier detection. Conduct exploratory data analysis (EDA) to uncover trends and relationships in the dataset.
- Validation: Test the framework by preparing data for a ML-based predictive maintenance model and evaluating its impact on model performance.
This thesis will create a foundation for predictive maintenance by ensuring data quality, reducing errors in ML models, and supporting more accurate failure predictions in aviation.
Time duration: 3 Months
Contact:
Dr. Thomas Feuerle, t.feuerle@tu-braunschweig.de
Parth Purohit, parth-yogeshbhai.purohit@tu-braunschweig.de