Digital Twin for Aircraft Wing Maintenance with ML-Based Deviation Detection

Abstract

A growing need in the aviation industry is the development of smart maintenance systems that have the ability to not only improve aircraft safety but also reduce operating costs and time. Existing inspection technologies are mostly time-based and do not allow one to find structural defects between maintenance intervals. Digital Twin (DT) technology allows obtaining a virtual copy of the asset in real-time through integration of sensors, computational models, cloud communication, and predictive analysis. In this work, we propose a DT-based framework for aircraft wing maintenance with ML-based deviation detection. Our framework includes a physical aircraft wing model with a Flex sensor and MPU6050 accelerometer connected to an ESP32 microcontroller for real-time data collection. Data from the sensors are transmitted to the ThingSpeak cloud service and synchronised with a virtual twin of the physical wing created using MATLAB/Simulink and Simscape. Decision Tree classifier uses strain and vibration characteristics to classify the state of the wing as normal or abnormal to perform predictive maintenance.