In-service Load Monitoring for an UAV Digital Twin

Abstract

A load monitoring system plays an important role in recovering the actual load spectra of aeronautical structures, contributing to the online evolution of airframe digital twins. In scenarios where many aircrafts lack on-board strain sensors during the service phase, yet some strain data is available during the test flight phase, our innovative approach utilises deep learning-based flight-strain prediction and an inverse-direct approach for in-service load monitoring. Initially, a deep learning approach is employed during the test flight prior to service to establish a flight parameter-strain prediction model. This model, incorporating time series features of flight and strain data, exhibits superior predictive accuracy compared to traditional regression methods. Moving into the subsequent service phase, the flight parameter-strain prediction model seamlessly integrates with an inverse-direct load monitoring method. This integrated approach facilitates real-time monitoring of full-field load distribution, relying solely on flight parameters. Validation of the approach utilises flight test data from an unmanned aerial vehicle, revealing better performance compared with the strain-measurement-based method. Notably, our method's efficacy extends across diverse aircraft types, as it does not rely on on-board strain sensors during the service phase.