Rotor Component Load Reconstruction For Fiber Bragg-Instrumented Rotor Blades

Abstract

Monitoring the loads on key rotor components during operation can help to reduce maintenance expenses and life cycle costs for helicopter operators. As such loads are typically challenging to measure, this study proposes a methodology for predicting loads on such components - including the pitch-link, the inter-blade damper and the rotor mast bend- ing moment - based on the data obtained from a fiber Bragg-instrumented pre-series AW09 rotor blade. A full-scale whirl-tower rotor test campaign was conducted to investigate the relationship between instrumented rotor components and strain signals. Time-domain signals were converted to frequency-domain features, which were used to train a su- pervised regression model predicting the Fourier coefficients of the target load sensor based on the instrumented blade data. A linear regression analysis was conducted to assess the influence of specific blade instrumentation characteris- tics, including local deformation and sensor radial position, on model performance. Additionally, the performance of a neural network model was compared to that of the linear regression. It was demonstrated that an instrumented blade can be an effective means of estimating the loads on other rotor components. The accuracy of this approach was found to be significantly superior to that obtained by using only the control input data for the regression. The regression exhibited an improvement in accuracy as the number of sensors increased. While neural networks showed potential to capture complex, non-linear relationships in the data, a simpler linear regression model was sufficient to reconstruct the main characteristics of the target signal.