Model-Based Indirect Load Measurement for Aircraft Landing Gear Using FBG Sensors and Dynamic Graph Networks

Accurate measurement and quantitative evaluation of aircraft landing gear loads are essential for ensuring structural integrity and flight safety. Since no dedicated sensor can directly measure landing gear loads, the assessment must rely on indirect estimation from strain responses obtained bystrain sensors. However, most conventional and learning-based approaches fail to model the spatial topology of measurement points or to account for variations caused by different buffer-stroke conditions, resulting in calibration inconsistency and reduced measurement precision. To address these limitations, we propose a model-based indirect load-measurement method that employs a dynamic graph network (DGN) to estimate loads from distributed strain data. By integratingchannel-wise self-attention (CSA) with dynamic graph con volution (DGCN), the model captures both local and global dependencies in the strain-load relationship, thereby enhancing robustness across varying buffer strokes. Ground load calibration tests were performed on a full-scale aircraft right landing gear, where fiber Bragg grating (FBG) sensors measured strain responses associated with longitudinal (X), vertical (Y), and lateral (Z) load components under three representative buffer-stroke settings. Using K-fold cross-validation (K = 14), the proposed approach achieved high-accuracy load estimation with mean absolute percentage errors (MAPEs) of (4.125 ± 1.743)%, (1.347 ± 0.421)%, and (4.196 ± 1.419)% in the X-, Y-, and Z-directions, respectively. Furthermore, the overall measurementuncertainty of our built indirect load-measurement system was experimentally determined to be 2.5%, which indicates a compact overall uncertainty level for the proposed indirect measurement system and reflects the stable performance of the method. Source code is available at https://github.com/Mu-Tang/CSA-DGCN.