Machine Learning-Driven Finite Element Analysis for Structural Assessment of Agricultural Drone Arms
Year: 2026
Type: conference-paper
Abstract: Agricultural hexacopters operate under highly stochastic conditions where uncertainties in payload and flight maneuvers significantly impact structural reliability. This study presents a multi-fidelity uncertainty quantification (UQ) framework to evaluate the structural integrity of the SB1 RP08 drone arm, constructed from Carbon Fiber Reinforced Polymer (CFRP). The methodology integrates Latin Hypercube Sampling (LHS) for efficient design space exploration, high-fidelity Finite Element Analysis (FEA) for structural response training, and Polynomial Regression surrogate modeling to reduce computational costs. A Monte Carlo Simulation (MCS) with$\mathbf{1 0}^{\mathbf{7}}$iterations was conducted to propagate input uncertainties specifically total mass and pitch angle into the probabilistic stress response. The results reveal distinct structural behaviors in the Hover case exhibited a higher mean maximum von Mises stress of 589.99 MPa with a narrow probability distribution, whereas the Forward Flight case yielded a lower mean stress of 436.99 MPa but displayed a significantly wider distribution. This finding indicates that while forward flight incurs lower average loads due to thrust redistribution, the introduction of pitch angle uncertainty creates greater variability, making it a critical scenario for reliability assessment. The proposed framework demonstrates a robust and computationally efficient approach for reliability based design optimization in unmanned aerial vehicles.
Institution: Suranaree University of Technology
Language: English
Subfield: Computational Mechanics
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7202391982