Optimization of vibration behavior of carbon fibre reinforced polylactic acid based drone quadcopter arms using ANOVA and support vector regression algorithm

The dynamic behavior of quadcopter arms is notably sensitive to variations in thickness, internal infill architecture, and perforation geometry, particularly in lightweight quadcopter systems. This study aimed to evaluate the vibration characteristics of additively manufactured perforated quadcopter arms using a combination of experimental and numerical methods and machine learning techniques. The specimens were fabricated from carbon-fiber-reinforced polylactic acid (PLA-CF) using fused deposition modelling (FDM) technique employing a gyroid infill structure. The configurations for vibration testing were determined using a Taguchi’s L27 orthogonal array, which utilized three parameters at three levels: arm thickness (8 mm, 10 mm, and 12 mm), horizontal spacing between perforations (3 mm, 6.5 mm, and 10 mm), and infill density (20%, 35%, and 50%). The first six natural frequency modes were extracted using ANSYS finite element package and analysed to determine the effects of the three selected design parameters on the vibrational performance of the quadcopter arm. Dynamic stress was analysed using a sequential modal-harmonic-transient analysis procedure, and the resulting maximum principal stresses were assessed against Rankine's maximum principal stress criterion, which confirmed a minimum factor of safety of 2.14 across all 27 configurations. Additionally, a radial basis function support vector regression (SVR) surrogate model, tuned by exhaustive grid search under leave-one-out cross-validation, was employed to derive predictive relationships between the design parameters and the response variables (leave-one-out R 2 = 0.943 for frequency and 0.924 for dynamic stress). Coupling the surrogate with a differential evolution optimiser subject to a resonance-avoidance constraint identified an optimal configuration of 12.00 mm thickness, 10.00 mm horizontal spacing and 20% infill density, predicting a fundamental frequency of 923.27 Hz and a dynamic stress of 9.137 MPa. This optimum coincides with an experimentally fabricated and tested specimen (DA25: 922.77 Hz, 9.1475 MPa), providing direct physical validation of the surrogate-based optimisation framework, with a prediction error of 0.05% in frequency.