Bio-Inspired Sensitivity-Weighted NSGA-II Optimization of a 6-UPS Parallel Loading Mechanism for Aero-Engine Pylon Vector-Force Loading

Authors:
Open Access
Publisher website
Google Scholar
Abstract
Structural static testing is paramount for validating the structural integrity of critical aerospace components. However, conventional test rigs are often constrained to fixed loading axes and frequently induce parasitic torques. Accurate reproduction of aero-engine pylon flight loads therefore requires a mechanism that combines omnidirectional vector loading, high stiffness, and efficient force transmission. Achieving these coupled requirements is primarily a geometric synthesis problem, yet the associated workspace, stiffness, and load–capacity indices are nonlinear, mutually coupled, and expensive to evaluate over dense pose samples. To address this optimization bottleneck, this work develops a task-specific 6-UPS loading mechanism and a bio-inspired sensitivity-weighted NSGA-II algorithm for its geometric synthesis. Inspired by gene/locus-specific heterogeneity in biological evolution, the algorithm assigns variable-wise search intensities according to design-variable sensitivities, which are estimated using Multivariate Adaptive Regression Splines (MARS). In this way, influential design genes receive stronger local exploitation, whereas less sensitive ones retain broader exploration. Numerical simulations demonstrate that the proposed approach reduces computation time from about 30 h to 3 h relative to direct optimization with the baseline NSGA-II, while simultaneously improving workspace, stiffness, and load-carrying capacity. A hybrid physical prototype was further tested under 240 loaded pose conditions; the system maintained force magnitude errors below 0.64% (63.42 N) and directional deviations below 1.15°. These results support the efficacy of the proposed bio-inspired optimization-based design methodology for high-fidelity static testing of aero-engine pylons under the adopted hybrid setup.
Funding Information
  • National Natural Science Foundation of China (U22A20204)

This publication has 37 references indexed in Scilit: