Aerodynamically Constrained Machining Optimization of Compressor Blades Leading Edge Using a Non-Probabilistic Bounded Field Method

The aerodynamic performance and operational stability of aero-engine axial compressors are governed primarily by the machining accuracy of rotor blade leading edges (LE). Modern highly loaded, thin-walled compressor blades exhibit an exponential aerodynamic sensitivity to micron-scale LE manufacturing deviations, posing a critical challenge for high-precision small-batch production of aerospace blades. Traditional probabilistic uncertainty quantification (PUQ) methods rely heavily on large-sample Gaussian assumptions, which leads to statistical divergence and underestimation of extreme risks under small-sample constraints. Conventional non-probabilistic convex models suffer from inherent over-conservatism, and most existing approaches fail to establish a physically interpretable mapping between manufacturing defects and aerodynamic performance degradation. Focusing on steady-state near-design operating conditions, to address these limitations, this study proposes a novel Non-Probabilistic Bounded Field (NPBF) model for LE deviation dispersion quantification and aerodynamic penalty evaluation. The method integrates physics-informed modal decomposition, a hybrid MVEE-KDE uncertainty domain to balance boundary reliability and conservatism, and a Gaussian process regression-based Bayesian optimization framework for CNC segmented milling parameters. Within the framework of the proposed NPBF method, integrated with the Bayesian optimization algorithm, the co-evolution metric is improved by 21.3% and the computational efficiency is improved by 43.3% compared with the standard genetic algorithm, achieving simultaneous enhancement of solution accuracy and computational efficiency. Test results of titanium alloy blisk sectors made of TC11 and TC17 with different material specifications and dimensions demonstrate that the intra-blade geometric dispersion of the leading edge is reduced by up to 9.5%, while the inter-blade profile deviation is decreased by up to 4.5%. Meanwhile, this geometric homogenization effect suppresses premature boundary layer separation and effectively reduces the aerodynamic loss coefficient, establishing an initial baseline framework for aerodynamically constrained machining accuracy optimization for aerospace manufacturing under small-sample constraints.