<p>The performance and reliability of critical technologies are fundamentally constrained by microscopic quality characteristics (MQCs) forged during manufacturing. Identifying the pivotal few—the critical micro-quality characteristics (CMQCs)—is a long-standing challenge. This difficulty arises from the dual constraints of uncertain expert knowledge and the data scarcity common in small-batch production. This work introduces a hybrid fuzzy-Bayesian framework to resolve this challenge by decoupling a characteristic’s importance into its systemic “topological influence” and direct “intrinsic importance.” To systematically manage knowledge uncertainty, we propose an innovative fuzzy confidence interval consensus mechanism. This mechanism aggregates and quantifies imprecise judgments from multiple experts, translating them into robust connection weights within a complex network. Topological influence is then derived from this network via an improved LeaderRank algorithm. Concurrently, a Bayesian-optimized Extreme Gradient Boosting (BO-XGBoost) model distills intrinsic importance directly from sparse manufacturing data. Applied to the complex milling of a high-precision aero-engine blade with a dataset of 200 samples, the framework identified both acknowledged and previously obscured critical factors, achieving an identification accuracy of 92.50%—a significant improvement over conventional approaches. This research provides a robust solution for quality control under profound uncertainty. It demonstrates how fuzzy systems can effectively bridge the gap between qualitative human expertise and quantitative data-driven analysis in high-value manufacturing.</p>