<p>In five-axis machining of parts with variable curvature, excessive motion axis speeds beyond servo limits can destabilize the tool feed rate, inducing chatter and degrading surface quality and contour accuracy. To address this, a jerk-constrained speed partitioning algorithm is proposed to stabilize the cutter contact feed rate. First, a model for the speed, acceleration, and jerk of each axis were established based on the linear interpolation principle, and the large curvature inflection points and corner area were identified using the rate of change of the vector angle between tool positions. Second, for high-curvature corner areas, a speed partition optimization algorithm under jerk constraints was employed to dynamically optimize speeds that exceed servo limits, ensuring they remain within the constrained range. Finally, the effectiveness of the algorithm was validated through machining experiments on a split-flow impeller. Results show that the optimized total feed rate effectively improves the surface quality of the parts, with computational efficiency and contour accuracy increasing by approximately 20.28% and 43%, respectively. Additionally, the speed, acceleration, and jerk curves of each axis are within the limit range, which proves the feasibility and effectiveness of the algorithm.</p>