Surface Roughness and Burr Height in Helical Milling of Ti-6Al-4V: Factorial Analysis and Grouped Machine-Learning Validation

Helical milling can produce Ti-6Al-4V holes, but the combined effects of cutting conditions on wall surface roughness ( $R_a$ ), within-hole variability, and burr height remain unclear. A replicated 2⁴ full-factorial experiment examined tangential feed, axial helical pitch, fluid supply, and axial ultrasonic vibration across 48 holes. Roughness was quantified from 1536 readings in four axial zones; burr height was obtained from 12 circumferential measurements at the entrance of each hole. Factorial and mixed-effects analyses were supplemented by conservative paired-block inference for the fluid main effect and grouped validation of five regression models against a mean baseline. Within the fixed fluid–tool run structure, flood machining was associated with 12.94% lower hole-mean $R_a$ (2.355 to 2.050 µm; p = 0.0285) and 25.10% lower mean within-zone $R_a$ SD. Both associations had nominal paired-block support but did not survive FDR correction across six responses; fluid was confounded with tool identity and run position. Axial variability reductions were descriptive and unsupported by the conservative test, and no systematic top-to-bottom roughness gradient was detected. Ultrasonic vibration was unrelated to wall roughness but was associated with 26.3% higher burr height, whereas higher tangential feed was associated with 44.6% lower burr height. Burr height was the most predictable individual-hole response; hole-mean $R_a$ showed moderate predictability, and axial $R_a$ CV transferred poorly across replicates. Wide and incompletely calibrated Gaussian process intervals preclude individual-hole tolerance certification. The findings support multi-response assessment and response-specific validation of hole quality.