Have a personal or library account? Click to login
Feed Rate-Adaptive Predictive PH Interpolation for Real-Time NURBS Toolpath Execution in High-Speed CNC Machining Cover

Feed Rate-Adaptive Predictive PH Interpolation for Real-Time NURBS Toolpath Execution in High-Speed CNC Machining

By: Guosheng TianORCID  
Open Access
|Oct 2025

Figures & Tables

Fig. 1.

NURBS curve related to the airfoil profile [75] NACA2415 (Axes show the toolpath coordinates in the XY plane: X-axis (mm) and Y-axis (mm).
NURBS curve related to the airfoil profile [75] NACA2415 (Axes show the toolpath coordinates in the XY plane: X-axis (mm) and Y-axis (mm).

Fig. 2.

After defining the sharp corners, the profile of the advancing speed on the airfoil creates the PH curve and the speed profile.
After defining the sharp corners, the profile of the advancing speed on the airfoil creates the PH curve and the speed profile.

Fig. 3.

Servo control system block diagram [16].
Servo control system block diagram [16].

Fig. 4.

Forward speed profile on an airfoil using a PH-based predictive algorithm.
Forward speed profile on an airfoil using a PH-based predictive algorithm.

Fig. 5.

Planar toolpath coordinates, X-axis and Y-axis (mm), of the airfoil profile at three orientations (0°, 45°, and −45°), used to evaluate the predictive PH-based interpolation algorithm under varying dynamic conditions.
Planar toolpath coordinates, X-axis and Y-axis (mm), of the airfoil profile at three orientations (0°, 45°, and −45°), used to evaluate the predictive PH-based interpolation algorithm under varying dynamic conditions.

Fig. 6.

Comparison of velocity profiles: (a) Forward velocity determination algorithm to limit the chord error. (b) Forward velocity determination algorithm based on curvature. (c) Prediction algorithm based on PH.
Comparison of velocity profiles: (a) Forward velocity determination algorithm to limit the chord error. (b) Forward velocity determination algorithm based on curvature. (c) Prediction algorithm based on PH.

Fig. 7.

Comparison of trajectory-tracking error: (a) Forward speed algorithm to limit string error; (b) Forward speed algorithm based on curvature; (c) PH-based forward-looking algorithm.
Comparison of trajectory-tracking error: (a) Forward speed algorithm to limit string error; (b) Forward speed algorithm based on curvature; (c) PH-based forward-looking algorithm.

Fig. 8.

Experiment hardware.
Experiment hardware.

Fig. 9.

Comparison of the path-following error when testing interpolation algorithms (a) Algorithm for determining the feed rate to limit the error and tri (b) Algorithm for determining the feed rate based on the curvature; (c) PH-based predictive algorithm.
Comparison of the path-following error when testing interpolation algorithms (a) Algorithm for determining the feed rate to limit the error and tri (b) Algorithm for determining the feed rate based on the curvature; (c) PH-based predictive algorithm.

Servo control system transformation function parameters [16]_

ParameterX axisY axis
a01.938 × 1091.904 × 109
a13.538 × 1073.496 × 107
a22.135 × 1052.120 × 105
a36.984 × 1026.948 × 102
a41.001.00
b01.938 × 1091.904 × 109
b13.476 × 1073.435 × 107
b21.471 × 1051.466 × 105

Performance comparison of interpolation algorithms_

Interpolation algorithmChord error-limited [47]Curvature-based feed rate [48]PH-based predictive interpolator
Error on each axis [µm]X axisMAX357.68194.7756.96
Root mean square75.6541.5320.30

Y axisMAX132.5522.6413.97
Root mean square26.328.413.54

Routing error [µm]MAX14.1147.12142.79
Root mean square35.1412.983.84
MIN16.7808.512.32

Time [s] 0.34310.45260.3440

Comparative analysis of the feed rate interpolation algorithms in literature_

StudyInterpolation methodMax path error [µm]RMS error [µm]Feed rate adaptivityPredictive controlIndustrial suitability
[47]Chord error-limited357.68 (X) / 132.55 (Y)75.65 / 26.32Limited
[48]Curvature-based194.77 (X) / 22.64 (Y)41.53 / 8.41Moderate
[78], [79]Jerk-limited cubic spline65 – 120 (AVG)22 – 30Moderate
[11], [14]Real-time PH interpolation∼60 – 100N/APartialHigh
This studyPH-based predictive57.85 (X) / 43.62 (Y)20.25 / 14.85High

Profile airfoil PH curve error comparison_

Error [µm]
Curve areaNo area divisionDivision by area

Base mean squareMaxBase mean squareMax

AB4.919313.430.50251.43
CD4.37929.780.54621.45

Hardware specifications used in the experiment_

ItemDescriptionApplication in industryReason for selection
Servo driveAC servo drive TECO TSTA-20CCommonly used in CNC machine tools for high-speed, closed-loop motor controlProvides stable torque-speed characteristics and is compatible with standard motion controllers
Servo motorsTECO TST0640 / TSB0845 AC servo motorsWidely implemented in industrial X–Y stages and CNC milling systemsHigh positioning accuracy, low latency response, and encoder feedback support
Motion controllerPCI-1240 motion control cardUsed in precision motion control applications including real-time CNC path trackingOffers deterministic motion planning and is compatible with bit-pattern interpolation
Digital linear scalesCARMAR high-resolution linear scales (1 µm resolution)Commonly used in coordinate measuring machines (CMMs) and precision CNC tablesProvides reliable high-resolution feedback for path-following error evaluation

Simulation and experimental input parameters for evaluating the feed rate-adaptive predictive PH interpolation algorithm_

SymbolDescriptionValueUnit
VmaxMaximum feed rate3500mm/min
AmaxMaximum acceleration2450mm/s2
JmaxMaximum jerk5 × 104mm/s3
δLateral error limit1.0µm
κ¯ {\bar \kappa } Reference curve1.0mm−1
εmaxPath following error limit15.0µm
eApproximate curve error limit1.0µm

Summary of key mathematical parameters and control variables used in the predictive interpolation algorithm, including their physical meaning and relevance to CNC machining operations_

SymbolDescriptionUnitRelevance in CNC operation
TsSampling periodsDefines time resolution for interpolator output
VkFeed rate at step kmm/sInstantaneous tool motion speed
AkAcceleration at step kmm/s2Dynamic response of tool motion
δChord error tolerancemmMaximum deviation from ideal path for profile generation
κ(u)Scalar curvaturemm−1Describes path sharpness, used for adaptive feed rate
kthCurvature thresholdmm−1Defines boundary to detect sharp corners
AmaxMaximum allowable accelerationmm/s2Mechanical limit of CNC axis acceleration
VmaxMaximum allowable feed ratemm/sCap of commanded feed velocity
ΔsArc-length step per interpolationmmTool travel distance per cycle
LLookahead arc windowmmUsed for predicting upcoming curvature transitions
ɛePath-following errormmActual deviation from commanded path
ϕTangent angle of toolpathradUsed for transforming Cartesian error into tangential form

Experimental machining accuracy using PH-based predictive interpolation_

MetricX axis [µm]Y axis [µm]Combined error [µm]
Maximum path-following error57.8543.6214.10
RMS path-following error20.2514.853.84
Minimum error observed2.312.312.31
Average deviation from reference trajectory18.2113.74
Machining tolerance achieved (from ruler data)±15±15Within allowed range

Airfoil curve numerical simulation at three angles_

Interpolation algorithm–45°45°
Error on each axis [µm]X axisMAX57.9743.5138.71
Root mean square20.3014.9314.19

Y axisMAX13.9838.8543.66
Root mean square3.5614.2414.99

Routing error [µm] MAX14.1114.1114.12
Root mean square35.143.843.84
MIN16.7802.312.31

Predictive interpolation parameter definitions_

ParameterDescriptionValue / rangeUnit
amaxMaximum allowable tangential acceleration2m/s2
VmaxMaximum allowable feed rate1m/s
ɛSmall positive constant to prevent division by zero in curvature-based calculations10−6
ΔsArc-length step per interpolation cycle0.05 – 0.2mm
LLookahead arc length used for predictive curvature filtering5 – 10mm
Language: English
Page range: 284 - 299
Submitted on: May 10, 2025
Accepted on: Sep 4, 2025
Published on: Oct 23, 2025
Published by: Slovak Academy of Sciences, Institute of Measurement Science
In partnership with: Paradigm Publishing Services
Publication frequency: Volume open

© 2025 Guosheng Tian, published by Slovak Academy of Sciences, Institute of Measurement Science
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.