Predicting CNC Machining Errors Before the First Cut: A PO-ELM Intelligence Framework for Toolpath-Aware Quality Control in Smart Manufacturing

Theophilus Ikiedideike¹, Stanislaus Chinemerem Philip²

¹ PhD Student, Department of Mechanical Engineering (Production Option), Rivers State University, Port Harcourt, Nigeria.

² Student, Department of Mechanical Engineering (Production Option), Rivers State University, Port Harcourt, Nigeria. 

Abstract

Geometric inaccuracies in CNC milling remain a persistent barrier to zero-defect smart manufacturing. Existing approaches treat toolpath selection, machine performance monitoring, and error prediction as separate concerns, leaving a critical integration gap. This paper closes that gap by presenting a unified intelligence framework that couples toolpath strategy evaluation with a Partially-Optimized Extreme Learning Machine (PO-ELM) error predictor. Eight toolpath strategies-Zig-Zag, Contour, Spiral, Radial, Zig, One-Way, Morph, and Follow Periphery-were benchmarked on aluminium 6061 components machined at Innoson Manufacturing Company, with Coordinate Measuring Machine (CMM) inspection providing ground-truth geometric error data. The Morph strategy achieved the shortest machining time (22.4 min) and lowest power consumption (2.6 kW), while Follow Periphery delivered the best surface finish (Ra = 0.59 μm) and tightest dimensional deviation (0.011 mm). Vibration levels above 2.0 m/s² were strongly correlated with positioning errors exceeding 8 μm and thermal growth above 12 μm, providing actionable in-process monitoring thresholds. The PO-ELM model, trained on multi-source machining data, predicted surface roughness with R² = 0.916 and geometric errors with residuals contained within ±0.002 mm across all samples. Feature importance analysis identified toolpath length (0.85), spindle speed (0.78), and feed rate (0.72) as the dominant predictors. The resulting Pareto optimization frontier quantifies the time–accuracy trade-off for each strategy, enabling data-driven process selection. The integrated framework demonstrated a 28.2% reduction in machining time and a 39.2% improvement in surface accuracy versus the worst-performing strategy, providing a replicable methodology for proactive quality control in smart CNC environments.    

Keywords: CNC machining, Extreme learning machine, Geometric error prediction, PO-ELM, Smart manufacturing, Surface roughness, Toolpath optimization.

References

1. Abeykoon, C. (2020) ‘Advances in modelling and monitoring of machining processes’,” J. Manuf. Mech. Internal Rev., 7(24), pp. 1–14.
2. Adebayo, A. (2023) ‘Neural network modeling of surface finish in precision milling’, J. Manuf. Syst., 6(6), pp. 112–121.
3. Adekunle, A. (2021) ‘A lookup table-based error compensation system for positional accuracy in CNC milling’,” J. Manuf. Process., 6(4), pp. 1358–1366.
4. Altintas, Y. (2012) Manufacturing automation: Metal cutting mechanics, machine tool vibrations, and CNC design. 2^nd ed. Cambridge, U.K.: Cambridge Univ. Press.
5. Chen, Y. (2021) ‘Support vector machine for tool wear prediction in titanium milling,’ J. Manuf. Process., 5(3), pp. 78–92.
6. Gologlu, C. and Sakarya, N. (2008) ‘The effects of cutter path strategies on surface roughness of pocket milling of 1.2738 steel based on Taguchi method,’ J. Mater. Process. Technol., 206(13), pp. 7–15.
7. Gopalakrishnan, O. (2022) ‘Neural network-based modeling of surface integrity in end milling process,’ J. Manuf. Process., 7(5), pp. 10–18.
8. Huang, G.B., Zhu, Q.Y. and Siew, C.K. (2006) ‘Extreme learning machine: Theory and applications,’ Neurocomputing, 70(3), pp. 489–501.
9. Igbokwe, N.C., Nwamekwe, C.O. and Okpala, C.C. (2026) ‘Manufacturing waste reduction through data-driven process optimization: Evidence from smart production systems’, International Journal of Technology, Health and Sustainability, 2(1), pp. 165-174. https://ijths.com/wp-content/uploads/IJTHS-020167.pdf
10. Ikiedideike, T. and Philip, S.C. (2026) ‘Economic justification for predictive maintenance in CNC machining: An industrial case study on cost and downtime reduction’, International Journal of Technology, Health and Sustainability, 2(1), pp. 71-75. https://ijths.com/wp-content/uploads/IJTHS-020149.pdf
11. Kamara, S. (2021) ‘Thermal-based adaptive control for dimensional accuracy in precision machining,’ Int. J. Precis. Eng., 7(2), pp. 764–775.
12. Khan, M. (2022) ‘A mechanistic model for prediction of tool deflection and its effect on surface error in end milling,’ J. Mater. Process. Technol., 2(9), pp. 11–48.
13. Lawal, I. (2022) ‘Power monitoring-based adaptive control for energy-efficient CNC turning,’ J. Cleaner Prod., 3(36), pp. 13–21.
14. Luo, J. (2022) ‘A modified extreme learning machine with partial optimization for industrial data modelling,’ IEEE Trans. Ind. Informat., 18(5), pp. 3421–3430.
15. Nwosu, O. (2021) ‘Performance analysis of partially-optimized extreme learning machine in industrial prediction applications,’ J. Manuf. Syst., 6(1), pp. 345–357.
16. Okon, P. (2022) ‘Parameter optimization strategies in partially-optimized extreme learning machines,’ Engineering Applications of Artificial. Intelligence, 1(12), pp. 10–34.
17. Olawale, K. (2020) ‘Optimization of machining parameters for improved surface roughness in CNC turning of aluminum 6061,’ Nigerian J. Technol., 39(2), pp. 455–462.
18. Singh, A. (2022) ‘Extreme learning machine for kerf width prediction in laser cutting,’ Int. J. Adv. Manuf. Technol., 3(7), pp. 45–59.
19. Wang, K. (2023) ‘Application of PO-ELM to predict power consumption of CNC machine tools,’ J. Manuf. Syst., 7(4), pp. 210–225.