Automated Detection and CNC Trajectory Generation for Pitting Repair in Aluminum Cylinder Heads Using YOLOv5

Flores J.; Magallanes J.; Figeroa J. · 2026 12th International Conference on Mechatronics and Robotics Engineering, ICMRE 2026 (2026) · DOI: 10.1109/icmre69538.2026.11533904

Resumen

The maintenance of aluminum cylinder heads in internal combustion engines still relies on manual visual inspection and artisanal repair planning, leading to inconsistent results, long intervention times, and a high rate of rework. Although industry has made progress in inspection and machining systems, there are currently no integrated solutions capable of automatically detecting defects and generating CNC toolpaths for their repair, representing a significant technological gap. To address this limitation, this study proposes an automated methodology that combines image processing, pit detection using YOLOv5, and a geometric conversion module that transforms each detected defect into a CNC trajectory adapted to its morphology. This workflow enables, for the first time, an automated closing of the inspection-repair loop. Experimental results show 94.7% accuracy in defect detection, processing times below 10 s per image, and a mean error of 0.28 mm in trajectory generation, validated on 300 real images. The proposed solution demonstrates the feasibility of integrating automatic detection with assisted repair, reducing operator dependence and process variability, and providing a significant advancement toward automation in automotive remanufacturing. This integrated approach demonstrates a viable solution for closing the inspection-repair loop in automotive remanufacturing, reducing human dependency and process variability.

Instituciones: Universidad Peruana de Ciencias Aplicadas