Open Access

ARTICLE

Steel Ball Defect Detection System Using Automatic Vertical Rotating Mechanism and Convolutional Neural Network

Yi-Ze Wu, Yi-Cheng Huang^*

Department of Mechanical Engineering, National Chung Hsing University, Taichung, 402202, Taiwan

* Corresponding Author: Yi-Cheng Huang. Email:

(This article belongs to the Special Issue: Selected Papers from the International Multi-Conference on Engineering and Technology Innovation 2024 (IMETI2024))

Computers, Materials & Continua 2025, 83(1), 97-114. https://doi.org/10.32604/cmc.2025.063441

Received 14 January 2025; Accepted 26 February 2025; Issue published 26 March 2025

Abstract

Precision steel balls are critical components in precision bearings. Surface defects on the steel balls will significantly reduce their useful life and cause linear or rotational transmission errors. Human visual inspection of precision steel balls demands significant labor work. Besides, human inspection cannot maintain consistent quality assurance. To address these limitations and reduce inspection time, a convolutional neural network (CNN) based optical inspection system has been developed that automatically detects steel ball defects using a novel designated vertical mechanism. During image detection processing, two key challenges were addressed and resolved. They are the reflection caused by the coaxial light onto the ball center and the image deformation appearing at the edge of the steel balls. The special vertical rotating mechanism utilizing a spinning rod along with a spiral track was developed to enable successful and reliable full steel ball surface inspection during the rod rotation. The combination of the spinning rod and the spiral rotating component effectively rotates the steel ball to facilitate capturing complete surface images. Geometric calculations demonstrate that the steel balls can be completely inspected through specific rotation degrees, with the surface fully captured in 12 photo shots. These images are then analyzed by a CNN to determine surface quality defects. This study presents a new inspection method that enables the entire examination of steel ball surfaces. The successful development of this innovative automated optical inspection system with CNN represents a significant advancement in inspection quality control for precision steel balls.

---

Keywords

---