Providing Faulty Thread Detection in Tie Rod End Grease Fittings with Machine Learning Method “YOLO” algorithm and “Smart-VS”

Abstract

The tie rod end transmits the movement from the steering box to the wheels of the vehicle. The
joint in the tie rod ends ensures that the pushing motion is transmitted in a mobile way. These joints must
be lubricated with grease to reduce the friction force. In order to easily perform the lubrication process,
there are greases that have an oily ball at the end and can be easily disassembled. Manufacturer-induced
faults may occur in greasers. Frequently encountered tooth production errors can cause serious problems.
A greaser with a thread defect can be fitted with a tight fit and disconnected from the assembly under
fatigue. For this reason, greasers must be checked and separated during assembly. Defect control systems
used in industry are generally carried out manually. At the same time, manual control varies according to
operator competence and initiative. High efficiency, low cost and objectivity can be achieved with
machine learning-based systems. One of the machine learning techniques, the deep learning algorithm
YOLO (You Only Look Once) is extremely fast and sharp. In the supervised learning process, it is
necessary to obtain the data that is desired to be determined, and to label and train the data. In this study,
a detailed investigation of YOLO-based object detection with object detection methods designed in recent
years, the effect of data labeling on detection results and Smart-VS Smart sensor performance
comparisons were made.