TITLE:
Simulation Study on Axis Offset Prediction of Convolutional Neural Network Based on Structured Light
AUTHORS:
Jingwen Zhang, Junkai Yu, Kun Xia, Zhe Gou, Jianping Tan
KEYWORDS:
Checkered Plate Structure Light, Shaft Offset, Convolutional Neural Network, OpenCV Post-Processing
JOURNAL NAME:
Open Journal of Applied Sciences,
Vol.15 No.11,
November
17,
2025
ABSTRACT: Helicopters assume a pivotal role in domains such as disaster relief, transportation, and military operations. The transmission system stands as one of the three essential components of a helicopter, with the tail drive shaft system tasked with transmitting power to the tail rotor. Although the supercritical design of helicopter tail drive shaft systems presents numerous merits both at home and abroad, the amplitude experiences a sharp increase when traversing the first-order critical speed. In the absence of vibration measurement, monitoring, and control, this can readily lead to overload, triggering collisions between the shaft system and other components, consequently resulting in accidents. With the escalating demand for measurement accuracy in the industrial realm, detecting and predicting the offset of high-speed rotating shafts represents an extremely challenging yet crucial objective in mechanical engineering. This research employs structured light technology to conduct simulation experiments on the projected structured light patterns of both straight and bent shafts. Structured light projects known patterns or light rays (e.g., laser beams or checkerboard patterns) onto the surfaces of objects, and surface deformation is computed based on the altered light patterns. Through 3Ds Max software model rendering and post-processing using OpenCV, we observed the variations of projected patterns across different shaft offsets, trained a Convolutional Neural Network (CNN) model, and applied the CNN to predict the offset values of unknown shafts.