Dynamic Yarn-Tension Detection Using Machine Vision Combined with a Tension Observer.

Area of Science:

• Textile Engineering
• Applied Physics
• Embedded Systems Engineering

Background:

• Contact measurement of yarn tension causes stress, hairiness, and breakage.
• Existing machine vision systems have speed limitations due to image processing.
• Axially moving models for tension detection neglect motor vibration disturbances.

Purpose of the Study:

• To develop an embedded system integrating machine vision with a tension observer for enhanced yarn tension detection.
• To improve the accuracy and update rate of non-contact yarn tension measurement.
• To address the limitations of speed and vibration disturbances in current methods.

Main Methods:

• Established a differential equation for transverse string dynamics using Hamilton's principle.
• Implemented image acquisition on a field-programmable gate array (FPGA) and processing on a multi-core digital signal processor (DSP).
• Utilized adaptive weighted data fusion in a programmable logic controller (PLC) to combine machine vision and tension observer data.

Main Results:

• The combined tension detection method demonstrated improved accuracy compared to individual non-contact methods.
• The system achieved a faster update rate, overcoming the inadequate sampling rate of pure machine vision.
• The brightest centerline grey value was used to determine the feature line for yarn vibration frequency analysis.

Conclusions:

• The proposed embedded system effectively enhances yarn tension detection accuracy and speed.
• The integration of machine vision and a tension observer provides a robust solution for real-time monitoring.
• This system offers a viable foundation for future real-time control applications in textile manufacturing.