关键词: 奇异值分解算法, 纱线振动, 图像处理, 纱线张力, 非接触式检测

Abstract:

Objective Yarn tension is closely related to product quality and production efficiency. The size and stability of tension run through each process from spinning to manufacturing. The excessive tension of yarn will lead to irreversible deformation of yarn, which will not only increase the yarn breaking rate, but also affect the mechanical strength, surface performance, dyeing performance and process structure of the fabric. The excessive tension of yarn will lead to poor formation of fabric organization, poor structure and poor elasticity. Modern technology requires the size and stability of yarn tension is increasingly high, so it is extremely important to realize the real-time measurement of yarn tension in operation.

Method The singular value decomposition (SVD) algorithm is adopted to obtain the amplitude frequency information by reducing the dimension of video image data, recombining vibration displacement extraction, and iterative denoising. With the help of fast Fourier transform, the yarn vibration time domain characteristics are converted into frequency domain characteristics and draw the frequency domain map, and finally, the yarn vibration monitoring experiment platform is built to test the feasibility and reliability of the algorithm.

Results An experimental set-up was built and experimentally verified to test the feasibility and reliability of the provided scheme. Different yarn running speeds were set and the experimentally derived tensions were compared with the measured tension magnitudes during yarn movement. The results indicated that when the speed of yarn movement was increased, the vibration amplitude of the yarn became smaller, the vibration frequency of the yarn larger, and the tension of the yarn larger. The tension of the yarn and the vibration frequency of the yarn were positively correlated, which is consistent with the theoretical equation of yarn vibration. Statistical results of yarn tension calculated by conventional image processing showed that when the yarn motion speed was 50-70 mm/s, the experimental value of yarn tension was close to the measured value of yarn tension with an absolute error of no more than 4%. However, when the yarn speed exceeded 75 mm/s. the yarn was irreversibly deformed due to the excessive tension and friction between the yarn and mechanical structure such as yarn guide wheels, and the yarn demonstrated a sudden change in the linear density, resulting in an absolute error of more than 10% occurs between the experimental and test values. The algorithm was computationally fast and accurate, and the yarn tension could be measured in real time with good performance.

Conclusion The results of experiments show that the non-contact yarn tension measurement based on machine vision successfully solves the problem of inaccurate tension values caused by the contact between the yarn and the measuring element during the contact yarn tension measurement, and the measurement accuracy can meet the performance requirements of most textile processes for yarn tension.

Key words: singular value decomposition algorithm, yarn vibration, image processing, yarn tension, non-contact detection