Yarn periodical errors determination using three signal processing approaches

This paper presents a study undertaken to identify the type and location of yarn periodical errors using three different signal processing approaches based on FFT – Fast Fourier Transform, FWHT – Fast Walsh–Hadamard Transform and FDFI – Fast Impulse Frequency Determination. Available commercial equipment is based exclusively on an FFT approach which is unable to clearly detect all types of periodical yarn errors, particularly impulse errors. The theoretical basis of each the three signal processing techniques is described. Their performance when applied to several simulated errors, namely, pulse errors and impulse errors is analyzed in detail. Finally, the three different techniquesʼ ability to successfully characterize errors in actual data taken from real textile yarns with known sinusoidal errors is explored and commented upon.