Detection of Wrap-Up Incidents in a Napping Machine

In the textile industry, a napping machine is used to raise pile on the surface of the web (knit or woven fabric). As a result of the napping machine’s high speed planetary motion, the web can get tangled in the machine and induces damage to both the machine and the web. By averting wrap-up incidents, it is possible to save maintenance costs, reduce machine downtime, and make the work environment safer. This paper introduces a method for predicting wrap-up incidents in a napping machine to avoid costly damage to the machine. The task of detecting wrap-up incidents is achieved by using indirect sensing of vibration signals from the napping machine. The data collected from the napping machine are represented by discrete wavelet transforms. The features extracted from the coefficients of the discrete wavelet transforms are used as inputs to a multilayer neural network. Once the neural network is trained by using the data specific to the napping machine, data from the machine are processed and fed to the neural network for online wrap-up incident prediction. Several experiments are conducted on a test napping machine, to verify and validate the proposed wrap-up detection scheme. It was found that the vibration signals along the horizontal direction of the main shaft of the napping machine provides an impressive 100% correct wrap-up detection signal.