Iot System and Convolutional Neural Networks for Anomaly Detection in Fabric Dyeing Machines in A Textile Company.

Abstract

In the process of fabric dyeing in textile companies, several problems can occur, mostly due to a kinking in the fabric ropes, which stop and originate losses such as degradé or veteaduras in production. The objective of this article was to develop an IoT system and convolutional neural networks (CNN) for the detection of anomalies in fabric dyeing machines in a textile company. Since traditionally the dyeing machines have a magnet, which takes approximately 2.5 minutes to give a complete rotation. SCRUM methodology was used for the development of the project and supervised learning algorithms were used to detect anomalies in the machines.

As a result, it was decided to place an IP camera in the hatch of each rope and with a supervised algorithm based on convolutional neural networks, an accuracy of 85% was obtained in detecting if there is a rope in "stop or in motion" sending alerts in 2 seconds maximum to the operators through smartwatches connected to the local network via WIFI.