PREDICTIVE MAINTENANCE IN CASE OF FILTER BLOCKAGE BY TRAINING AN AI

As a pioneer in the field of modern software development, we have set ourselves the goal in a new project together with our colleagues from SME 4.0 Competence Centre for Textile Networks Hahn-Schickard to train an AI for the anticipatory filter change in water treatment plants.

This water pump contains a filter that filters large and fine particles from the water flowing through it and, if it works well, will eventually block them. When this happens, the machines receive a signal and trigger an - as yet - unpredictable emergency stop. The machines shut down unexpectedly and suddenly stop working until the filter has been changed.

tepcon recognised early on that this reduces the productivity of the machines and has collected data from a water treatment plant over the last two years. How does this work? In a water treatment plant a pump pumps water against a filter. The pump causes a vibration in the system. In this project, among other things, the vibration data is converted into graphic images in order to identify vibration patterns depending on the degree of filter contamination. The aim is to use the data from a structure-borne sound sensor attached to the pump to predict the status or life cycle of the water filter and when it will become clogged. For the purposes of predictive maintenance, a filter change will in future be indicated with a lead time of three days.

In order to get closer to the project goal of pattern recognition and, based on this, to train AI, Hahn-Schickard use regression models from the field of machine learning and - where possible - deep learning. In this way, values are created from information.

Diese Website speichert Cookies auf Ihrem Computer. Diese ermöglichen eine bessere Dienstbarkeit unserer Website. Mehr Infos über die von uns eingesetzten Cookies finden Sie in unserer Datennutzungserklärung.

Please fill in the form bellow