ENBIS Spring Meeting 2017

28 – 30 May 2017; Monastery of Schlägl in Upper Austria Abstract submission: 11 November 2016 – 5 March 2017

Optimized Production Processes Using Sensor Data Analysis

29 May 2017, 11:35 – 11:55

Abstract

Submitted by
Alessandro Tarchini
Authors
Simone Lombardi (MathWorks), Sarah Drewes (MathWorks)
Abstract
Mondi Gronau cooperated with MathWorks Consulting to optimize production processes in the polymer film industry. The considered approach minimized waste production, downtimes and energy consumption while increasing production quality.

In the main part of the project, a recommendation system was created that combines automated sensor data analysis and human experience. In this application, process information is constantly
updated. Deviations trigger a warning message for the machine operator, such than he can intervene to reduce waste production.

These recommendations are based on prediction models created with MATLAB using machine learning on historical sensor data. Per machine, hundreds of sensors (temperature, pressure, etc.) are monitored per minute.
At the same time, quality states of the produced polymer film are also automatically captured.
The sensor data is cleaned and consolidated with the state information, and different machine learning methods are evaluated on historical data. The most robust models with the best prediction accuracy is then used for the predictions.
The system is integrated into the existing IT infrastructure and is used for an increasing number of production machines.

Return to programme