Mathematical methods of image processing and Condition Monitoring play a key role in industrial production – in our »Re(Pro)³« project we show how predictive models help to reduce waste and optimize processes sustainably.
In the project we combine Condition Monitoring and Predictive Maintenance with automated quality monitoring. In this way, we help companies to detect errors in production at an earlier stage – both in the product itself and in the manufacturing process, in order to significantly reduce reject rates through the digitalization of production processes. The core idea of Re(Pro)³ is to combine methods from image processing and Condition Monitoring in order to develop an image data-based system that integrates temporal changes in product qualities and process variables caused by wear, defects or misconfiguration. The aim is to sustainably reduce avoidable rejects in automated production processes.
Classic Condition Monitoring keeps an eye on the status of machines, plants or production systems during operation. It is based on the analysis of measured values (e.g. temperature, vibrations, pressure, acoustic signals) obtained from sensors or imaging systems.
Predictive Maintenance goes beyond Condition Monitoring by using data analysis, algorithms (e.g. Machine Learning) and predictive models to make forecasts about the future condition of a system. It determines the optimum time for maintenance before a failure occurs.
Automated Surface Inspection for Flawless Products Including Changes Over Time
Automated surface inspection systems monitor the quality of products using imaging processes. These systems detect and classify defects on surfaces. Defective products are then sorted out or reworked. However, further analysis, e.g. of the development over time, does not currently take place. However, in addition to operating and control errors, the cause of defective products often lies in the wear of individual components of the production system. By observing how product quality changes over time, it is possible to predict at an early stage when fault tolerances will be exceeded and maintenance of the production system will be required.
By analyzing defect patterns, which are determined in advance based on virtual inspection planning, and combining them with other measurement data from the production plant, the causes of defects can be identified more precisely and the development of defects over time can be predicted. On this basis, targeted measures can be taken, such as adapting the system control through intelligent control or the timely replacement of certain components that are likely to fail. This means that fewer faulty products are produced, downtimes in the production process are reduced and the sustainability of the entire production process is significantly increased.
Applications and Software Framework
The result of our project is a software framework for the sustainable, resource-optimized operation of production facilities using image-based processes. We see future applications in various sectors such as the textile industry, the plastics and metal industry, the packaging industry or the building materials industry (e.g. wall cladding made of fiber materials).
For further information and details, please visit our project website: www.itwm.fraunhofer.de/repro3-en