The SUPREEMO technologies are focused around the MIDIH architecture where, and different tiers were used to support the experiment architecture.
Collecting data and maintenance logs through devices that covers the whole production line.
Customised algorithms for data fusion and advanced analytics that provide meaningful results.
Responsible for collecting data, and through middlewares apply AI algorithms.
Fetching data, and offers user with visualization components, meaningful analytics and DSS system.
The SUPREEMO industrial pilot plant was a food-processing SME for which we had multiannual data on energy consumption and other process attributes, as well as the equipment maintenance log files. In the edge tier we installed a set of non-intrusive custom sensors to monitor electric loads in energy demanding equipment. The sensor PCB was designed, printed and evaluated during SUPREEMO, to collect load data in high frequency and enable the cloud analytics. In the fog tier we deployed a gateway to collect and pre-process the sensor data and transfer them to the platform tier. Pre-processing of the high frequency measurements was essential to reduce the required bandwidth and ensure uninterrupted data transfer to the cloud. In the Platform tier we developed customised machine learning and deep learning algorithms for data fusion and analysis. The results were fed to the Enterprise tier hosting the business application, to provide an interactive interface to inform the process operators on the status of their equipment, and support decision making to improve process availability and energy use.
The results of the SUPREEMO experimentation were evaluated through a set of KPIs.
Reduction of Electricity Consumption Costs
SUPREEMO offered to the pilot an energy monitoring system that analyses consumptions and presents indicators and reports towards actions that reduce the total energy costs of the pilot. During the experiment we assessed recent and past equipment upgrades, we selected and evaluated (using simulations) a set of scenarios suitable for the pilot, and proposed solutions each leading to 3-5% reduction of the current energy costs.
Fault Prediction Accuracy
The collected high frequency load data enabled the identification of signal anomalies indicative of deterioration of the industrial equipment or other problems. Using deep learning technologies, SUPREEMO identified deviations from normal operation and classified some of these deviations early warnings for machine malfunctions. It was later verified and all these warnings were associated with machine malfunctions (and vice versa) which took place a few days or weeks after the warning.
Overall User Acceptance
SUPREEMO developed a decision support system with a friendly interface to allow non-experienced in ICT users to be informed about the energy consumption and device health. The system received an over 85% rating from the pilot personnel in terms of Ease of use (usability), Friendliness of interface, Understandable analytics, Meaningful results and the system’s Integration potential:
Get in touch with us. Would love to answer your questions!
is a co-funded Application Experiment under the MIDIH Open Call 2