Smart Monitoring for Energy Efficiency and Predictive Maintenance – Application to Electric Motors Retrofitting


Experiment description

The SUPREEMO experiment proposed a data-driven approach using CPS and IoT technologies to deliver solutions for energy efficiency and predictive maintenance. The experiment involved the non-intrusive load monitoring (NILM) of industrial Electric Motor-Driven Systems (EMDS). EMDS are commonly found in pumps, fans, compressors and material handling equipment, they consume around 2/3 of the electrical power used in industry, and their estimated energy efficiency potential is around 10%.

The SUPREEMO system was installed and tested in the olive oil refinery plant of ELSAP S.A., and custom sensors were installed on various devices ranging from pumps to centrifugal units, pressure chambers and boilers. The sensors were designed, printed and tested during SUPREEMO, and they collected high-frequency electric load data to enable advanced analytics of the obtained signals.

The experiment created a 60TB database of hard-to-find industrial data, which enabled the development, training and testing of custom ML and DL algorithms to analyze high frequency energy load data. It finally presented a suite of robust tools for fault prediction in industrial equipment operation (with prediction windows ranging from a few days to a few weeks), as well as a general methodology to propose tangible energy efficiency measures. In summary, the experiment:

–          developed a robust system for the early detection of signal anomalies, possibly leading to equipment malfunction, and therefore energy or production losses, and eventually unexpected device breakdowns.

–          linked this system to the DSS, where the facility personnel had access to realtime equipment status information, and alerts when signal anomalies occurred, so that inspections and (if needed) predictive maintenance actions can be planed.

–          applied systematic procedures for the identification of energy efficiency potential and practical solutions to improve energy efficiency and energy costs, and delivered a set of solutions based on the constraints and practicalities of the pilot process.

Technical impact

The SUPREEMO solution utilized Cloud – Edge architecture to enable the collection, processing, transmission and storage of the large volumes of electricity datastreams. The implementation of this system had several challenges. Using the MIDIH components it was possible to reduce the cost, increase the integration potential with 3rd party services and the scalability; and deliver a robust and compact working version in less than 4 months. Among the different components of the MIDIH architecture, the Kafka broker was used to handle the data exchange between the fog device and the cloud infrastructure, and MongoDB was used for the databases. The selection of these components allowed the transfer and storage of the very large volumes of data, and the realtime visualization of the collected measurements. SPARK enabled the creation of the live data streams to feed the analysis modules, and ensured the necessary speed in data flow for realtime analysis.

Economical / Business impact

The initial business concept was to offer advanced, flexible and affordable Industry 4.0 solutions for the reduction of energy costs, and the improvement of process efficiency by monitoring the device health and predicting future malfunctions.

The experiment met all three key performance indicators:

  • Energy saving solutions with over 3-5% reduction of total energy costs;
  • Equipment fault prediction and classification accuracy over 90%;
  • A user-friendly Decision Support System (DSS) featuring an overall user satisfaction score above 80%;

and enabled improvements on various business aspects, including energy efficiency and process sustainability, equipment efficiency and maintainability, and production losses due to equipment unavailability. All these translate to overall reduction of operational and maintenance costs if the pilot plant continues to use the tools tested and developed during SUPREEMO.

Now that the experiment is completed, the developed tools have been installed and tested in real industrial conditions, and their results have been validated against real observations (showing for instance that all the predicted signal anomalies resulted to malfunctions). The collaboration with the industrial pilot ELSAP S.A. was excellent, and their input was extremely important to understand the real needs of future customers, and how to place the SUPREEMO solution within an industrial perspective.

The procedures and systems developed under SUPREEMO can be easily applied to other industrial plants in the future. The DEMOKRITOS team is now working into the next steps to improve these tools and bring them closer to the market, and significant progress has been made in seeking additional pilots and funding.

Information source: Midih



Smart Monitoring for Energy Efficiency and Predictive Maintenance – Application to Electric Motors Retrofitting