Hypermodelling strategies on multi-stream time-series data for operational optimization (HYMOP)

Keywords

fleet-based data analytics, hyper-modelling, hybrid-modelling, prognostics, diagnostics, root cause analysis, performance benchmarking

Project description

HYMOP aims to impact the total cost of ownership of systems by optimizing the maintenance and maximizing the availability and the output quality of industrial assets. This not only improves the economic effectiveness of these assets but at the same time contributes to the improvement of their environmental performance, in terms of optimal use of material and energy resources as well as the promotion of renewable energy production.

HYMOP is a basic research project (SBO) with the academic objective to push the scientific state-of-the-art by:

• designing an infrastructure for gathering and processing multiple streams of high-frequency raw sensor measurements;
• bridging the existing gap between physical and data-driven modelling of the behaviour of components, systems and fleets on the basis of these incoming data streams;
• developing richer and more scalable composite models (hypermodels) on the basis of component, system and fleet models that encompass the relations between components in a system, and systems in a fleet;
• evaluating models and hypermodels in real-time in order to predict anomalies, assess physical degradation of a system and to quantify its remaining lifetime;
• providing valuable decision support for operation optimization actions and taking proactive measures to prevent the unexpected downtime of systems due to equipment failure.

These new developments will lead to innovative data analysis and modelling techniques that are able to cope with significant amounts of streaming data in a real-time context, in line with major challenges confirmed by Flemish lead user companies related to:

• robust and cost-effective capturing, storing and processing of multi-stream high-frequency time-series data originating from different sensors embedded in different components of different assets within a fleet;
• pro-active rather than reactive exploitation of the data to predict the future behaviour of the assets, to estimate remaining useful life, to identify anomalies and to support planning and decision making related to operation and maintenance.

For HYMOP, we developed innovative modelling and data processing/analysis techniques that are able to cope with large amounts of complex data in real-time. Thanks to advanced data mining techniques, we exploited the opportunities available in data recorded by sensorised and connected industrial assets. Finally, we chaired all valorisation and dissemination activities of the project.

Project Partners

Industrial Advisory Committe

Project details

October 2015 - September 2019

With the support of (met de steun van)

Further information

Here