Dynamic Network Toolbox for Data-Driven Model Learning and Diagnostics

Descripción del proyecto

Herramientas innovadoras para el análisis de datos de sistemas dinámicos

Los sistemas complejos de ingeniería, como los sistemas eléctricos, los procesos de producción, las redes de transporte y los sistemas biomédicos, requieren la capacidad de modelar, supervisar, optimizar y controlar subsistemas dinámicos interconectados. Asimismo, los modelos precisos y los datos actualizados son indispensables para garantizar unas operaciones seguras, eficientes y sólidas. Sin embargo, el «software» estándar existente carece de la eficacia necesaria para el análisis de datos y el aprendizaje automático que aprovecha la estructura de interconexión física de los subsistemas. En este contexto, el equipo del proyecto SysDyNetTool, financiado por el Consejo Europeo de Investigación, pretende desarrollar una caja de herramientas basada en MATLAB para el análisis de datos de sistemas dinámicos interconectados espacialmente. Esta caja de herramientas ofrecerá una valiosa ayuda a ingenieros, investigadores y estudiantes y se guiará por las aportaciones de las partes interesadas del sector. El equipo del proyecto persistirá en la investigación y el perfeccionamiento de métodos para mejorar las herramientas existentes.

Objetivo

Increasing demands on the safe and efficient operation of engineering systems require the ability to model, monitor, optimize and control complex dynamic systems that are spatially interconnected as networks of dynamic subsystems. Examples can be found e.g. in distributed (smart) power systems, industrial production processes, transportation networks, biomedical systems, etcetera.
Operational decisions are being made on the basis of models and data, while keeping the network models up-to-date over time is key for the ability to guarantee safe, efficient and robust operation. While sensor data is playing an increasing role as a basis for modeling, diagnostics, decision making and predictive maintenance, there are currently no standard software tools available for data analytics and machine learning where effective use is made of the physical interconnection structure of the constituting subsystems.
The results of the ERC Advanced Research project Data-driven modeling in dynamic networks (2016-2022) will be translated into an effective general purpose GUI supported MATLAB-based toolbox for data analytics, including data-driven dynamic modelling and diagnostics, for the situation of spatially interconnected linear dynamic systems. The scientific methods and algorithms will be turned into into effective workflows to be used by engineers, researchers and students for applications in a variety of engineering domains. A professional software architecture and first steps of an implementation have already been realized. The project will be guided by stakeholders from industry, among which MathWorks Inc, ASML and ABB, who will also be involved in the development of use cases. Parallel to this project, the research into effective methods will continue, allowing to overcome shortcomings in the current tools that might appear. A further plan for exploitation of the toolbox will be made by the end of the project, and will be dependent on the proof-of-concept evaluation.