The advent of Artificial Intelligence has also revolutionised Industry 4.0, affecting the way we perceive, monitor and diagnose problems in industrial environments.
All this change and at times uncertainty, leads industrial companies and workers to face challenges, such as the variability of working conditions, limitations to transfer large volumes of data securely and efficiently or the challenges of sustainability.
What technological solutions does the NEXTMON project offer?
The NEXTMON project hopes to deliver technological solutions to address monitoring challenges that have already been identified but remain unresolved commercially in a variety of sectors and markets. The scientific and technological objectives of the NEXTMON project will focus on the following lines:
- Hybridisation of (DAQ) acquisition systems with IIoT technologies.
- Monitoring of the condition of systems with variable operating regimes.
- Reduction in the number of sources of data required for monitoring.
- Integration of information from multimodal 2D, 3D and hyperspectral sensors.
- Life cycle management of AI models for different types of information.
- Federated systems with distributed intelligence.
Scientific objectives and lines of action
Monitoring of 1D signals in variable regimes: Order Tracking (OT) is an effective technique to analyse non-stationary vibration signals, monitor and diagnose faults.
Separation of signal sources: Various techniques under the umbrella of Blind Source Separation (BBS) enable signals to be separated according to their source of origin.
Monitoring the health status of wave energy devices for predictive maintenance:
- Vibration analysis: vibration monitoring allows for the early detection of deviations from normal behaviour, and thus timely maintenance to prevent unexpected faults.
- ESA (Electrical Signature Analysis): encompasses a variety of techniques for monitoring the status of electrical machines by analysing electrical signals, such as current and voltage. MCSA (Motor Current Signature Analysis) is a methodology that involves analysing the frequency of electrical current from motors or electrical generators to detect and diagnose faults or anomalies in the system.
2D, 3D and hyperspectral perception in recycling streams:
- Spatial sensor co-registration (HSI, RGB, 3D)
- Dimensionality reduction, sensor optimisation
- Development of advanced, multimodal, on-line implementable perception models.
- MLOps: Continuous learning and data-centred techniques. Life cycle management of AI models for different types of information.
- Integration of information from multimodal 2D, 3D and hyperspectral sensors.
Data management system. The evolution of information technologies and the development of new tools for the storage and management of streaming data now allow for the advanced management of the data collected through tools aimed at time series data management, distributed storage or streaming processing. Use of InfluxDB as a time series database and its subsequent integration into analysis flows.
Federated systems: Federated learning, as a key methodology within this framework, enables collaborative training of artificial intelligence models without sharing the underlying data, addressing fundamental data privacy and security concerns. Within the scope of the AISYM4MED project, INYCOM is developing a federated system that allows health data sources to be used in the training of the same algorithm without the need to consolidate them beforehand.
Project reference: CPP2023-010810
Programme: 2023 AEI Public-Private Partnership Projects
