Periodic Reporting for period 2 - MOIRA (MOnItoRing of large scale complex technologicAl systems)
Período documentado: 2023-03-01 hasta 2025-08-31
Modern technological systems increase in scale and are becoming more and more complex and sophisticated. Parallel, the revolution in electronics, digital technology and communications have drastically modified and expanded the physical diversity, scope, processing capabilities and complexity of the monitoring equipment and infrastructure used. Millions of networked sensors are being embedded in the physical world sensing, creating and communicating data. The amount of data available for capturing has been exploding and the era of Big Data is already here, as the Internet of Things (IoT) is becoming a reality. The main question which arises is how, following which steps and with which tools the data can be transformed to information and knowledge. The initial objectives of MOIRA were i) the development of novel signal processing tools for the monitoring of industrial processes based on machine learning methods applied on heterogeneous time series, ii) the application of data mining technologies for the estimation of Key Performance Indicators which determine the operational profit, iii) the conception, development and validation of methodologies for automated monitoring of cyber physical system fleets, iv) the multi sensor machine condition monitoring under variable operating conditions.
MOIRA - MOnItoRing of large scale complex technologicAl systems - brought together early stage researchers and experienced specialists from key players in academia and industry across Europe covering different scientific disciplines and industrial stakeholders from a broad range of backgrounds to optimally tackle the challenges ahead. The MOIRA Fellows were trained in innovative PhD topics as well as receiving specific theoretical and practical education in the fields of mechanical engineering and computer science, focusing towards the online early accurate identification of abnormal incidents with minimum false alarms and missed detections. They successfully developed signal processing and machine learning tools to monitor rotating machinery, e.g. gearboxes, motors, robots, aircraft engines, complex packaging machines, as well as human beings. The research activities led to important results and conclusions which have been published in top conferences and journals. MOIRA project achieved the main goal training 15 ESRs as the new generation of expert engineers and scientists in the field of Knowledge Discovery for online on time accurate monitoring/fault detection/diagnosis/prognosis of large scale complex technological systems for enhanced safety, minimal environmental impact, production and energy efficiency optimisation. The 15 ESRs joined forces and successfully developed novel methods, tools and technologies to analyse heterogeneous data structures captured by various sensors, extract and fuse valuable information at a unit level and at the fleet level, taking real-time monitoring, diagnostics and prognostics in the industrial sectors of Aerospace, Mining, Automotive, Health Care and Packaging to the next level.
MOIRA - MOnItoRing of large scale complex technologicAl systems - brought together early stage researchers and experienced specialists from key players in academia and industry across Europe covering different scientific disciplines and industrial stakeholders from a broad range of backgrounds to optimally tackle the challenges ahead. The MOIRA Fellows were trained in innovative PhD topics as well as receiving specific theoretical and practical education in the fields of mechanical engineering and computer science, focusing towards the online early accurate identification of abnormal incidents with minimum false alarms and missed detections. They successfully developed signal processing and machine learning tools to monitor rotating machinery, e.g. gearboxes, motors, robots, aircraft engines, complex packaging machines, as well as human beings. The research activities led to important results and conclusions which have been published in top conferences and journals. MOIRA project achieved the main goal training 15 ESRs as the new generation of expert engineers and scientists in the field of Knowledge Discovery for online on time accurate monitoring/fault detection/diagnosis/prognosis of large scale complex technological systems for enhanced safety, minimal environmental impact, production and energy efficiency optimisation. The 15 ESRs joined forces and successfully developed novel methods, tools and technologies to analyse heterogeneous data structures captured by various sensors, extract and fuse valuable information at a unit level and at the fleet level, taking real-time monitoring, diagnostics and prognostics in the industrial sectors of Aerospace, Mining, Automotive, Health Care and Packaging to the next level.
The performed work perform in the project was organised in 3 researcher WPs: Processing of Heterogeneous Data (WP1); Fleet Monitoring (WP2); and Multi-Sensor Diagnostics & Prognostics (WP3). WP4 focused on the communication, dissemination and exploitation of the MOIRA results. WP5 and WP7 on management and Ethics run smoothly and via GA meetings and SB meetings, organized twice a year, the project was followed-up closely. WP6 was on training. The Individual training programs were adjusted to the needs of the ESRs, while all foreseen network wide training courses took place.
WP1 focused on "Information extraction from heterogeneous data sources". The lead of this WP was executed by AMC VIBRO. The main achievements are:
− the development of advanced methods for the extraction and exploitation from Heterogeneous signals the correlation of system wear and monitoring indicators
− the development of novel Deep Learning methods for time-series information data fusion
− the development of denoising signal processing techniques for non-Gaussian noise filtering
− the development of data fusion techniques for the exploitation of heterogeneous data
The Work Package 2 (WP2) focused on the development of “Fleet monitoring” methodologies applied to high-frequency, continuous time data captured from a fleet of industrial machines, vehicles and aircraft in operation.
The main achievements are:
• The development of unsupervised self-learning methods based on fleet streaming data
• The development of novel approaches in dynamic modelling, probabilistic fleet models and information geometry
• The development of novel approach in inverse modelling and model selection
• The application of Transfer Learning approaches on simulation data and experimental data for classifications and anomaly detection in vehicle fleets
WP3 focused on the development of advanced methodologies for fault detection, diagnostics, and prognostics in complex technological systems, with a strong emphasis on industrial applications such as aerospace engines, heavy machinery, and autonomous platforms. The WP was led by SAFRAN, leveraging industrial expertise to ensure that research remains aligned with field requirements.
The main achievements are:
• Signal processing and diagnostics: The development of tools for extracting relevant fault signatures from vibratory and acoustic signals acquired in highly nonstationary and noisy environments, with direct application to aircraft engines and mechanical subsystems.
• Health prognostics: The exploration of robust prognostic frameworks, including uncertainty modeling and data-driven forecasting approaches, to anticipate the evolution of component degradation.
• Sensor validation and fusion: The investigation of multi-sensor strategies, sensor selection, and information fusion for enhanced system observability and reliable diagnostics.
• Human-in-the-loop learning: The development of methodologies enabling autonomous agents to integrate expert feedback and learn optimal maintenance behaviors over time.
WP1 focused on "Information extraction from heterogeneous data sources". The lead of this WP was executed by AMC VIBRO. The main achievements are:
− the development of advanced methods for the extraction and exploitation from Heterogeneous signals the correlation of system wear and monitoring indicators
− the development of novel Deep Learning methods for time-series information data fusion
− the development of denoising signal processing techniques for non-Gaussian noise filtering
− the development of data fusion techniques for the exploitation of heterogeneous data
The Work Package 2 (WP2) focused on the development of “Fleet monitoring” methodologies applied to high-frequency, continuous time data captured from a fleet of industrial machines, vehicles and aircraft in operation.
The main achievements are:
• The development of unsupervised self-learning methods based on fleet streaming data
• The development of novel approaches in dynamic modelling, probabilistic fleet models and information geometry
• The development of novel approach in inverse modelling and model selection
• The application of Transfer Learning approaches on simulation data and experimental data for classifications and anomaly detection in vehicle fleets
WP3 focused on the development of advanced methodologies for fault detection, diagnostics, and prognostics in complex technological systems, with a strong emphasis on industrial applications such as aerospace engines, heavy machinery, and autonomous platforms. The WP was led by SAFRAN, leveraging industrial expertise to ensure that research remains aligned with field requirements.
The main achievements are:
• Signal processing and diagnostics: The development of tools for extracting relevant fault signatures from vibratory and acoustic signals acquired in highly nonstationary and noisy environments, with direct application to aircraft engines and mechanical subsystems.
• Health prognostics: The exploration of robust prognostic frameworks, including uncertainty modeling and data-driven forecasting approaches, to anticipate the evolution of component degradation.
• Sensor validation and fusion: The investigation of multi-sensor strategies, sensor selection, and information fusion for enhanced system observability and reliable diagnostics.
• Human-in-the-loop learning: The development of methodologies enabling autonomous agents to integrate expert feedback and learn optimal maintenance behaviors over time.
MOIRA brought together partners from Academia and Industry and trained 15 ESRs in the use of advanced research methodologies to bridge the three important inter-/multi-disciplinary technological challenges of monitoring large-scale technological systems, e.g. the information extraction from heterogeneous data, the fleet monitoring and the multi-sensor diagnostics & prognostics. The ESRs with the support of their supervisors proposed new signal processing and machine learning techniques with direct application in the Aerospace, Mining, Packaging and Health care domains. Approximately 100 articles were presented and published in international conferences and journals, where the methods were applied and validated on real signals from the abovementioned domain demonstrating the direct socioeconomic impact. The developed content is required in the modern economics, industry needs new methodologies, algorithms, technologies and tools that will be able to solve the complex problems of large-scale technological systems. The methodologies developed extract information from massive heterogeneous data structures (continuous, discrete, event driven, scaled) covering operations under variable and non-stationary conditions, achieving the minimization of missing detections and false alarms.