Session: 01-01/05-07 Joint Session: Aero-Engine Control & Diagnostics

Paper Number: 152290

Natural Noise Feature Extraction in Sensor Data for Enabling Digital Twins and Real-Time Anomaly Detection 

Digital Twins (DT) are essential for complex systems because they provide a real-time digital replica of physical assets, allowing for continuous monitoring, simulation, and optimization. This enables predictive maintenance, operational efficiency, and enhanced decision-making. A fundamental component of digital twins is anomaly detection, which identifies deviations from normal operations and ensures system reliability. Data-driven anomaly detection is vital for DT of complex systems like turbofans as it is ensuring improved resilience, timely response, and preventing costly damage or downtime. The significant presence of natural noise in real-world sensor data from turbofans, caused by system dynamics and operating conditions, poses challenges for accurate anomaly detection. This research addresses the gap in synthetic data generation, forecasting, and simulation by integrating natural noise into simulated data. 

This study explores feature engineering techniques for extracting and identifying natural noise behavior from experimental turbo fan data, which can be leveraged in digital twin models of cyber-physical systems. After extracting the behavior of natural noise from experimental data, real-life-like synthetic noise can be generated. This approach enables the creation of more realistic system states by injecting natural noise into data generated through simulations based on physical/numerical models. This method provides an accurate benchmark for comparing simulated data with real-time data collected during the system’s actual operation, facilitating improved accuracy in real-time, data-driven anomaly detection, with potential benefits for predictive maintenance and system diagnostics.

Presenting Author: Antonio Ficarella Green Engine Lab, Department of Engineering for Innovation, University of Salento

Presenting Author Biography: Ali AGHAZADEH ARDEBILI holds a Ph.D. in Engineering Risk Analysis from the University of Trieste and a 2nd Ph.D. at the UniSalento in Complex System's Engineering; he is currently working at the department of the Research and Development of HSPI SpA and collaborating with CRISR research center and Data Lab at University of Salento on cyber-Physical Systems Resilience through Digital Twining. His research field is Digital Twins, UAS/RPAS/AAM, Cyber-Physical-Social Systems, uncertainty analysis/assessment, engineering/project risks, Critical Infrastructures resilience, and Data-Oriented solutions. He has 4 patents, and 8 years of professional career experience in international engineering design/consultant companies. He speaks English, Italian, Persian, Azerbaijani, and Turkish. In 2018 he was selected among the 13 redefiners of the next 100 years in N100 Symposium out of 650 applicants from 55 countries; he was peace ambassador certified by the global peace chain from 2018 to 2020. In 2019 he won the TATA Steel challenge award for a sustainable solution for HYSARNA industrial steel production process.

Authors:

Ali Aghazadeh Ardebili HSPI Roma
Antonella Longo Data Lab, Department of Engineering for Innovation, University of Salento
Antonio Ficarella Green Engine Lab, Department of Engineering for Innovation, University of Salento