PSI - Issue 75

Mohamed El Yazrhi et al. / Procedia Structural Integrity 75 (2025) 262–275 Mohamed El Yazrhi , Jean-Yves Disson / Structural Integrity Procedia (2025)

266

5

NOMAD computes the Extreme Response Spectrum (ERS) and the Fatigue Damage Spectrum (FDS) using the time domain method described in Section 2. When enabled, a pre-trained machine learning model classifies the usage condition based on vibration indicators, and in the medium term, this classification can also integrate additional data sources such as CAN signals from vehicles. To ensure that the system operates in real time despite the computational intensity of FDS calculations, NOMAD uses parallel processing techniques to reduce execution time. Stress tests have shown that the platform can reliably handle 3-axis signals sampled at 10 kHz with up to 300 frequency points. It is also possible to limit the ERS/FDS calculations to specific frequency bands identified as critical during equipment qualification tests or early design phases. 3.3. Data storage and prognosis For each processed block, NOMAD stores key information in its embedded memory, including the timestamp, calculated ERS and FDS indicators, the identified usage class (if classification is enabled), and GPS coordinates in vehicle applications. In cases where predefined thresholds are exceeded — such as during extreme vibration events — the system can also save the raw time signal for further analysis. This real-time processing approach enables continuous monitoring throughout the entire operational lifespan of the equipment while minimizing data volume. Instead of storing tens of thousands of raw samples per block, NOMAD records only a few hundred ERS and FDS values, significantly reducing memory requirements. Additionally, the system constantly updates the envelope of ERS and the cumulative sum of FDS and compares them against the reference limits defined during equipment qualification. As long as these indicators remain below the thresholds, the equipment is considered safe. By combining this monitoring with fatigue extrapolation methods described in Section 4, NOMAD also provides an estimate of the remaining useful life (RUL), enabling predictive maintenance and avoiding unexpected failures.

Figure 2 : Real-Time Fatigue Monitoring and Remaining Life Estimation in Track Driving Mode