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)

274

13

relatively well under such conditions. The rolling statistics in the same figure confirm that signal variations remain bounded and slowly varying, which limits the risk of large prediction errors.

Figure 7: Stationarity analysis of block damage signals at two representative frequencies (258.7 Hz and 507.5 Hz)

However, this also points to an important limitation: the highway dataset used here may not fully represent more complex, variable, or harsh vibration environments. The comparison was carried out under conditions where damage accumulation behaves almost ideally. In such cases, more sophisticated probabilistic approaches do not have a significant advantage over simpler models. To strengthen the general conclusions, future work should include evaluations on larger and more diverse datasets, especially those exhibiting strong non-stationary behavior or sudden transient events. It would also be valuable to investigate alternative prediction methods, including hybrid or adaptive models, and assess their robustness under less predictable operating conditions. 6. Conclusion A proof of concept embedded in a small box (95*60*45 mm) has been defined and carried out to assess the possibility of processing in real time the acceleration measured at the foot of a sensitive equipment to monitor the harshness of its mechanical environment. The processing is perform in temporal domain, in accordance with the recommendations of standards as AFNOR XP X50-144 and the disjoint block method, which allows all types of signals to be processed. Traditionally, vibration data are collected for a short period with temporary instrumentation and analyzed offline to calculate ERS and FDS, which characterize respectively the shock harshness and the induced-fatigue harshness. Although FDS in particular is computationally intensive in time-domain, we have demonstrated the possibility to calculate in real time on board with the POC NOMAD the ERS and the FDS. These performances lead to the new features that NOMAD offers: • The ERS and the FDS are continuously calculated for each time disjoint blocks during all the life of the equipment, validating at any time that the equipment is safe, as long as the envelop off all ERS and the sum off all FDS remain each below the qualification ones. • The variability of ERS and the FDS are assessed with a high degree of accuracy and confidence, which lead to assess the predicted ERS and FDS for future usage by introducing statistical approaches based on XRS and XFS spectra.