A vibration-based approach for health monitoring of tie-rods under uncertain environmental conditions

Automatic Detection of Real Damage in Operating Tie-Rods

“…Finally the ROC curves for this case are presented in Figure 20: again, it is possible to observe how the performance of the damage-detection algorithm was improved by the adoption of the data-cleansing procedure. Additionally in this case, the strategy could successfully detect damage, and the results were remarkable considering the extent and severity of the damage in a location close to the constraints where the index is less sensitive to damage, as shown in [24]. Moreover, it is worth noticing that, in this case, the baseline set was shorter than that of damage at 𝜉 = 𝐿 (compare timelines reported in Figures 11 and 17) due to the fact that the corrosion attack started earlier and that the tie-rod was used for another experimental test between 1 January 2020 and 22 April 2020.…”

“…The position of the accelerometer is indicated by a black-solid vertical line at ξ = L/10: it is possible to observe that such a position potentially allowed for the identification of the considered vibration modes. The tie-rods were equipped with sensors to replicate a long-term structural health monitoring system for research purposes (e.g., [24,38]). More specifically, each tie-rod was equipped with four general-purpose industrial piezo-electric accelerometers (PCB 603C01 model with a sensitivity of 10.2 mV/(m/s 2 ) and full scale of ±490 m/s 2 ).…”

“…If the procedure described in Section 2 can be successfully used to detect damage in a tie-rod, as shown in [ 24 ], a fundamental step toward real applications is the development of a completely automatic algorithm with no need for human expert supervision. The automatization of the procedure to obtain the data needed by the strategy described in Section 2 is explained below.…”

“…To overcome these limits, the authors of [ 24 ] proposed a vibration-based, data-driven approach to tie-rod damage detection that showed the potential to detect damage under the effects of an uncontrolled environment without knowledge of physical variables (e.g., tie-rod tension) and without the need for a reference structure. The key point of the proposed approach is that when a pattern of modal parameters is considered instead of just a single one, the damage-detection problem can be treated as a multivariate outlier detection problem.…”

“…This strategy relies on a damage feature defined only by tie-rod eigenfrequencies which could be identified with a single accelerometer properly placed on the monitored tie-rod. This allows for a simpler experimental set-up with respect to the one adopted in [ 24 ]. The obvious economic consequences of adopting a simple experimental set-up can help the transition from research to real applications.…”