Passive detection is a general methodology that estimates a structure's condition using ambient noise as probing waves, which could be excited in many industrial scenarios, such as during aircraft flight. It is beneficial for many online monitoring scenarios because it does not need any wave excitation equipment. Its principle is that the system’s impulse response can be approximated by the cross-correlation of ambient noise measured by a pair of receivers under the diffuse field assumption, meaning that waves arrive from every direction with equal probability. In practice, however, this assumption may not hold, but the sources could be unpredictable and strongly directional, such that the passive detection cannot be applied. To address this problem, this paper proposes a wave manipulation scheme by exploring the fact that the convergence of cross-correlation relies on waves from the stationary phase zone (SPZ) and is contaminated by sources outside SPZ, where SPZ is defined by the region almost on the same line as the sensor pair. After a quantitative formulation of SPZ given the considered frequency range, artificial reflectors and absorbers are designed to enhance SPZ and remove non-SPZ waves, such that the convergence of cross-correlation can be restored. The proposed method is justified to be efficient by experimental data and successfully applied to the passive ice detection of plate-like structures. The application scenarios of the passive detection methodology are thus largely expanded, because the diffuse field assumption about the source distribution is no longer a must.

Passive ultrasonic guided waves techniques provide a promising alternative to active measurements for the monitoring of structures such as pipes, ship hulls or aircraft fuselage, by taking advantage of the otherwise unwanted mechanical noise caused by ambient excitation. They enable the use of a receivers-only instrumentation such as optical fiber-based sensors, instead of conventional piezoelectric transducers. Consequently, the generation of these waves is entirely dependent on the properties of the ambient excitation (power, frequency content, spatial distribution, etc.), that may vary significantly among use cases and industries. In this study, we investigate the use of passive ultrasonic guided waves techniques for ship hull monitoring, through an experimental campaign in a research vessel under operational conditions. Several piezoelectric transducers are bonded to the inside of the ship’s hull, and signals are recorded while sailing. Active measurements are also performed using separate transducers for actuation, for comparison purposes. The time-frequency content of raw passive signals is first studied, and contributions from different sources of mechanical and electronical noise are identified. Passive guided waves signals are then successfully reconstructed using existing techniques on a low frequency range, and compare pretty well to active signals. The influence of the ship’s sonar, which pollutes measurements at higher frequencies, is also discussed. These results pave the way for the Structural Health Monitoring of ships using new, innovative passive instrumentations such as fiber Bragg gratings.

Pipes play a crucial role in transporting essential resources such as water, oil, and gas across industrial,
urban, and environmental infrastructures. Monitoring of the flow capacity in such extended structures has is a still persisting issue, potentially resulting in operational disruptions and serious safety risks. The current demand on these systems' reliability have been increasing maintenance costs. Therefore, any advance in diagnosis methods and tools is benefic.
This study presents a prognostic and health monitoring approach that utilizes flow-induced acoustic emissions to detect and characterize pipeline blockages. An analytical model and a finite element is developed to capture the acoustic signatures flow-induced disturbances on the structure, and how the acoustic wave propagation is affected by the level of clogging. This reveals features highly sensitive to the changes of the flow rate.
These insights drive the development of a machine learning-based predictive maintenance strategy, validated on real-case datasets. The results demonstrate exceptional accuracy, with most classifiers achieving 100% detection rates for clogging presence, shape, and severity. Additionally, model generalization tests show that machine learning algorithms adapt more effectively to varying clogging thickness than clogging shape. This research is the first step for more enhancing predictive maintenance and ensuring the reliability of industrial pipelines.

Monitoring of structures in operation using guided waves must take into account ambient noise which can significantly perturb signals. However, recent signal processing techniques originating from geophysics, demonstrated that this ambient noise includes important information about structure's health. Indeed, it is possible to recover the impulse response between two sensors by cross correlating the ambient noise acquired by these two sensors. This approach, called "passive method", offers several advantages compared to the traditional "active method". Indeed, no emission device is thus necessary and lighter less intrusive sensors unable to emit waves, like Fiber Bragg Gratings on optical fibers, can then be used, offering new possibilities of system operating in severe environments for instance. Theoretically, these methods are described by assuming several hypotheses especially the equipartition of the ambient noise. However, determining if a given ambient noise meets the needed hypotheses is still an open question. Moreover, while ambient noise for various applications has been mostly studied in the audible range, we are interested in dealing with frequencies around 100 kHz, in which we still manage to recover some exploitable signal. In this paper, we study ambient noise excitation through the case of a flow of water inside a pipe, producing noise though its interaction with the structure. First, a numercial model is proposed, in which a turbulent boundary layer is considered as the noise source. It is modeled using the Corcos model to generate a pressure field applied at the inner boundary of the pipe. This model is also compared to experimental data acquired in a duct for different flows. The acquired noise is characterized through its autocorrelation term, for which it is found that its evolution was comparable to the one at high frequencies, where the spectrum is overall influenced by viscosity. Properties of reconstructed signals using a passive method, namely the cross-correlation of the acquired noise signals, are also studied for both simulation and experiment. We focus on the reconstruction of helical wave packets, which can be described according to the number of revolutions performed in the path between a sensor and an other, as these wave packets seem especially impacted by the ambient noise nature.