Browsing by Subject "Hidden Markov modeling"
Now showing 1 - 1 of 1
- Results Per Page
- Sort Options
Item Acoustoelasticity and data processing for acoustic-based corrosion monitoring in structures(2020-05) Dubuc, Brennan Donovan; Salamone, Salvatore; Ferron, Raissa; Kallivokas, Loukas; Wilson, PrestonCorrosion is one of greatest concerns in a variety structures, from cable-stayed and suspension bridges to prestressed and post-tensioned concrete members. The common theme in these structures is corrosion of the load-bearing steel strands. These members carry significantly larger stresses compared to steel reinforcing bars, and corrosion can therefore be particularly catastrophic. Aiming towards real-time diagnosis and prognosis, this thesis investigates active and passive acoustic methods for nondestructive corrosion monitoring in steel strands and prestressed concrete, while incorporating advanced data processing techniques. The main goal is to take advantage of complex acoustic data in new ways, allowing the extraction of richer and more robust corrosion information for longterm monitoring. Guided waves and acoustic emission constitute the active and passive acoustic methods considered, with each designed to target a unique aspect of the corrosion process. Guided waves are used to actively interrogate the stress redistribution within a corroding strand, which may point to loss of load-carrying capacity. On the other hand, acoustic emission is used to passively monitor corrosion and its various mechanisms (e.g. concrete cracking and steel pitting). Several data processing techniques are adapted and proposed to realize these aims, including time-frequency transforms, modal modulation, data fusion, topological data analysis, and hidden Markov modeling. In addition, acoustoelasticity theory is advanced in order to predict the effect of stress on guided wave propagation in strands. Particular emphasis is placed on higher-order guided wave modes, which possess several advantageous characteristics for corrosion-induced stress monitoring. To validate the abovementioned acoustic methods, accelerated corrosion experiments were conducted on loaded strands and small-scale prestressed concrete specimens. The experiments were designed to evaluate the performance of guided waves in monitoring corrosion-induced stress redistribution, as well as acoustic emission in monitoring the evolution of corrosion mechanisms. The results showed that higher-order guided wave modes were able to reveal the underlying trend in stress redistribution, as well as critical moments like wire fracture. In addition, the topology of acoustic emission data was shown to indicate mechanisms appearing at the onset of corrosion. Combining this information with traditional frequency analyses through hidden Markov modeling then allowed for a realtime automated diagnosis of the corrosion process.