Storm and wastewater pipes form a significant proportion of the UK's hydraulic infrastructure, the maintenance of which presents key challenges. Ageing pipes face stresses from climate change and urbanisation, demanding robust monitoring solutions. Current monitoring practices utilise spot sensing methods, which fail to provide real-time continuous data on the internal hydraulics of these turbulent pipe flows. While emerging studies utilise free-surface dynamics, characterising turbulence via the fluctuating wall pressure in partially-filled pipes remains under-explored. Fibre-optic sensing non-invasively monitors these conduits from within the pipe wall through its response to the wall pressure. This thesis details the development, implementation, and analysis of a novel flush-mounted fibre-optic sensor designed to infer properties of turbulent partially-filled pipe flows directly from its response to wall pressure fluctuations.
Firstly, the Approximate Bayesian Computation-sequential Monte Carlo scheme is adapted to infer a high-dimensional binary system parameter encoding boundary condition information of a thin plate structure housing a fibre-optic sensor. Applied to vibrational data, it infers boundary condition configurations, with metrics demonstrating how convergence depends on user-design choices.
This is followed by the development and application of a novel flush-mounted fibre-optic sensor in an experimental facility, where its response to two partially-filled pipe flow regimes was analysed. Cross-correlation analyses of sensor data enabled accurate recovery of bulk flow velocity. A score-based Metropolis-Hastings algorithm was developed to identify optimal frequency bands retaining flow dynamics of interest. Azimuthal variation of sensor response to wall pressure fluctuations was recovered, including streamwise correlation structures, frequency-wavenumber spectra, and large-scale motion signatures. Their organisation enabled linkage to secondary flow motions. These findings were compared against Large-Eddy Simulation data of equivalent flow regimes. While showing reasonable qualitative agreement, simulation limitations led to slight quantitative differences between results. Regardless, findings support the sensors ability to recover complex turbulent phenomena in partially-filled pipe flows.
