This thesis describes applications of NMR techniques to flowing liquid streams to obtain quantitative information about the contents of the streams. The quantitative accuracy of NMR spectroscopy for composition measurement of liquid mixtures is measured as ±0.34 mol% and ±1 mol% for static and flowing mixtures respectively.
The effects of flow on NMR spectroscopy are analysed using the residence time distributions of the streams in the magnet and the detection coil. Algorithms are developed for automated analysis of the NMR spectra of the mixtures, in which automatic phase and baseline correction are performed together. A peak-assignment algorithm is written that identifies components in a mixture based on the patterns observed in the pure-component spectra. Automated composition analysis of mixture spectra is performed using these algorithms in less than 4 minutes with an accuracy of ±0.66 mol%. A mathematical model is derived for the NMR spectrum of a mixture that considers the spectrum a weighted sum of pure-component spectra shifted in frequency. The experimental lineshape observed in an inhomogeneous magnetic field is poorly fitted by a Lorentzian lineshape, so a new model lineshape is developed based on the distribution of resonance frequencies across the sample. Volume selective NMR spectroscopy using the STEAM and PROJSAT pulse sequences is optimised to give quantitative results from well-defined volumes with minimal signal contamination. The STEAM pulse sequence is modified to include flow-compensated slice selection gradients. The accuracy of the compositions measured from volume selective spectra is measured as ±1 mol% and ±2 mol% for static and flowing mixtures respectively. Pulsed field gradient NMR sequences using double echoes for flow compensation are tested on flowing water, then used to determine the droplet size distributions of flowing emulsions. Flow images are acquired of a vertical liquid jets showing the narrowing and acceleration of the jet and the entrainment of the surrounding water.