The application of signal analysis techniques based on chaos theory to flow regime identification.

The aim of the research presented in this thesis has been to develop an objective measurement technique to improve the detection of flow patterns in closed ducts. This activity is important for the safe and efficient running of many processes, particularly within the oil production, nuclear power, chemical and process industries. Signal analysis techniques based on nonlinear dynamic (chaos) theory have been applied to simulated and experimental transducer signals measuring properties of gas-liquid (air-water) flows in horizontal and vertical pipes. The techniques provide a method of measuring properties of the signals that are related to patterns within the signals. Signals from various non-invasive transducers (including differential pressure transducers, an electrical conductance transducer, a light attenuation transducer, an ultrasonic transducer and a gamma-ray densitometer) have been analysed. Signal analysis techniques include the use of singular value decomposition, the correlation integral and power spectra analysis. The results of signal analysis on the simulated signals illustrate their potential for flow regime identification. When applied to experimental signals it is shown that changes in some of the signal characteristics correlate well with changes in the flow regimes. Discernment between horizontal stratified-wavy, plug and slug and vertical slug and bubbly flow regimes has been achieved. The most successful analysis technique (using singular value composition) is more robust than previously used techniques and can be computed much more efficiently.