The motivation of the work is to investigate a new, non-intrusive condition monitoring system for gas turbines with capabilities for earlier identification of any changes and the possibility of locating the source of the faults. This thesis documents experimental research conducted on a laboratory-scale gas turbine to assess the monitoring capabilities of Acoustic Emission (AE). In particular it focuses on understanding the AE behaviour of gas turbines under various normal and faulty running conditions. A series of tests was performed with the turbine running normally, either idling or with load. Two abnormal running configurations were also instrumented in which the impeller was either prevented from rotation or removed entirely. With the help of demodulated resonance analysis and an ANN it was possible to identify two types of AE; a background broadband source which is associated with gas flow and flow resistance, and a set of spectral frequency peaks which are associated with reverberation in the exhaust and coupling between the alternator and the turbine. A second series of experiments was carried out with an impeller which had been damaged by removal of the tips of some of the blades (two damaged blades and four damaged blades). The results show the potential capability of AE to identify gas turbine blade faults. The AE records showed two obvious indicators of blade faults, the first being that the energy in the AE signals becomes much higher and is distinctly periodic at higher speeds, and the second being the appearance of particular pulse patterns which can be characterized in the demodulated frequency domain.