Experimental Measurements of Gas Turbine Flow Capacity Using a Novel Transient Technique

Abstract

Nozzle guide vane (NGV) flow capacity is perhaps the most important parameter for engine optimization. Inaccurate evaluation of capacity would lead to incorrect performance evaluation, and unmatched stages. A new semi-transient technique has been developed and demonstrated that will allow turbine designers to measure experimentally the effective throat area of an annular cascade of nozzle guide vanes at engine-representative Mach number, Reynolds number, and mainstream-to-cooling-flow pressure ratio. The technique allows NGV capacity to be measured to bias and precision uncertainties to 95% confidence of ±0.546% and ±0.028%, which compares well to large scale industrial facilities. Order-of-magnitude cost savings are offered over typical continuously running industrial facilities by running in blowdown mode from a receiver tank, thus removing the need for large scale compressor plant. To demonstrate the technique, a high mass flow rate blowdown tunnel was developed and commissioned at the University of Oxford, and the capacity of a high-pressure NGV from a large civil aircraft engine was experimentally determined. Experimental results are presented, which allow the precision error to be accurately calculated. A detailed uncertainty analysis is given from which the bias error is computed. It is shown that the low precision error the new technique offers means that it is ideally suited to investigations in which secondary influences on capacity are the subject of the investigation. The technique is of industrial significance because methods to compute engine capacity analytically or computationally do not yet provide sufficient accuracy for engine optimization, and the new technique offers equivalent accuracy at a much reduced cost over conventional experimental techniques. By performing an uncertainty analysis using experimental data it is shown that the increase in uncertainty due to the semi-transient (as opposed to steady state) nature of the technique is approximately 0.004% (to 95% confidence), and is entirely negligible. The experimentally measured trend of capacity against pressure ratio is compared with simple 1D, 2D, and 3D inviscid models, and an analytical correction for total pressure loss is performed. It is shown that while a simple 3D model is better than a 1D model (up to 1.5% improvement) for crude estimates of engine capacity, experimental trends are poorly predicted by such simple techniques. An analytical correction for total pressure loss increases the difference between 3D prediction and experiment. The experimental data demonstrate the complex nature of the process by which nozzle capacity is determined and the need for accurate, low-cost experimental techniques for capacity measurement. Correction to engine conditions is discussed.