Significane of loss models in aerothermodynamic simulation for axial turbines

University dissertation from Stockholm : Energiteknik

Author: Ning Wei; Kth.; [2000]

Keywords: ;

Abstract: This thesis deals with a study of the significance of lossmodels and their applications in simulation and optimisation ofaxial turbines. The design of a turbomachine system is a verycomplex engineering operation which can be looked upon as aniterative procedure made of various steps. Computational toolsfor simulating and optimising turbomachines are needed nowadaysin order to design turbines more effectively and efficiently.In order to use the tools to analyse elaborate thermodynamiccycles and optimise the cycles for different assignments,aerothermodynamic performances over each blade row in turbineshave to be predicted in a correct trend and reasonable accuratein a wide operating range. Because the flow in a turbine iscomplex and many mechanisms of the flow losses in turbine havenot be known well, loss models are needed, not only in thepreliminary process of mean line prediction but also in thefurther process of through flow calculation, in the simulationand optimisation of turbines. Most of the loss models areempirical while some of them are established with combinationof test data and analysis of physical origins of thelosses.The objectives of this thesis are to further understand anddevelop loss models and therefore to achieve useful guides forapplying the models properly in turbine aerothermodynamicsimulation and optimisation.In this thesis, concepts of the flow field, loss mechanismsand classification and definition of loss coefficient in axialturbines are introduced in Chapter 1. In Chapter 2, some lossmodels published in the literature for axial turbines arereviewed and studied in detail. An effort is also made, in thischapter, to develop a simple method for approximatelypredictingoverall film cooling losses in a turbine blade rowbased on equation of mass, momentum and energy conservation inmixing flows. In Chapter 3, five axial turbine stages, whichhave been used for the study, are introduced. These turbinestages are with different geometrical parameters, with impulseand reaction and untwisted and free vortex blading, and work atdifferent flow conditions, in subsonic and supersonic flow andcooling and non-cooling conditions. In Chapter 4, mean lineperformances of these turbine stages were simulated withdifferent loss models which are the models byAinley/Mathieson/Dunham/Came (AMDC), Kacker/Okapuu, Craig/Coxand Moustaph/Kacker supplemented with the new developed coolingloss prediction method. The simulations were performed withinlarge operating ranges which cover design and off-designpoints. The predicted results from the mean line performancecalculation in the loss models were compared with theexperimental data. In Chapter 5, an analysis of the optimumpitch/chord ratio for both the stator and rotor in one of theseturbine stages was made through studying the total lossespredicted in the loss models.It was found that all these loss models give the same trendof overall performance compared with the trend of experimentalresults on the turbine stages. For the impulse turbine stages,it was showed that all these loss models seem to overestimatelosses in the whole operation range and the AMDC model is onewhich most overestimates the losses while the Craig/Cox modelgives the loss closest to the experimental results at and nearthe design point. When the aspect ratios in the impulse turbinestages increase about 25%, the results showed that the modelsgive about 15-30% lower total losses. For the free vortexblading turbine, it was found that the losses areunderpredicted in all the models, especially on the off-designoperation points, because the high losses at the hub wallcaused by the blading of free vortex are underestimated byusing the parameters at mean line. For the high pressure andtemperature cooled turbine stage, the predicted results showedthat all these models predict higher efficiencies than theexperiments because the lack of cooling loss calculation in themodels. The AMDC and Craig/Cox models seem to give closeresults to the experiments in this cooling turbine stagebecause of the fortuity of overestimated uncooled turbinelosses. After adding the cooling losses predicted with the newdeveloped method in the thesis, the trends of simulatedperformances in this turbine stage have better agreement withthe trend of experimental data. In the analysis of the optimumpitch/chord ratio, the results showed that the pitch/chordratio evaluated by total losses is not very critical and allthe models gave similar low loss regions of pitch/chord ratiosin which the total losses do not change significantly. Thisimplies there will be no significant difference between theseloss models if they are employed to obtain the optimumpitch/chord ratio in turbine optimisation process.