This thesis investigates the research developments and performance estimation of radial inflow turbines. This is in aid of the current project being undertaken by the Queensland
Geothermal Energy Centre of Excellence (QGECE). QGECE have set the goal of designing and building a working 1MW turbine by mid 2014 and the intermediate goal of a 20kW working turbine in the laboratory loop currently under construction in Mansergh Shaw Building 45 at the University of Queensland. They are conducting research into using supercritical carbon dioxide in a closed loop Brayton cycle to produce geothermal power from hot dry rocks in the ground at Innamincka, South Australia.
This thesis focuses on the review of radial turbine performance within the power loop, including prediction theory and modelling software from Concepts NREC. It provides details
of a FORTRAN program from NASA on the design analysis of radial turbines around which a Python code was constructed in order to automate its execution. Results were obtained
from the FORTRAN program and the RITAL software from Concepts NREC for two separate flow conditions (geothermal site and laboratory model), each for air and CO2. Results were compared for the air and CO2 conditions to illustrate the feasibility of using CO2 as the working fluid. Also for the laboratory flow conditions, the results from the FORTRAN program were compared with those from RITAL in order to emphasize or nullify the validity of the models.
The literature review revealed the current technologies in turbomachinery as well as the current progress of the project directed by QGECE and the timeframe for deliverables. The review also highlighted two distinct computer based solvers for modelling radial in-flow turbine performance, they included a FORTRAN program from NASA and software from an
American based company Concepts NREC. Research into turbochargers was also carried out to enable a better understanding of its implementation into the laboratory loop as the
The motivation for using radial turbines was established through specific speed calculations for the highest efficiency. It was found that radial turbines are most efficient when their
specific speeds are Ns = 0.4 – 1.0. For the conditions proposed by QGECE, the specific speed of the turbine in the loop lies within these bounds. It was also motivated by the fact that radial turbines are easier to manufacture than their axial counterparts. A standardized table of nomenclature of the turbine was obtained for ease of understanding in future references.
The software from Concepts NREC comprised of a meanline solver called RITAL and a 3D geometric solver called AxCent. RITAL was used to compute the performance analysis of the turbine for the conditions in the laboratory and at the geothermal site for both air and CO2. Results showed that the CO2 model for the laboratory flow conditions was less efficient than the air model. It also showed that it required larger geometries of the turbine and a lower shaft speed. The CO2 model for the geothermal site conditions had equal efficiency to the air model, however it produced 31.6% less power due to the smaller required geometries of the turbine.
The FORTRAN program automated by the Python code was also used to compute the performance analysis of the turbine for the conditions in the laboratory and at the geothermal site for both air and CO2. This was conducted by analysing the performance of the turbine at conditions on and near those proposed in order to illustrate methods to improve performance. The results also showed that CO2 had reduced efficiency but generally within 2% of the air model.
A comparison between the FORTRAN program and RITAL software was carried out in order to attain the validity of the models. This comparison revealed that the FORTRAN program produced turbine geometries that are below the RITAL results and efficiencies which are above those from RITAL. It was deduced that the FORTAN code produces a more ideal and
conservative estimate of performance than RITAL. The rotor exit relative flow angle had the greatest difference error due to the fact that the FORTRAN program calculates exit values at several places on the rotor whereas RITAL calculates an overall value.
Finally the validity of the model was assumed unresolved as there was no experimental data to compare with the computer based solvers. It was recommended that some changes be
made to the FORTRAN and Python format statements for easier code execution as well as a comparison between the FORTRAN code and RITAL source code in order to resolve the issue of the sensitivity of the FORTRAN program, to improve its loss model and calculate more output values. Once these have been completed it was recommended to compare with experimental data in order to bring insight into the accuracy of the simulation to real life situations.