ALBUQUERQUE, Rodrigo Barbosa da Fonseca e; http://lattes.cnpq.br/8519519088025779
Resumo:
This work presents the development of a low cost computational methodology for the
conceptual design optimization of axial-flow hydraulic turbines (propeller turbines). The
methodology has been developed with a quasi-two dimensional flow model, employing
empirical correlations for cascade losses and flow deviations. The study is based on the
conservation principles for mass, energy and momentum. The radial equilibrium equation is
included in order to achieve a more realistic flow field. For reducing the number of design
variables, the runner blading stagger, chord-pitch ratio and camber are parameterized in terms
of their values at the hub, mean and tip stations. The design optimization algorithm has been
coded in MatLab™ language. This code searches for a basic geometry that maximizes the
turbine efficiency, given the design flow rate, rotational speed and bounds for the design
variables and also for the available head.
Two optimization techniques have been applied: a gradient based local search method
and a population set-based global search algorithm. The gradient based technique is a
standard Sequential Quadratic Programming, using the fmincon function from MatLab™,
which searches for local minimizers starting from an initial point. For the global searches, it is
presented the Controlled Random Search Algorithms, some of their versions and their
advantages in the design optimization problem.
An application example of the methodology is presented and discussed for the
optimization of a real tube type propeller turbine, previously tested in a laboratory rig. The
optimized solutions are compared with the original turbine design, showing the performance
improvements, according to the hydrodynamic modeling. Suggestions for methodology
improvements are also made.