Plenary Lecture

Numerical Optimization in Hydrodynamic Design

Professor John S. Anagnostopoulos
School of Mechanical Engineering
National Technical University of Athens,
Heroon Polytechniou 9, 15780 Zografou, Athens, Greece
j.anagno@fluid.mech.ntua.gr

Abstract: CFD analysis has become a valuable tool in manufacturing of various hydraulic components and machinery and is being increasingly used by the industry not only in the design stage but also in maintenance and troubleshooting, due to its important contribution to the understanding of the development and of the effects of various steady and unsteady or dynamic flow mechanisms.
The inverse design and design optimization methods constitute more recent approaches in engineering, which can exploit the progress in performance simulation methods and the continuous increase of computing power. The aim is to achieve a more automated design procedure and thus to reduce the cost and the duration of this stage, while improving as much as possible the hydrodynamic behaviour of the final products.
The various optimization methods usually combine a number of design variables to create a cost or objective function, the value of which can be computed from the numerical simulation results of the corresponding flow field. Therefore the fluid problem is transformed to a mathematical minimization or maximization problem. The success of such a procedure depends mainly on the careful selection and incorporation of the design variables, which must be as few as possible to reduce the computer time of the algorithm, but at the same time adequate to permit the consideration of all the design possibilities and constraints.
In the simplest optimization type the target is the minimization or maximization of a single flow quantity or mechanism (i.e. flow rate, hydraulic efficiency, hydraulic losses, cavitation etc). However, as will be shown in some examples, the obtained optimal shapes may be strange and/or non-acceptable from the engineering point of view, since the mathematical procedure can arbitrarily select and combine the free design variables from their initial variation range. Therefore, further parametric studies or reconsideration of the problem constraints may be required.
The performance of multi-objective optimizations, where more than one goals (that may be multidisciplinary) are simultaneously desired, can provide superior final designs as will be shown in some examples, since interactions between the objectives are taken into account. In this procedure the progress of the optimizer is towards the so-called Pareto front rather than towards an absolute minimum or maximum. This allows the engineer to choose an acceptable trade-off between the various goals by picking a point somewhere along the Pareto front.
However, increasing the number of objectives complicates the optimization algorithm and makes more difficult the analysis and interpretation of the results. Therefore, the important in this approach is to select only the most critical for the design objectives, which in some complex cases may not be obvious or easy to incorporate. Hence, in every case the experience and the knowledge of fluid mechanics theory are prerequisites for a successful and efficient hydrodynamic design, but both the procedure and the results can be significantly improved with the use of numerical optimization tools.

Brief Biography of the Speaker:
John Anagnostopoulos is currently assistant professor in the School of Mechanical Engineering at the National Technical University of Athens (NTUA), Greece. He received his BS in Mechanical Engineering (1985), and his Ph.D. in Computational Fluid Mechanics (1991), both from the NTUA. He worked for several years as principal researcher in various research projects and as R&T consultant. He specialized in the numerical modelling of the flow field and flow mechanisms in various industrial and physical processes, including pulverized coal combustion, fouling, coal grinding, electrostatic precipitation, atmospheric flows and pollutant dispersion, pollutant formation and photochemical kinetics, pulsating flows, steel continuous casting, metal thermal spraying, mechanical erosion wear, centrifugal pumps and pumping installations, impulse water turbines.
He has developed several comprehensive computer codes: COal Combustion Algorithm (COCA), Modeling of Atmosperic Pollution, (MAP), COal Grinding Algorithm (COGA), Simulation of ELectrostatic Filters (SELF), FLow Automated Solver (FLAS), and he has been involved in feasibility studies for various industrial innovations.
His current interests include the flow analysis and hydrodynamic design optimization in hydraulic turbomachinery and in micropumps, using Eulerian and Lagrangian methods, as well as the optimal sizing and design of hydroelectric, pumped-storage, and hybrid power plants.