Multiobjective Optimization Better Propulsion Design

  • Published
  • By Joseph Gordon, Douglas Davis
  • Propulsion
WRIGHT-PATTERSON AIR FORCE BASE, Ohio --  Optimally performing propulsion systems require that the competing needs (i.e., individual design parameters) of various components be considered holistically--enter Air Force Research Laboratory research partner Prairie View A&M University and its successful demonstration of a solution that does just that. Known as multiobjective optimization, the approach merges elements of computational fluid dynamics and surrogate modeling techniques to achieve desired functional results; it also reduces the sizeable computing costs of optimizing complex propulsion designs via traditional means.

With the goal of optimizing a supersonic inlet for demonstration purposes, PVA&MU's Dr. Ziaul Huque employed Pareto-Optimal Front, a method involving design-variable surfaces wherein the optimal design can be chosen based on the importance of multiple objective functions. In this case, Dr. Huque selected a response-surface-based methodology, prompting CFD simulations of various inlet designs based on a matrix of variables/parameters for developing the response surface. This optimization effort is easily extended to follow-up activities in order to include more design variables and objective functions (e.g., nozzle and combustor).

Also completed was a study of the inlet's performance trends, with mass flow rate through the inlet and entropy gain used as the objective functions. Angle of attack, axial length from the ramp tip to the cowl tip, and inlet Mach number were the three variables used for the CFD simulations, with inlet dynamic pressure kept constant at 1500 psf. Execution of the inlet CFD simulations occurred via CHEM, a general-purpose, multidimensional, multispecies, viscous chemistry solver built upon a rules-based specification system. The standard, least square method generates response surface by the statistical code JMP. The elitist NGSA [Non-Dominated-Sorting Genetic Algorithm] II determined the Pareto-optimal solution. Each Pareto-optimal solution represents a different compromise between design objectives. This gives the designer a choice to select a design compromise that best suits the requirements from a set of optimal solutions.