Design of a low noise turbofan duct via an acoustic gradient-enhanced Kriging method

Abstract

The Kriging models which are frequently used in aerodynamic shape optimization may become computationally inefficient when solving problems with large numbers of design variables. One solution to this problem would be the application of gradient-enhanced Kriging model. A gradient-enhanced Kriging and acoustic adjoint-method approach to duct acoustic problems is developed, aimed to improve the efficiency and accuracy of the existing Kriging approach at acoustic problems with many design parameters. To our knowledge, it is the first application of gradient-enhanced Kriging for duct acoustic problem. It employs a Kriging response surface in the parameter space, augmented with gradients obtained from the acoustic adjoint equations efficiently. The present paper aims at describing the potential of the gradient-enhanced Kriging method for low noise turbofan duct design. Prior to the optimization process, the implementation of the unsteady aeroacoustic adjoint method in shape optimization is validated by comparing the gradient values with that obtained by finite differences. In this work, the ordinary Kriging model and gradient-enhanced Kriging method are applied firstly to a benchmark functions and the results show that the additional gradient information can significantly enhance the accuracy of Kriging model. And then, the original Kriging-based, adjoint-based and the gradient-enhanced Kriging method are all used to model 50 variable duct acoustic problems, respectively. The test results show that this approach whose gradient information is introduced by using acoustic adjoint method developed from multimode LEE, named as acoustic gradient-enhanced Kriging, can significantly enhance the accuracy of Kriging models when the gradient data are available and thus provide an optimized low noise intake while maintaining the aerodynamic performance.

Machine Learning Reduced Order Model for Cost and Emission Assessment of a Pyrolysis System

“…The Kriging method, also known as the Gaussian process regression method, is a metamodeling technique that employs the Gaussian process to develop mathematical piecewise linear equations representing interpolation of the input data based on prior covariances . The Kriging method has been applied to various spatial analysis applications and computer experiment applications. − However, it has not been used extensively for modeling pyrolysis reactions.…”

Optimization of a vacuum cleaner fan suction and shaft power using Kriging surrogate model and MIGA

“…A recent trend in the field of turbomachinery optimization involves the combination of surrogate models and single or multi-objective optimization methods to enhance the aeroacoustic or vibrational characteristics of different fluid machinery such as aero-engine fan, 37 small centrifugal fan, 38 centrifugal pump, 39 helicopter rotor, 40 turbofan duct, 41 forwardcurved blades centrifugal fan 36 and horizontal-axis wind turbine. 42 Despite the mentioned investigations in this section, detailed review of the published papers proves that concurrent optimization of the vacuum cleaner fan components (impeller, diffuser and return channel) has never been carried out.…”

“…A recent trend in the field of turbomachinery optimization involves the combination of surrogate models and single or multi-objective optimization methods to enhance the aeroacoustic or vibrational characteristics of different fluid machinery such as aero-engine fan, 37 small centrifugal fan, 38 centrifugal pump, 39 helicopter rotor, 40 turbofan duct, 41 forward-curved blades centrifugal fan 36 and horizontal-axis wind turbine. 42…”