Search results

The purpose of this work is to apply and compare surrogate-assisted and multi-fidelity, multi-objective optimization (MOO) algorithms to simulation-based aerodynamic design exploration.

The three algorithms for multi-objective aerodynamic optimization compared in this work are the combination of evolutionary algorithms, design space reduction and surrogate models, the multi-fidelity point-by-point Pareto set identification and the multi-fidelity sequential domain patching (SDP) Pareto set identification. The algorithms are applied to three cases, namely, an analytical test case, the design of transonic airfoil shapes and the design of subsonic wing shapes, and are evaluated based on the resulting best possible trade-offs and the computational overhead.

The results show that all three algorithms yield comparable best possible trade-offs for all the test cases. For the aerodynamic test cases, the multi-fidelity Pareto set identification algorithms outperform the surrogate-assisted evolutionary algorithm by up to 50 per cent in terms of cost. Furthermore, the point-by-point algorithm is around 27 per cent more efficient than the SDP algorithm.

The novelty of this work includes the first applications of the SDP algorithm to multi-fidelity aerodynamic design exploration, the first comparison of these multi-fidelity MOO algorithms and new results of a complex simulation-based multi-objective aerodynamic design of subsonic wing shapes involving two conflicting criteria, several nonlinear constraints and over ten design variables.

Compared with the low-fidelity model, the high-fidelity model has both the advantage of high accuracy, and the disadvantage of low efficiency and high cost. A series of multi-fidelity surrogate modelling method were developed to give full play to the respective advantages of both low-fidelity and high-fidelity models. However, most multi-fidelity surrogate modelling methods are sensitive to the amount of high-fidelity data. The purpose of this paper is to propose a multi fidelity surrogate modelling method whose accuracy is less dependent on the amount of high-fidelity data.

A multi-fidelity surrogate modelling method based on neural networks was proposed in this paper, which utilizes transfer learning ideas to explore the correlation between different fidelity datasets. A low-fidelity neural network was built by using a sufficient amount of low-fidelity data, which was then finetuned by a very small amount of HF data to obtain a multi-fidelity neural network based on this correlation.

Numerical examples were used in this paper, which proved the validity of the proposed method, and the influence of neural network hyper-parameters on the prediction accuracy of the multi-fidelity model was discussed.

Through the comparison with existing methods, case study shows that when the number of high-fidelity sample points is very small, the R-square of the proposed model exceeds the existing model by more than 0.3, which shows that the proposed method can be applied to reducing the cost of complex engineering design problems.

The purpose of this paper is to reduce the overall computational time of aerodynamic shape optimization that involves accurate high-fidelity simulation models.

The proposed approach is based on the surrogate-based optimization paradigm. In particular, multi-fidelity surrogate models are used in the optimization process in place of the computationally expensive high-fidelity model. The multi-fidelity surrogate is constructed using physics-based low-fidelity models and a proper correction. This work introduces a novel correction methodology – referred to as the adaptive response prediction (ARP). The ARP technique corrects the low-fidelity model response, represented by the airfoil pressure distribution, through suitable horizontal and vertical adjustments.

Numerical investigations show the feasibility of solving real-world problems involving optimization of transonic airfoil shapes and accurate computational fluid dynamics simulation models of such surfaces. The results show that the proposed approach outperforms traditional surrogate-based approaches.

The proposed aerodynamic design optimization algorithm is novel and holistic. In particular, the ARP correction technique is original. The algorithm is useful for fast design of aerodynamic surfaces using high-fidelity simulation data in moderately sized search spaces, which is challenging using conventional methods because of excessive computational costs.

Aerodynamic shape optimisation is a complex problem usually governed by transonic non-linear aerodynamics, a high dimensional design space and high computational cost. Consequently, the use of a numerical simulation approach can become prohibitive for some applications. This paper aims to propose a computationally efficient multi-fidelity method for the optimisation of two-dimensional axisymmetric aero-engine nacelles.

The nacelle optimisation approach combines a gradient-free algorithm with a multi-fidelity surrogate model. Machine learning based on artificial neural networks (ANN) is used as the modelling technique because of its ability to handle non-linear behaviour. The multi-fidelity method combines Reynolds-averaged Navier Stokes and Euler CFD calculations as high- and low-fidelity, respectively.

Ratios of low- and high-fidelity training samples to degrees of freedom of n_LF/n_DOFs = 50 and n_HF/n_DOFs = 12.5 provided a surrogate model with a root mean squared error less than 5% and a similar convergence to the optimal design space when compared with the equivalent CFD-in-the-loop optimisation. Similar nacelle geometries and aerodynamic flow topologies were obtained for down-selected designs with a reduction of 92% in the computational cost. This highlights the potential benefits of this multi-fidelity approach for aerodynamic optimisation within a preliminary design stage.

The application of a multi-fidelity technique based on ANN to the aerodynamic shape optimisation problem of isolated nacelles is the key novelty of this work. The multi-fidelity aspect of the method advances current practices based on single-fidelity surrogate models and offers further reductions in computational cost to meet industrial design timescales. Additionally, guidelines in terms of low- and high-fidelity sample sizes relative to the number of design variables have been established.

Extreme support vector regression (ESVR) has been widely used in the design, analysis and optimization of engineering systems of its fast training speed and good computational ability. However, the ESVR model is only able to utilize one-fidelity information of engineering system. To solve this issue, this paper extends extreme support vector regression (ESVR) to a multi-fidelity surrogate (MFS) model which can make use of a few expensive but higher-fidelity (HF) samples and a lot of inaccurate but cheap low-fidelity (LF) samples, named ESVR-MFS.

In the ESVR-MFS model, a kernel matrix is designed to evaluate the relationship between the HF and LF samples. The root mean square error of HF samples is used as the training error metric, and the optimal hyper-parameters of the kernel matrix are obtained through a heuristic algorithm.

A number of numerical problems and three engineering problems are used to compare the ESVR-MFS model with the single-fidelity ESVR model and two benchmark MFS models. The results show that the ESVR-MFS model exhibits competitive performance in both numerical cases and practical cases tested in this work.

The proposed approach exhibits great capability for practical multi-fidelity engineering design problems.

A MFS model is proposed based on ESVR, which can make full use of the advantages of both HF data and LF data to achieve optimal results at same or lower cost.

Reliable modeling of induction hardening requires a multi-physical approach, which makes it time-consuming. In designing an induction hardening system, combining such model with an optimization technique allows managing a high number of design variables. However, this could lead to a tremendous overall computational cost. This paper aims to reduce the computational time of an optimal design problem by making use of multi-fidelity modeling and parallel computing.

In the multi-fidelity framework, the “high-fidelity” model couples the electromagnetic, thermal and metallurgical fields. It predicts the phase transformations during both the heating and cooling stages. The “low-fidelity” model is instead limited to the heating step. Its inaccuracy is counterbalanced by its cheapness, which makes it suitable for exploring the design space in optimization. Then, the use of co-Kriging allows merging information from different fidelity models and predicting good design candidates. Field evaluations of both models occur in parallel.

In the design of an induction heating system, the synergy between the “high-fidelity” and “low-fidelity” model, together with use of surrogates and parallel computing could reduce up to one order of magnitude the overall computational cost.

On one hand, multi-physical modeling of induction hardening implies a better understanding of the process, resulting in further potential process improvements. On the other hand, the optimization technique could be applied to many other computationally intensive real-life problems.

This paper highlights how parallel multi-fidelity optimization could be used in designing an induction hardening system.

This paper aims to present a multi-fidelity surrogate-based optimization (MFSBO) method for computationally accurate and efficient design of microfluidic concentration gradient generators (µCGGs).

Cokriging-based multi-fidelity surrogate model (MFSM) is constructed to combine data with varying fidelities and computational costs to accelerate the optimization process and improve design accuracy. An adaptive sampling approach based on parallel infill of multiple low-fidelity (LF) samples without notably adding computation burden is developed. The proposed optimization framework is compared with a surrogate-based optimization (SBO) method that relies on data from a single source, and a conventional multi-fidelity adaptive sampling and optimization method in terms of the convergence rate and design accuracy.

The results demonstrate that proposed MFSBO method allows faster convergence and better designs than SBO for all case studies with 49% more reduction in the objective function value on average. It is also found that parallel infill (MFSBO-4) with four LF samples, enables more robust, efficient and accurate designs than conventional multi-fidelity infill (MFSBO-1) that only adopts one LF sample during each iteration for more complex optimization problems.

A MFSM based on cokriging method is constructed to utilize data with varying fidelities, accuracies and computational costs for µCGG design. A parallel infill strategy based on multiple infill criteria is developed to accelerate the convergence and improve the design accuracy of optimization. The proposed methodology is proved to be a feasible method for µCGG design and its computational efficiency is verified.

A multi-disciplinary robust design optimization method for micro Mars entry probe (no more than 0.8 m in diameter) is proposed. The purpose of this paper is to design a Mars entry probe, not only the geometric configuration, but the trajectory and thermal protection system (TPS). In the design optimization, the uncertainties of atmospheric and aerodynamic parameters are taken into account. The probability distribution information of the uncertainties are supposed to be unknown in the design. To ensure accuracy levels, time-consuming numerical models are coupled in the optimization. Multi-fidelity approach is designed for model management to balance the computational cost and accuracy.

Uncertainties which cannot defined by usual Gaussian probability distribution are modeled with degree of belief, and optimized through with multiple-objective optimization method. The optimization objectives are set to be the thermal performance of the probe TPS and the corresponding belief values. Robust Pareto front is obtained by an improved multi-objective density estimator algorithm. Multi-fidelity management is performed with an Artificial Neural Network (ANN) surrogate model. Analytical model is used first, and then with the improvement of accuracy, rather complex numerical models are activated. ANN updates the database during the optimization, and makes the solutions finally converge to a high-level accuracy.

The optimization method provides a way for conducting complex design optimization involving multi-discipline and multi-fidelity models. Uncertainty effects are analyzed and optimized through multi-disciplinary robust design. Because of the micro size, and uncertain impacts of aerodynamic and atmospheric parameters, simulation results show the performance trade-off by the uncertainties. Therefore an effective robust design is necessary for micro entry probe, particularly when details of model parameter are not available.

The optimization is performed through a new developed multi-objective density estimator algorithm. Affinity propagation algorithm partitions adaptively the samples by passing and analyzing messages between data points. Local principle component techniques are employed to resample and reproduce new individuals in each cluster. A strategy similar to NSGA-II selects data with better performance, and converges to the Pareto front. Models with different fidelity levels are incorporated in the multi-disciplinary design via ANN surrogate model. Database of aerodynamic coefficients is updated in an online manner. The computational time is greatly reduced while keeping nearly the same accuracy level of high fidelity model.

The purpose of this paper is to present a method for analysis and optimization of morphing wing. Moreover, a numerical advantage of morphing airfoil wing, typically assessed in simplified two-dimensional analysis is found using higher fidelity methods.

Because of multi-point nature of morphing wing optimization, an approach for optimization by analysis is presented. Starting from naïve parametrization, multi-fidelity aerodynamic data are used to construct response surface model. From the model, many significant information are extracted related to parameters effect on objective; hence, design sensitivity and, ultimately, optimal solution can be found.

The method was tested on benchmark problem, with some easy-to-predict results. All of them were confirmed, along with additional information on morphing trailing edge wings. It was found that wing with morphing trailing edge has around 10 per cent lower drag for the same lift requirement when compared to conventional design.

It is demonstrated that providing a smooth surface on wing gives substantial improvement in multi-purpose aircrafts. Details on how this is achieved are described. The metodology and results presented in current paper can be used in further development of morphing wing.

Most of literature describing morphing airfoil design, optimization or calculations, performs only 2D analysis. Furthermore, the comparison is often based on low-fidelity aerodynamic models. This paper uses 3D, multi-fidelity aerodynamic models. The results confirm that this approach reveals information unavailable with simplified models.

The purpose of this paper is to investigate an active flow control technique called Shock Control Bump (SCB) for drag reduction using evolutionary algorithms.

A hierarchical genetic algorithm (HGA) consisting of multi‐fidelity models in three hierarchical topological layers is explored to speed up the design optimization process. The top layer consists of a single sub‐population operating on a precise model. On the middle layer, two sub‐populations operate on a model of intermediate accuracy. The bottom layer, consisting of four sub‐populations (two for each middle layer populations), operates on a coarse model. It is well‐known that genetic algorithms (GAs) are different from deterministic optimization tools in mimicking biological evolution based on Darwinian principle. In HGAs process, each population is handled by GA and the best genetic information obtained in the second or third layer migrates to the first or second layer for refinement.

The method was validated on a real life optimization problem consisting of two‐dimensional SCB design optimization installed on a natural laminar flow airfoil (RAE5243). Numerical results show that HGA is more efficient and achieves more drag reduction compared to a single population based GA.

Although the idea of HGA approach is not new, the novelty of this paper is to combine it with mesh/meshless methods and multi‐fidelity flow analyzers. To take the full benefit of using hierarchical topology, the following conditions are implemented: the first layer uses a precise meshless Euler solver with fine cloud of points, the second layer uses a hybrid mesh/meshless Euler solver with intermediate mesh/clouds of points, the third layer uses a less fine mesh with Euler solver to explore efficiently the search space with large mutation span.