Aerodynamic shape design using evolutionary algorithms and new gradient-assisted metamodels

Surrogate models and mixtures of experts in aerodynamic performance prediction for aircraft mission analysis

“…Exploration criteria aim to "fill the gaps" between existing sample points to ensure that the samples are evenly distributed spatially. This category includes sequential space-filling sampling plans such as that of Sobol' [90] and the Halton sampling sequence [21], as well as an adaptive approach that locates infill points with the highest estimated error (e.g., using the kriging variance as a metric). In general, maximizing the variance when adding samples tends to maximize the inter-site distances (D-optimality) [15].…”

“…Essentially, surrogate models are used as low-cost substitutes to replace expensive evaluations when the original physics-based models are used in any computationally intensive tasks (e.g., repeated analysis or optimization) [20]. These approximation models are also known as metamodels [87] or models of models [28,63].…”

“…The Halton sampling sequence, which is a space-filling lowdiscrepancy method [21], is used to generate training samples to construct the surrogate models, as well as to generate the clustering training data for the ME models. The discrepancy in this case refers to the departure of the sampling points from a uniform distribution, thus ensuring an even distribution of samples over the input space.…”

Multi‐objective turbomachinery optimization using a gradient‐enhanced multi‐layer perceptron

“…This idea has been tried in conjunction with both kriging [5] and ANNs [6]. The expected advantage of a gradient-enhanced RSM is a higher accuracy of the response surface, especially noticeable in higher-dimensional design spaces across gaps in the sampled objective values.…”

Linear Evolutionary Algorithm

“…The major disadvantages of EAs are their poor performance in handling constraints, long computational time, and high computational complexity, especially when the solution space is hard to explore. To overcome these difficulties, some 'more intelligent' rules and /or hybrid techniques such as evolutionary-gradient search (EGS) have been developed to extend EAs to overcome the slow convergence phenomena of the EAs near the optimum solution [24][25][26][27]. In addition, improving fitness function, crossover and mutation operators, selection mechanisms, and adaptive controlling of parameter settings all enhance EA's efficiency and performance.…”