演化算法中近似模型构造方法及其工程应用

演化算法在求解大型、复杂的工程优化问题时，由于大量耗时的详细分析计算，导致算法的优化效率很低。文中将均匀设计、径向基神经网络技术以及演化算法结合起来，发展了一种“基于均匀设计的逐步修正近似模型构造方法”。利用该方法可以建立目标及约束的近似模型，不仅避免了耗时的详细分析计算，而且提高了演化算法的效率。演化算法采用粒子群优化算法，以六峰值驼背测试函数以及某栽人返回舱气动布局优化设计作为算例，验证了该方法的有效性。

A Less Time-Consuming Method of Evolutionary Algorithm for Complex Engineering Optimization Problems

One essential difficulty in employing evolutionary algorithms in complex engineering optimization tasks is the huge time consumption due to the high complexity of performance analyses and the large number of evaluations needed in the evolutionary optimization. To address this problem, approximation models that are computationally efficient are often used to reduce the computation time. In this paper, we develop an approach, MMAM-UD (Multi-Modification Approximation Model Method Based on Uniform Design), to construct approximation models. There are three reasons for conceiving the approach: (1) population of evolutionary algorithms is initialized by uniform design, which makes individual distributing uniform; then these individuals can reflect more information in optimization region; so, evolutionary algorithms will converge quickly, have good robustness, and approximation models will be more accurate; (2) radial basis neural network can approximate functions very well; (3) in each evolution, the best individual is close to local optimum or global optimum, so the individual can efficiently modify approximation models. Finally, we give two examples: (1) six-hump camel back function, (2) cosmonaut transportation vehicle optimization design. By using the approach proposed in this paper, evolutionary algorithm not only converges correctly but also makes high fidelity analyses only 7.28% for example (1) and 7.82% for example (2) respectively of what was required previously.