一种改善球形目标RAM优化速度的方法

Design of radar absorbing material for spherical target using speedy method

采用遗传算法设计球形目标雷达吸波材料时 ,由于适应度函数是以级数形式给出 ,且随物体尺寸的增大 ,级数收敛减慢 ,优化时间成倍增长。为提高设计效率 ,依物体电尺寸的大小分段使用不同的方法进行优化设计 ;其次 ,采用浮点数编码 ,可以节省计算机内存 ,简化优化过程 ,从而大大缩短优化设计的时间。为避免算法陷入局部最优 ,采用分段变参数遗传算法 ,提高进化的效率。最后 ,用Matlab设计算法 ,通过算例证明方法的快捷、有效。

Genetic algorithm is an efficient optimum method. It has been applied in electromagnetic area such as the design of RAM for canonical targets. But for electrically large spherical targets, it will take quite a long time to get the optimal RAM and become intractable to design those structures. In this paper, we proposed a speedy genetic algorithm. First, in genetic algorithm, we represented the chromosome with floating point number not binary digits. This will not only eliminate the converting between binary and decimal, but also save the compute memory a lot. So, the time to get one generation can be reduced ten to eleven times. We also utilized a special strategy to avoid trapping in premature convergence. Second, the optimum problem was divided into three sections according to the target electrical dimension. Those are low-frequency section?resonant section and high-frequency section. Different sections use different techniques. It's timesaving too. In the end, we fulfilled the algorithm with Matlab and presented some examples. Those examples demonstrated that the methods we proposed are speedy and efficient.