多磁源目标分辨的优化方法探讨

Discussion of Optimization Method on Multi-Magnetic Source Discrimination

在多磁源目标分辨研究中,解析非线性方程成为一个难题,优化方法可以有效解决这一问题。利用高斯-牛顿算法和遗传算法对多磁源目标分辨进行了求解,并全面分析对比了两种算法的特点和规律。结果表明,合理的初始值可以加快两者解的收敛。两者计算误差都随磁源数目增加呈增加趋势。此外,当磁源数目在4个以内时,高斯-牛顿法计算效率高,遗传算法较耗时;磁源数目多于4个时,高斯-牛顿法失效,遗传算法计算结果可靠。研究可为多磁源目标分辨工作提供有力支撑。

In the research on multi-magnetic source discrimination, the non-linear equation is difficult to analyze, optimization methods are available to solve the problem. In this paper, Gauss-Newton Algorithm and Genetic Algorithm were used to review the multi-magnetic source inversion. Through contrast and analysis, the results show that reasonable initial iteration values can accelerate the convergence. The error of two methods inclines to increase with the dipole number increased. Furthermore, Gauss-Newton Algorithm has high efficiency while Genetic Algorithm costs more times in the case that the dipole number is lower than four. As the number of dipole is increasing, Gauss-Newton Algorithm fails while Genetic Algorithm is still effective. The research can provide a strong support for the multi-magnetic source resolution.