[论文解读] AN OVERVIEW OF NUMERICAL AND ANALYTICAL METHODS FOR SOLVING ORDINARY DIFFERENTIAL EQUATIONS

Byakatonda Denis|arXiv (Cornell University)|Jan 1, 2020

Numerical methods for differential equations参考文献 18被引用 3

一句话总结

本文全面概述了求解常微分方程（ODEs）的解析方法与数值方法，对比了在 MATLAB 中使用欧拉法与四阶龙格-库塔法的性能。结果表明，龙格-库塔法在较小步长下显著优于欧拉法，具有更高的精度，并建议未来研究 ODE 及刚性系统时采用先进的数值求解器和开源工具。

ABSTRACT

Differential Equations are among the most important Mathematical tools used in creating models in the science, engineering, economics, mathematics, physics, aeronautics, astronomy, dynamics, biology, chemistry, medicine, environmental sciences, social sciences, banking and many other areas [7]. A differential equation that has only one independent variable is called an Ordinary Differential Equation (ODE), and all derivatives in it are taken with respect to that variable. Most often, the variable is time, t; although, I will use x in this paper as the independent variable. The differential equation where the unknown function depends on two or more variables is referred to as Partial Differential Equations (PDE). Ordinary differential equations can be solved by a variety of methods, analytical and numerical. Although there are many analytic methods for finding the solution of differential equations, there exist quite a number of differential equations that cannot be solved analytically [8]. This means that the solution cannot be expressed as the sum of a finite number of elementary functions (polynomials, exponentials, trigonometric, and hyperbolic functions). For simple differential equations, it is possible to find closed form solutions [9]. But many differential equations arising in applications are so complicated that it is sometimes impractical to have solution formulas; or at least if a solution formula is available, it may involve integrals that can be calculated only by using a numerical quadrature formula. In either case, numerical methods provide a powerful alternative tool for solving the differential equations under the prescribed initial condition or conditions [9]. In this paper, I present the basic and commonly used numerical and analytical methods of solving ordinary differential equations.

研究动机与目标

• 调查并对比求解常微分方程的解析方法与数值方法。
• 评估欧拉法与四阶龙格-库塔法的精度与收敛性。
• 评估步长对数值解精度的影响。
• 推荐求解 ODE 的计算工具与未来研究方向，特别是对刚性方程与系统。

提出的方法

• 本研究采用解析技术求解具有常系数与变系数的一阶与二阶线性 ODE。
• 通过不同步长（h = 0.1, 0.05, 0.01）的欧拉法与四阶龙格-库塔法计算数值解。
• 使用 MATLAB 实现并模拟两种数值算法，进行对比分析。
• 通过将数值结果与精确的解析解对比，进行误差分析。
• 生成图形对比，可视化不同步长下的收敛性与精度。
• 通过离散点上的绝对误差度量评估数值方法的性能。

实验结果

研究问题

• RQ1 当使用相同步长求解同一 ODE 时，欧拉法与四阶龙格-库塔法在精度上如何比较？
• RQ2 减小步长在多大程度上能提高 ODE 数值解的精度？
• RQ3 随着步长趋近于零，数值解如何趋近于精确解析解？
• RQ4 对于初值问题，欧拉法与龙格-库塔法中哪种方法收敛更快且误差更小？
• RQ5 在解析解不可用时，求解 ODE 的最有效计算工具与策略是什么？

主要发现

• 在相同步长下，四阶龙格-库塔法产生的误差显著小于欧拉法，当 h = 0.01 时误差低至 1.69 × 10⁻⁸。
• 当 h = 0.01 时，龙格-库塔法的最大误差为 1.69 × 10⁻⁸，而欧拉法在 x = 0.1 处的最大误差为 0.010196522。
• 更小的步长始终能降低数值误差，且随着 h → 0，两种方法均表现出向精确解收敛的特性。
• 当 h < 0.05 时，龙格-库塔法的解曲线与精确解几乎无法区分，表明其具有更优的稳定性和精度。
• 即使在小步长下，欧拉法仍表现出较大误差，凸显其在实际应用中的局限性。
• 本研究证实，龙格-库塔法在求解 ODE 初值问题方面比欧拉法更有效、更高效。