The Incremental Approximation and Errors | 增量近似与误差

📚 The Incremental Approximation and Errors | 增量近似与误差

In calculus, when a function undergoes a small change in its input, the resulting change in the output can be approximated using the derivative. This technique, known as incremental approximation or linear approximation, provides a practical way to estimate values and analyse errors in measurements. Understanding how errors propagate is crucial in applied mathematics and the sciences.

在微积分中，当函数的输入发生微小变化时，输出的相应改变可以用导数来近似。这种技巧称为增量近似或线性近似，为估算数值和分析测量误差提供了一种实用的方法。理解误差如何传播在应用数学和科学中至关重要。

1. The Concept of Small Changes | 微小变化的概念

graph of the function is nearly linear, so Δy can be approximated well using the derivative at x.

假设我们有一个可微函数 y = f(x)。如果自变量 x 发生了微小的变化 Δx，那么因变量 y 的改变量为 Δy = f(x+Δx) − f(x)。对于极小的 Δx，函数的图像几乎是线性的，因此 Δy 可以用函数在 x 处的导数很好地近似。

2. Differentials and Their Notation | 微分及其符号

The differential dy is defined as dy = f ' l increment in x. When dx is taken to be the same as the finite change Δx, we write dy = f ' s a linear function of dx and serves as a practical approximation to the actual change Δy.

微分 dy 被定义为 dy = f '(x)dx，其中 dx 是 x 的任意小增量。当 dx 取为有限变化 Δx 时，我们记作 dy = f '(x)Δx。微分 dy 是 dx 的线性函数，可作为实际改变量 Δy 的一个实用近似。

dy = f '(x) dx    and    Δy ≈ dy = f '(x) Δx

dy = f '(x) dx    且    Δy ≈ dy = f '(x) Δx

3. The Linear Approximation Formula | 线性近似公式

ear approximation for the function value near a point a. For x close to a, f(x) ≈ f(a) + f ' ate errors in measurement, and check the sensitivity of outputs to small input variations.

通过重新整理微分关系，我们得到函数在点 a 附近的线性近似。当 x 接近 a 时，f(x) ≈ f(a) + f '(a)(x−a)。这就是在 a 处的切线方程，可以用来估算复杂的函数值、评估测量误差，以及检验输出对微小输入变化的敏感度。

f(x) ≈ f(a) + f '(a)(x − a)

f(x) ≈ f(a) + f '(a)(x − a)

4. Approximating the Increment Δy Using f '(x) | 使用导数近似增量 Δy

e set a = x and replace a+Δx by x+Δx in the formula. The resulting approximation is Δy ≈ f ' he volume of a sphere can be approximated from the error in its radius using dV = 4πr² dr.

要直接估算当 x 从 a 变到 a+Δx 时的 Δy，我们将 a = x，并用 x+Δx 替换 a+Δx。得到的近似为 Δy ≈ f '(x)Δx。这在物理和工程中广泛使用：例如，球体体积的误差可以通过其半径的误差，使用 dV = 4πr² dr 来近似。

5. Understanding the Approximation Error | 理解近似误差

or depends on the curvature of the function and the size of Δx. Mathematically, we can show that the error is of order (Δx)², provided f '' the error diminishes much faster than Δx, making the linear approximation highly reliable.

增量近似带有一定的误差 E = Δy − dy。这个误差取决于函数的弯曲程度和 Δx 的大小。从数学上可以证明，只要 f ''(x) 连续，该误差就是 (Δx)² 数量级的。也就是说，对于足够小的 Δx，误差比 Δx 减小得更快，使得线性近似非常可靠。

E = Δy − dy = ½ f ''(c) (Δx)²   for some c between x and x+Δx

E = Δy − dy = ½ f ''(c) (Δx)²   其中 c 介于 x 和 x+Δx 之间

6. Absolute, Relative and Percentage Errors | 绝对误差、相对误差与百分比误差

In measurement contexts, three types of error are commonly reported. Absolute error is |Δy − dy|, the magnitude of the difference between the true increment and its approximation. Relative error divides this absolute error by the true value |Δy| (or sometimes by the approximate value) and is dimensionless. Percentage error is the relative error multiplied by 100%.

在测量情境中，通常会报告三种误差。绝对误差是 |Δy − dy|，即真实增量与其近似值之差的绝对值。相对误差是将这个绝对误差除以真实值 |Δy|（有时也用近似值），并且没有量纲。百分比误差则是用相对误差乘以 100%。

广告 / Sponsored

Error Type	Definition
Absolute error	|Δy − dy|
相对误差	绝对误差 / |Δy|
Percentage error	相对误差 × 100%

7. Propagation of Errors Through a Function | 误差通过函数的传播

g the derivative. We treat δx as a typical magnitude of error in x, and then the propagated absolute error in y is approximately |f '(x)| ived from differentiation but are often summarised in standard error-propagation formulas.

如果输入 x 的测量有一个小的不确定度 δx，那么输出 y = f(x) 的不确定度可以用导数来估计。我们将 δx 视为 x 误差的典型大小，那么 y 的传播绝对误差大约为 |f '(x)| δx。对于和、差、积、商，相应的规则可以从微分推导出来，但通常总结为标准误差传播公式。

δy ≈ |f '(x)| δx

δy ≈ |f '(x)| δx

This approach tells us that the relative error in y is |f '(x) / f(x)| δx, highlighting which measurements must be made most precisely to minimise overall error.

这种方法告诉我们，y 的相对误差是 |f '(x) / f(x)| δx，从而凸显出哪些测量必须最精确地完成，以使得总误差最小。

8. Worked Example – Estimating a Square Root | 示例 – 估算平方根

Let us estimate √17 using f(x) = √x, with a point a = 16 where the value is easy. We have f '(x) = 1/(2√x), so f '(16) = 1/8. Set Δx = 1. Then √17 ≈ f(16) + f ' roughly 0.0019. This rapid convergence illustrates the power of incremental approximation.

让我们用 f(x) = √x 和 a = 16 这个容易求值的点来估算 √17。f '(x) = 1/(2√x)，因此 f '(16) = 1/8。取 Δx = 1，则 √17 ≈ f(16) + f '(16)×1 = 4 + 1/8 = 4.125。真值大约为 4.1231，绝对误差只有大约 0.0019。这种快速收敛展示了增量近似的威力。

√17 ≈ 4.125   (true ≈ 4.1231056)

√17 ≈ 4.125   (真实值 ≈ 4.1231056)

If we had used a smaller step, say Δx = 0.5 with a = 16.5, we would need f (16.5)=√16.5, which is not simpler. Choosing a convenient approximation point is key.

如果我们用更小的步长，比如 a = 16.5、Δx = 0.5，就需要计算 f(16.5)=√16.5，这并不简单。选择一个方便的近似点是关键。

9. Common Pitfalls and How to Avoid Them | 常见陷阱及如何避免

One common mistake is to forget that the approximation Δy ≈ f ' en approximating compound functions, one must correctly apply the chain rule. Finally, many students confuse dx with Δx; in the approximation context they are equal numerically, but conceptually dx is an infinitesimal while Δx is a finite difference.

一个常见错误是忘记近似 Δy ≈ f '(x)Δx 只在 Δx 很小时才成立。使用过大的 Δx 会导致显著误差。另一个陷阱是错误识别自变量：在近似复合函数时，必须正确地使用链式法则。最后，很多学生把 dx 和 Δx 混淆；在近似语境下它们数值相等，但在概念上 dx 是无穷小，而 Δx 是有限差值。

Also, when computing percentage errors, ensure the denominator is the true value (if known) or the best estimate, and always state which is used. In error propagation, never add absolute errors of independent measurements directly without considering whether the function is a sum or a product.

在计算百分比误差时，要确保分母是真实值（如果已知）或最佳估计值，并始终说明使用的是哪一个。在误差传播中，不要在没有考虑函数是求和还是求积的情况下，直接将独立测量的绝对误差相加。

10. Summary and Key Takeaways | 总结与要点

The incremental approximation y(x+Δx) ≈ y(x) + y ' r is proportional to (Δx)², ensuring high accuracy for small increments. Mastering this topic equips students to handle sensitive error analyses in physics, economics, and other fields where small variances matter.

增量近似 y(x+Δx) ≈ y(x) + y '(x)Δx 是导数定义的直接推论。它能够快速估算函数值和测量误差。近似误差与 (Δx)² 成正比，确保了微小增量下的高精度。掌握这一专题能让学生有能力处理物理、经济学以及其他微小差异都很重要的领域中的灵敏误差分析。

Published by TutorHao | Mathematics Revision Series | aleveler.com

更多咨询请联系16621398022（同微信）