Choose a suitable point (x0,y0) so that f(x0,y0) and the partial derivatives are easy to compute.
Determine Δx=x−x0 and Δy=y−y0, requiring ∣Δx∣,∣Δy∣ to be small.
Compute f(x0,y0), fx(x0,y0) and fy(x0,y0).
Substitute into the approximation formula.
The Usual Approximation Formulas
When ∣x∣,∣y∣ are small (taking x0=0,y0=0):
(1+x)m(1+y)n≈1+mx+ny
sin(x+y)≈x+y
ex+y≈1+x+y
ln(1+x+y)≈x+y
Error Estimation
The Absolute Error and the Relative Error
Mathematical formula
A formula is a symbolic expression of quantitative relations and structural patterns, and an essential tool for solving mathematical problems. Key formulas should be mastered and applied flexibly.
The error estimation formulas
Let z=f(x,y), and let the absolute error bounds of x and y be δx and δy respectively, that is, ∣Δx∣≤δx and ∣Δy∣≤δy. Then the
absolute error bound of z is:
δz=∣fx∣δx+∣fy∣δy
relative error bound of z is:
∣z∣δz=ffxδx+ffyδy
An Intuitive Understanding of Error Propagation
The absolute error = the absolute value of each partial derivative × the absolute error of the corresponding independent variable, and then sum.
The relative error = the relative error of each variable multiplied by the “elasticity” of that variable with respect to the function (∂x∂z⋅zx), and then sum.
For the product z=xy: the relative error = the sum of the relative errors, that is, ∣z∣δz=∣x∣δx+∣y∣δy.
For the quotient z=yx: the relative error is likewise the sum of the relative errors.
Worked Examples
Example 1: Approximate Computation
Use the total differential to approximate (1.02)3+(1.97)3.
approximate computationthe total differentialnumerical computation
Approach:
Approximation with the total differential means replacing the increment of the function value by its first-order linear term; the key is to choose a good base point.
Steps:
Put f(x,y)=x3+y3, and take x0=1,y0=2; then Δx=0.02 and Δy=−0.03
Answer: (1.02)3+(1.97)3≈2.95 (the exact value is about 2.951)
Example 2: Error Estimation
A rectangle has length x=5m and width y=3m. The absolute error in measuring both the length and the width is 0.01m. Find the absolute error and the relative error of the area of the rectangle.
error estimationthe absolute errorthe relative errorthe area of a rectangle
Approach:
The absolute error is given by the total differential, and the relative error is then obtained by dividing by the value of the function itself.
Steps:
The area S=xy
Sx=y=3 and Sy=x=5
The absolute error: δS=∣Sx∣δx+∣Sy∣δy=3×0.01+5×0.01=0.08 m2
The area S=5×3=15 m2
The relative error: SδS=150.08≈0.0053=0.53%
Answer: the absolute error is about 0.08 m2, and the relative error is about 0.53%
Example 3: Error Estimation of a Physical Quantity
Use a simple pendulum to measure the gravitational acceleration g=T24π2l. The length l=100cm and the period T=2s have been measured; if the absolute error of l is 0.1cm and that of T is 0.004s, find the absolute error and the relative error of g.
Just as in the previous example, but with the measurement errors of the physical quantities substituted into the formula for the total differential.
Steps:
g=T24π2l
∂l∂g=T24π2 and ∂T∂g=−T38π2l
At l=100,T=2: ∂l∂g=π2 and ∂T∂g=88π2×100=100π2
The absolute error: δg=π2×0.1+100π2×0.004=0.1π2+0.4π2=0.5π2≈4.93 cm/s2
g=44π2×100=100π2≈986.96 cm/s2
The relative error: gδg=100π20.5π2=0.005=0.5%
Answer: the absolute error is about 4.93 cm/s2, and the relative error is about 0.5%
Exercises
Exercise 1
Use the total differential to approximate (1.04)2.02.
approximate computationthe total differentiala power with a variable exponent
Approach:
Once the base point has been chosen, a single linear approximation with the total differential does the job.
Steps:
Put f(x,y)=xy, and take x0=1,y0=2; then Δx=0.04 and Δy=0.02
f(1,2)=1
fx=yxy−1, and fx(1,2)=2
fy=xylnx, and fy(1,2)=0
f(1.04,2.02)≈1+2×0.04+0×0.02=1.08
Answer: (1.04)2.02≈1.08 (the exact value is about 1.0824)
Exercise 2
A cylinder has radius r=5cm and height h=10cm. If the absolute errors of both the radius and the height are 0.1cm, find the absolute error and the relative error of the volume of the cylinder.
error estimationthe volume of a cylinderthe absolute errorthe relative error
Approach:
The error of the volume of a cylinder is handled with the total differential in the same way.
Steps:
The volume V=πr2h
Vr=2πrh and Vh=πr2
At r=5,h=10: Vr=100π and Vh=25π
The absolute error: δV=100π×0.1+25π×0.1=12.5π≈39.27 cm3
V=π×25×10=250π cm3
The relative error: VδV=250π12.5π=0.05=5%
Answer: the absolute error is about 39.27 cm3, and the relative error is about 5%
Summary
Symbols Used in This Article
Symbol
Type
Pronunciation/Explanation
Meaning in This Article
δx
Absolute error bound
delta x
The absolute error bound of x
δz
Absolute error bound
delta z
The absolute error bound of z
Δx
Increment
delta x
The increment of x
x0,y0
Base point
x zero, y zero
The base point used in the approximate computation