Numerical Derivatives and Statistics – ENGR 216, Comprehensive Exam Review – Study Notes

Source: ENGR 216 Comprehensive Exam Practice Bank | Texas A&M University

Tags: finite differences, forward difference, numerical derivative, Fourier's law, momentum, normal distribution, z-score, probability, standard deviation, ENGR 216


TL;DR

When you have discrete data points instead of a continuous function, finite difference methods approximate derivatives numerically. Forward difference uses the current point and the next one. Separately, the normal distribution is the backbone of engineering statistics: z-scores convert any normal variable into a standard form so you can look up probabilities in a table.


Key Terms

Finite difference method

A numerical technique for approximating derivatives from tabulated data. Three main types: forward, backward, and central.

Forward difference

Uses the current data point and the next one: df/dx ≈ (f(x_{i+1}) − f(x_i)) / (x_{i+1} − x_i). Lower accuracy than central difference, but only needs data ahead of the point.

Backward difference

Uses the current data point and the previous one: df/dx ≈ (f(x_i) − f(x_{i−1})) / (x_i − x_{i−1}).

Central difference

Uses the points on either side: df/dx ≈ (f(x_{i+1}) − f(x_{i−1})) / (x_{i+1} − x_{i−1}). More accurate than forward or backward alone.

Fourier's Law of Heat Conduction

q = k · A · (dT/dx), where k is thermal conductivity, A is cross-sectional area, and dT/dx is the temperature gradient.

Normal distribution (Gaussian distribution)

A symmetric, bell-shaped probability distribution fully described by its mean (μ) and standard deviation (σ). About 68% of values fall within ±1σ, 95% within ±2σ, and 99.7% within ±3σ.

Z-score (standard score)

z = (x − μ) / σ. Converts any normally distributed variable into the standard normal distribution (μ = 0, σ = 1) so you can use a single probability table.


Core Content

Forward Finite Difference: Thermal Conduction

Given data:

  • Point 4: x = 0.04 m, T = 82.6°C

  • Point 5: x = 0.05 m, T = 94.5°C

The forward difference approximation for dT/dx at Point 4:

dT/dx ≈ (T₅ − T₄) / (x₅ − x₄) = (94.5 − 82.6) / (0.05 − 0.04) = 11.9 / 0.01 = 1190 °C/m

Applying Fourier's Law:

q = k · A · (dT/dx) = 0.053 × 0.73 × 1190 ≈ 46.0 kW

Forward Finite Difference: Momentum

A 2.0 kg ball at x = 9.1 cm when t = 0.3 s and x = 15.8 cm when t = 0.4 s.

Velocity at t = 0.3 s (forward difference):

v = (x₂ − x₁) / (t₂ − t₁) = (15.8 − 9.1) / (0.4 − 0.3) = 6.7 / 0.1 = 67 cm/s = 0.67 m/s

Momentum:

p = m · v = 2.0 × 0.67 = 1.34 kg·m/s

Normal Distribution: Setting Up the Problem

For any normal distribution problem, the workflow is:

  • Identify μ, σ, and the boundary value x.

  • Compute z = (x − μ) / σ.

  • Look up the cumulative probability from a z-table or use the complement as needed.

Worked Example: Parachute Safety

μ = 200 m, σ = 30 m. Damage occurs if the parachute opens below 100 m.

z = (100 − 200) / 30 = −100/30 = −3.33

P(Z < −3.33) ≈ 0.0004 (from z-table)

The probability of damage is roughly 0.04%, which is very small but not zero.

Worked Example: Air Filtration

The filter catches 90% of particles, meaning 10% pass through. Particles are normally distributed with μ = 0.5 microns, σ = 0.2 microns.

The filter catches the largest 90%, so the 10% that pass through are the smallest particles. We need the diameter below which 10% of particles fall.

z for P(Z < z) = 0.10 is z ≈ −1.28

x = μ + z · σ = 0.5 + (−1.28)(0.2) = 0.5 − 0.256 = 0.244 microns

The largest particle that passes through the filter is approximately 0.24 microns in diameter.


Formulas

Forward difference: df/dx ≈ (f_{i+1} − f_i) / (x_{i+1} − x_i)

Backward difference: df/dx ≈ (f_i − f_{i−1}) / (x_i − x_{i−1})

Central difference: df/dx ≈ (f_{i+1} − f_{i−1}) / (x_{i+1} − x_{i−1})

Fourier's Law: q = k · A · (dT/dx)

Momentum: p = m · v

Z-score: z = (x − μ) / σ


Why It Matters / Exam Flags

⚠️ The exam will specify which difference method to use (forward, backward, or central). Read the question carefully and do not default to central difference unless instructed.

⚠️ Watch your units in Fourier's Law problems. If k is in kW/(m·°C), your answer will be in kW. Convert if the question asks for watts.

⚠️ For normal distribution problems, draw a quick sketch of the bell curve and shade the region you need. This prevents sign errors and helps you decide whether to use P(Z < z) or 1 − P(Z < z).

⚠️ The air filtration problem requires careful reading: "catches 90%" means the 10th percentile passes through, not the 90th.


Practice Q&A

Q: When would you choose forward difference over central difference?

A: When you only have data at the current point and the next point (no previous point available), or when the problem specifically requires forward difference.

Q: A temperature reading at x = 0.02 m is 60°C and at x = 0.03 m is 71°C. What is the forward-difference estimate of dT/dx at x = 0.02 m?

A: (71 − 60) / (0.03 − 0.02) = 11 / 0.01 = 1100 °C/m.

Q: If a z-score is −2.5, what does that mean physically?

A: The value is 2.5 standard deviations below the mean. In a standard normal table, P(Z < −2.5) ≈ 0.0062, so about 0.62% of values fall below this point.

Q: In the parachute problem, what would the probability of damage be if σ increased to 50 m?

A: z = (100 − 200)/50 = −2.0, giving P(Z < −2.0) ≈ 0.0228, or about 2.3%. A larger standard deviation means less predictable opening altitude and higher risk.


Related Terms / Search Tags

finite difference, forward difference, backward difference, central difference, numerical derivative, Fourier's law, thermal conductivity, heat conduction, temperature gradient, momentum, velocity, normal distribution, Gaussian, bell curve, z-score, standard score, cumulative probability, z-table, percentile, ENGR 216, Texas A&M