Introduction & Context

In process engineering, Particle Size Distribution (PSD) analysis is critical for ensuring product consistency and quality. This reference sheet provides a standardized approach to determining the required sample size for statistical quality control. By applying the rigorous problem-solving framework established by Cengel and Cimbala, engineers can transition from qualitative observations to quantitative, statistically significant sampling strategies. This methodology is essential for minimizing analytical costs while maintaining strict adherence to confidence intervals in industrial production environments.

Methodology & Formulas

The determination of the required sample size is governed by the relationship between the population variance, the desired confidence level, and the allowable margin of error. The following steps outline the logical flow for calculating the sample size (n).

1. Statistical Foundation

The fundamental calculation for the raw sample size is derived from the Z-score, the population standard deviation, and the margin of error:

\[ n_{raw} = \left( \frac{Z \cdot \sigma}{E} \right)^2 \]

2. Finite Population Correction (FPC)

When the calculated sample size represents a significant portion of the total batch, the Finite Population Correction factor must be applied to adjust the estimate:

\[ n_{adj} = \frac{n_{raw}}{1 + \frac{n_{raw} - 1}{N}} \]

3. Validity and Threshold Criteria

The following table defines the operational constraints and validity thresholds for the application of these formulas:

Parameter	Condition/Threshold	Action/Requirement
Standard Deviation	\(\sigma \leq 0\)	Invalid input; must be positive.
Margin of Error	\(E \leq 0\)	Invalid input; must be positive.
Confidence Level	\(0.90 \leq \text{Confidence} \leq 0.99\)	Valid range for Z-score selection.
Sensitivity	\(\frac{E}{\sigma} < 0.01\)	Formula unstable; margin of error too small.
Population Size	\(n_{raw} > 0.05 \cdot N\)	Apply Finite Population Correction (FPC).

4. Implementation Logic

To ensure statistical validity, the final sample size n_final is determined by rounding the adjusted sample size to the nearest integer:

\[ n_{final} = \lceil n_{adj} \rceil \]

This ensures that the sampling strategy remains robust against the inherent variability of the production process while adhering to the specified confidence requirements.

To achieve a representative sample for Particle Size Distribution (PSD) analysis, you must minimize segregation and bias during collection. Follow these best practices:

Use mechanical sampling devices like rotary splitters rather than manual scoop sampling to avoid operator bias.
Ensure the sample size is statistically significant relative to the largest particle diameter present in the batch.
Collect multiple increments from different spatial locations within the bulk flow to account for potential stratification.
Maintain consistent moisture levels during sampling to prevent fine particles from adhering to larger ones or container walls.

The sample mass must be sufficient to ensure that the number of particles analyzed is large enough to provide a stable distribution curve. If the mass is too low:

The analysis may suffer from poor reproducibility between runs.
Rare, large particles may be underrepresented, leading to a skewed distribution.
The signal-to-noise ratio in laser diffraction or image analysis systems may decrease, causing measurement instability.

Cohesive powders often form agglomerates that can lead to inaccurate PSD readings. To mitigate this:

Use anti-static measures, such as ionizing bars or grounded equipment, to reduce electrostatic attraction.
Incorporate a dispersion step, such as high-shear mixing or ultrasonic treatment, if the analysis method allows for wet dispersion.
Minimize the time between sampling and analysis to prevent moisture-induced caking.
Store samples in airtight, inert containers to prevent environmental humidity from altering the particle surface chemistry.

Worked Example: Flour Quality Control for PSD Analysis

A flour milling process requires routine quality control of particle size distribution (PSD). The engineering team needs to determine the sample size necessary to estimate the mean particle size (D₅₀) with a specified precision, assuming steady-state production and a normally distributed PSD.

Known Input Parameters:

Population Standard Deviation (σ): 15.0 µm
Desired Margin of Error (E): 5.0 µm
Confidence Level: 95%
Total Batch Population Size (N): 10,000 units

Step-by-Step Calculation:

Identify the Z-score corresponding to the 95% confidence level. From standard statistical tables, Z = 1.96.
Calculate the raw sample size using the formula for estimating a population mean: \( n_{\text{raw}} = \left( \frac{Z \cdot \sigma}{E} \right)^2 \).
Substituting the known values: \( n_{\text{raw}} = \left( \frac{1.96 \times 15.0}{5.0} \right)^2 = 34.574 \).
Check the need for a Finite Population Correction (FPC). The rule applies if \( n_{\text{raw}} > 0.05 \times N \). Here, \( 0.05 \times 10000 = 500 \). Since \( 34.574 < 500 \), no correction is required (FPC factor = 1.0).
Round up the raw sample size to the nearest whole number to ensure the margin of error is not exceeded. Thus, \( n_{\text{final}} = \lceil 34.574 \rceil = 35 \).

Final Answer: The required number of samples to collect is 35 samples.

"Un projet n'est jamais trop grand s'il est bien conçu." — André Citroën

"La difficulté attire l'homme de caractère, car c'est en l'étreignant qu'il se réalise." — Charles de Gaulle