Tolerance Stack-Up Analysis Calculator

Perform 1D Tolerance Stack‑up Analysis for Mechanical Assemblies using Arithmetic, Probabilistic, and Monte Carlo Methods.

What is Tolerance Stack-Up Analysis?

Tolerance stack-up analysis is a fundamental process in mechanical design used to study the accumulation of variation in an assembly. Each component of a product has dimensional tolerances (a permitted variation in its size). This analysis calculates the combined effect of these tolerances to predict the final dimensional variation of the assembly, ensuring that parts fit together correctly and the product functions as intended. A tolerance stack-up analysis allows to assess the effect of geometrical errors before manufacturing. Consequently, any defects can be easily fixed at a very low cost and in a short time.

Brief History

The concept of interchangeable parts dates back to the Industrial Revolution, but the systematic analysis of their accumulation gained momentum in the 20th century, especially with the rise of mass production. The need to predict and control variation became critical. Walter A. Shewhart's work in the 1920s on Statistical Process Control (SPC) laid the groundwork for probabilistic methods. Later, the development and standardization of Geometric Dimensioning and Tolerancing (GD&T) provided a precise language for defining tolerances, making analyses more robust and reliable.

How This Tool Works (Methods)

This tool performs tolerance stack-up analysis using mathematical and statistical methodologies. It automatically determines the best calculation method based on the distributions assigned to the assembly dimensions.

Arithmetic (Worst-Case) Calculates absolute min/max by summing limits. Guarantees 100% fit but can be overly conservative.
Probabilistic (RSS) Used when all parts are Normally distributed. Statistically sums variances for a realistic tolerance spread.
Monte Carlo Simulation Automatically engaged for non-Normal distributions. Simulates thousands of assemblies to predict variation.

Supported Distributions & Manufacturing Relevance

Selecting the correct distribution is crucial for accurate analysis, as it models how a process actually performs in the real world:

  • Normal (Gaussian): Well-controlled processes (e.g., CNC machining). Bell-shaped curve centered on nominal.
  • Homogeneous (Uniform): Equal probability across tolerance. Highly conservative. Used for purchased parts checked with Go/No-Go gauges.
  • Triangular: Mode is most probable, linearly decreasing to limits. Used when operators try to manually center the process.
  • Weibull: Skewed distributions. Used for modeling non-linear variation like wear or grinding processes.
  • Lognormal: Right-skewed distribution. Common in processes with multiplicative errors such as coating thickness.
  • Beta: Highly flexible bounded distribution. Useful for bounded processes like surface finish or percentage-based measurements.
  • Exponential: Right-skewed memoryless distribution. Useful for modeling failure rates or defect spacing.

Dimensions

No dimensions added yet

Click "Add Dimension" below to start building your tolerance stack-up analysis

Stack-up Visualization

Detailed Results

View and export detailed calculation results and history.

Arithmetic Method (Worst-Case)

Nominal value: -

Tolerance: -

Maximum: -

Minimum: -

Probabilistic Method (RSS)

Nominal value: -

Standard deviation: -

Tolerance: -

Maximum: -

Minimum: -

Monte Carlo Results

Nominal value: -

Standard deviation: -

Tolerance: -

Maximum: -

Minimum: -

Mathematical Formulas

Normal Distribution

$$ Z = \sqrt{-2 \ln U} \cdot \cos(2\pi V) $$
$$ X = \mu + \sigma \cdot Z $$

Uniform Distribution

$$ X = A + (B - A) \cdot U $$

where \( U \sim \text{Uniform}(0,1) \)

Triangular Distribution

$$ F_c = \frac{C - A}{B - A} $$
$$ X = \begin{cases} A + \sqrt{U \cdot (B - A) \cdot (C - A)} & \text{if } U < F_c \\ B - \sqrt{(1 - U) \cdot (B - A) \cdot (B - C)} & \text{if } U \geq F_c \end{cases} $$

Weibull Distribution

$$ X = \gamma + \eta \cdot \left[-\ln(1 - U)\right]^{1/\beta} $$

Lognormal Distribution

$$ X = \gamma + \exp(\mu + \sigma \cdot Z) $$

where \( Z \sim N(0,1) \)

Beta Distribution

$$ X = A + (B - A) \cdot Y $$

where \( Y \sim \text{Beta}(\alpha,\beta) \)

Exponential Distribution

$$ X = \gamma - \frac{\ln(1 - U)}{\lambda} $$

Arithmetic Method (Worst-Case)

$$ \text{Nominal} = \sum_{i=1}^{n} \text{sign}_i \cdot \text{nominal}_i $$
$$ T_{\text{total}}^{+} = \sum_{i=1}^{n} T_i^{+} $$
$$ T_{\text{total}}^{-} = \sum_{i=1}^{n} T_i^{-} $$

Probabilistic Method (RSS)

$$ \sigma_{\text{total}} = \sqrt{\sum_{i=1}^{n} \sigma_i^2} $$
$$ T_{\text{total}} = 3 \cdot \sigma_{\text{total}} $$

Monte Carlo Method

$$ S_j = \sum_{i=1}^{n} \text{sign}_i \cdot x_{ij} $$
$$ \mu_{\text{MC}} = \frac{1}{N} \sum_{j=1}^{N} S_j $$
$$ \sigma_{\text{MC}} = \sqrt{\frac{1}{N-1} \sum_{j=1}^{N} (S_j - \mu_{\text{MC}})^2} $$

Symbol Legend

  • U, V: Uniform(0,1) random variables
  • Z: Standard normal random variable
  • X: Random variable for dimension
  • μ, σ: Mean and standard deviation
  • A, B, C: Parameters for distributions (min, max, mode)
  • β, η, γ: Weibull parameters (shape, scale, location)
  • α, β: Beta distribution shape parameters
  • λ: Exponential distribution rate parameter
  • Ti+, Ti-: Plus and minus tolerances for dimension i
  • σi: Standard deviation for dimension i
  • N: Number of Monte Carlo samples
  • Sj: j-th Monte Carlo sample of the stack