Calculate linear tolerance stacks using worst case, RSS, or reproducible Monte Carlo simulation. ISO 286 tolerance classes can be resolved by the validated fits core.
Define dimensions and run the analysis.
Combine signed dimensions and compare worst case, RSS and Monte Carlo results.
Each row needs a name, nominal value, direction (+/−), and either upper/lower deviations or an ISO 286 field. RSS needs independent scatter; Monte Carlo also needs a distribution, sample count and random seed.
The sign determines whether a dimension increases or decreases the functional size. A normal distribution is meaningful for statistical methods only when the process is reasonably symmetric and stable.
Worst case adds the unfavorable limits. RSS takes the square root of the summed variances. Monte Carlo samples the selected distributions and counts limit violations.
100 + 30 − 20 mm gives a 110 mm nominal. With ±0.10, ±0.05 and ±0.05 mm, worst case is ±0.20 mm; independent normal distributions give about ±0.122 mm for 1σ.
Sources and limits: General tolerance and probability theory; ISO 286 fields use the validated ISO 286 core. Correlation and process drift are not modeled.
Combine signed dimensions and compare worst case, RSS and Monte Carlo results.
A tolerance defines the permitted size range of a manufactured feature. Its position locates that range around the zero line, while the IT grade sets its width.
Worst case adds the unfavorable limits. RSS takes the square root of the summed variances. Monte Carlo samples the selected distributions and counts limit violations.
Each row needs a name, nominal value, direction (+/−), and either upper/lower deviations or an ISO 286 field. RSS needs independent scatter; Monte Carlo also needs a distribution, sample count and random seed. The sign determines whether a dimension increases or decreases the functional size. A normal distribution is meaningful for statistical methods only when the process is reasonably symmetric and stable.
Enter each dimension in the chain with a name, nominal value, direction (+/−) and deviations or an ISO 286 field. Then choose the evaluation method: worst case for the least favorable combination, RSS for statistically independent scatter, or Monte Carlo for a full distribution simulation.
100 + 30 − 20 mm gives a 110 mm nominal. With ±0.10, ±0.05 and ±0.05 mm, worst case is ±0.20 mm; independent normal distributions give about ±0.122 mm for 1σ.
The sign determines whether a dimension increases or decreases the functional size. A normal distribution is meaningful for statistical methods only when the process is reasonably symmetric and stable. Positive clearance means the hole remains larger than the shaft in the least favorable limits. A negative result is interference and needs separate assembly, temperature and joining checks.
Dimensions in millimetres, result as nominal value with upper/lower limit deviation.
Checking whether a multi-link dimension chain meets a required final size with sufficient margin, for example for stacked parts or an assembly with several toleranced interfaces.
General tolerance and probability theory; ISO 286 fields use the validated ISO 286 core. Correlation and process drift are not modeled.
Common mistake: Don't assign chain-member signs by feel – each dimension must be entered as + or − consistently with the chosen reference direction, or the chain gives a wrong nominal sum. RSS assumes statistically independent, mostly normally distributed individual tolerances; with correlated or systematic deviations it overestimates the achievable accuracy.
Worst case suits safety-critical work or small quantities with few chain links, since it covers the absolute least favorable combination; with many independent tolerances it is often unnecessarily conservative.
RSS takes the square root of the summed variances instead of the plain sum of magnitudes; statistically it is unlikely all dimensions hit their least favorable limit at once.
It allows arbitrary, including non-normal, distributions per chain link and directly estimates the share of assemblies that would violate a specification limit – useful for asymmetric or known real-world scatter.
Every dimension is combined at its least favorable limit. The method requires no distribution assumption.
RSS assumes independent dimensions. Monte Carlo models the selected distributions; correlations, process drift, and measurement uncertainty are explicitly excluded from version 1.0.