The coefficient of variation should be available within a reliability benchmark problem.

[mbaudin47]

The difficulty is that not all algorithms can provide this output.

This can be provided by two different implementations.

• The CV is a new attribute of ReliabilityBenchmarkResult. It value is set to 0.0 by default, but is set to a nonzero value by the algorithms which can estimate it.
• A new class ReliabilityBenchmarkResultWithCOV is created. This is a children class to ReliabilityBenchmarkResult, with a new CV attribute.

@m-balesdent : do you have any comment on this enhancement proposal? (is https://github.com/brevault Loïc's account ?)

[m-balesdent]

Hello,

It could be also nice to provide the user with an indicator to analyze the simulation budget taken by the different algorithms to compute a failure probability with a given CV.

Indeed, another valuable attribute to compare two simulation algorithms is to analyze their capacity to reduce the budget needed by Crude Monte Carlo to estimate the failure probability for the same CV.

It is mentioned at page 88 (Eq. 5.14) of V. Chabridon's PhD thesis or at page 37 of [Morio & Balesdent - Estimation of Rare Event Probabilities in Complex Aerospace and Other Systems : A Practical Approach, Elsevier, 2016]

PS: Loïc's account is : https://github.com/l-brevault