SPIRRID

Václav Sadílek
Institute of Structural Mechanics, Faculty of Civil Engineering, Brno University of Technology
e-mail:sadilek.v(at)fce.vutbr.cz

Description

SPIRRID - Tool for estimation of statistical characteristics of multivariate random functions (Python package). The implementation of the SPIRRID is highly configurable and provides the following features:

• The class SPIRRID can be configured for an arbitrary response function q(e = [], theta = []). The function q(e, theta) must be a "callable" object and must have one or more control variables e and one or more parameters theta.
• The parameters and randomization are specified using the tvars trait attribute. They are instances of the RV class representing a random variable that can be associated with probabilistic distribution from scipy.stats.distribution package.
• There are four sampling schemes that can be specified using the sampling_type trait attribute.
• The execution of the integration may be done using the numpy implementation or using a compiled C-code implementation that gets generated on demand for the current response function and randomization scheme.
• The control variable e can be n-dimensional, the range of the input array is specified using the evars parameter of the SPIRRID class. The statistical evaluation is performed for each combination of the entries contained in the range of the control variables.
• The class SPIRRID can also calculate the variance along with the mean value. It can be easily extended with the evaluation of further characteristics like covariance or skewness.
• State dependency between the attributes of the SPIRRID object is maintained automatically: If the input values and the configuration of the SPIRRID have been modified, the results get modified on demand upon the next access to the output values.

Acknowledgment

Development of SPIRRID software was supported by projects FAST-S-14-2443 and by the project CZ.1.07/2.3.00/30.0005 of Brno University of Technology.

Download

SPIRRID software can be downloaded here.