Skip to content

Latest commit

 

History

History
46 lines (31 loc) · 2.42 KB

UQLab_UMbridge.md

File metadata and controls

46 lines (31 loc) · 2.42 KB

UQLab client

UQLab is a general purpose Uncertainty Quantification framework based on Matlab. It is made of open-source scientific modules which are smoothly connected to perform uncertainty quantification through Monte Carlo simulation, sensitivity analysis, reliability analysis (computation of rare event probabilities), surrogate models (polynomial chaos expansions, Kriging, etc.), Bayesian inversion/calibration and many more.

UQLab fully supports UM-Bridge models. After registering at https://www.uqlab.com/, downloading and installing UQLab as indicated on the UQLab website, you can start UQLab using

uqlab

To run an UM-Bridge model, install Docker Desktop and start a docker model, e.g., the Euler-Bernoulli beam model, using the terminal command

>> docker run -it -p 4243:4243 linusseelinger/benchmark-muq-beam-propagation:latest

Add UM-Bridge to your Matlab path and create an UM-Bridge model as in the section "Matlab client" above:

addpath(fullfile('\YOUR\FULL\PATH\TO\UMBRIDGE\INSTALLATION', 'matlab')) % modify with the correct path
model = HTTPModel('http://localhost:4243', 'forward');

The following lines create the corresponding UQLab model object:

ModelOpts.mHandle = @(X) model.evaluate(X);
EulerBernoulliBeamModel = uq_createModel(ModelOpts);

Additional parameters can also be passed to the model, see the UQLab User Manual: The Model Module, Section 2.1.4.

The model can now be evaluated for one or more sets of input parameters. For example, the three-dimensional Euler-Bernoulli beam model with parameter space [1.00 ,1.05]^3 can be evaluated at two parameter sets as follows:

X = [1.00, 1.00, 1.01; ...
     1.04, 1.01, 1.05];
Y = uq_evalModel(EulerBernoulliBeamModel, X);

Each row of X and Y correspond to one set of input parameters. In general, let N denote the number of sample sets, and let N_in and N_out denote the input and output dimension of the model, respectively. Then X has dimension N x N_in, and Y has dimension N x N_out.

The UQLab model object can be used further for all types of UQ analyses, as explained in detail in the UQLab documentation.