Search
Close this search box.

Nodeworks 19.1 Release Announcement

We are pleased to announce the release of Nodeworks 19.1! This release includes a complete over-haul of the Optimization tool-set and the introduction of a new Uncertainty Quantification tool-set, which are offered together under a single Surrogate Modeling and Analysis menu option. This new tool-set includes many new features that enable not only optimization workflows, but also uncertainty quantification workflows like forward propagation of input uncertainties and sensitivity analysis. New MFiX features have been also been added, including Monitor readers.

Please use the new forum for questions and bug reports: https://mfix.netl.doe.gov/forum/c/nodeworks

Notable changes include:

  • Design of Experiments (DOE)
    • Support geometry parametrization inside MFiX
    • Added genetically optimized latin-hypercube
    • Correctly implemented central composite
    • Include/exclude samples
    • Import/export samples
  • Response Surface Methods (RSM)
    • Supports multiple matrix/response inputs
    • 11 response surface methods available
      • from scikits-learn library:
        • polynomial (with 16 different regressors)
        • Gaussian process model (GPM)
        • multilayer perceptron
        • support vector
        • desicion tree
        • random forest
      • from scipy library:
        • radial basis function (RBF)
        • nearest
        • cubic
        • linear
      • from pyearth library:
        • multivariate adaptive regression splines (MARS)
    • Support fitting multiple models at the same time
    • Assess the quality of the surrogate models constructed through various statistical error metrics calculated and user selectable cross-validation process.
    • Offer the user the selection of which response surface model to be used in the rest of the workflow
  • Optimization
    • Exposed 10 optimization methods in scipy library instead of hard-coding them, including differential evolution
    • "optimize" multiple times with random guesses inside the model space
    • Export results
  • Sensitivity Analysis (SA)
    • Exposed all 5 sensitivity analysis methods implemented in SALib
  • Forward propagation of input uncertainties
    • Handles 7 different type probability density functions (PDF) and user-prescribed distributions for aleatory uncertainties
    • Use any of the DOE methods for epistemic uncertainty
    • Exports the bounds to a file
    • Verified against Ishigami example.
  • Wizard
    • Automatically add and connect nodes for common use cases including
      • Sensitivity analysis
      • Forward propagation of uncertainty
      • Optimization
    • Includes 17 test functions
Share
Share
Share
Share