• documentation
main content

sensitivity analysis -凯发k8网页登录

analyze cost function sensitivity to model parameters using design of experiments (doe), monte carlo, and correlation techniques

use sensitivity analysis to evaluate how the parameters and states of a simulink® model influence the model output or model design requirements. you can evaluate your model in the sensitivity analyzer, or at the command line. you can speed up the evaluation using parallel computing or fast restart. in the sensitivity analyzer, after performing sensitivity analysis, you can export the analysis results to the parameter estimator or response optimizer apps. to learn more about sensitivity analysis and its applications, see what is sensitivity analysis?

apps

explore design space and determine most influential model parameters

functions

specify parameter space

specify probability distributions for model parameters
specify discrete grids of sample values for model parameters
combine parameter spaces defined for sensitivity analysis
create probability distribution object
truncate probability distribution object
set distribution of parameter in sdo.parameterspace object
add parameter to sdo.parameterspace or sdo.griddedspace object
remove parameter from sdo.parameterspace or sdo.griddedspace object
set grid values of parameters in gridded parameter space

generate samples

generate parameter samples for sensitivity analysis
scatter plot of samples
parameter sampling options for sdo.sample
options for sampling gridded parameters with sdo.sample

create simulation scenario

simulation scenario description

specify time-domain requirements

piecewise-linear amplitude bound
reference signal to track
step response bound on signal
impose elliptic bound on phase plane trajectory of two signals
impose region bound on phase plane trajectory of two signals

specify parameter requirements

impose function matching constraint on variable
impose monotonic constraint on variable
impose relational constraint on pair of variables
impose bounds on gradient magnitude of variable

specify frequency-domain requirements

bode magnitude bound
closed loop peak gain bound
gain and phase margin bounds
nichols response bound
damping ratio bound
natural frequency bound
settling time bound
singular value bound
evaluate cost function for samples
cost function evaluation options for sdo.evaluate
set up steady-state operating point computation
get design variables for optimization
initial state for estimation from simulink model
list of model file and path dependencies
set design variable value in model
analyze how model parameters influence cost function
analysis options for sdo.analyze

topics

evaluation basics

  • what is sensitivity analysis?
    simulink design optimization™ software performs global sensitivity analysis.
  • generate parameter samples for sensitivity analysis

    generate random samples of parameter values or specify a grid of values for sensitivity analysis in the sensitivity analyzer app or at the matlab® command line.

  • analyze relation between parameters and design requirements
    use visual and statistical analysis techniques to analyze the relationship between the parameters and design requirements.

  • validate sensitivity analysis by checking generated parameter values, evaluation results, and analysis results.

  • write a cost function for parameter estimation, response optimization, or sensitivity analysis. the cost function evaluates your design requirements using design variable values.

apps and programmatic workflow

  • identify key parameters for estimation (gui)
    this example shows how to use sensitivity analysis to narrow down the number of parameters that you need to estimate when fitting a model.

  • this example shows how to use sensitivity analysis to narrow down the number of parameters that you need to estimate to fit a model.
  • design exploration using parameter sampling (gui)
    this example shows how to sample and explore a design space using the sensitivity analyzer.

  • this example shows how to sample and explore a design space.
  • explore design reliability using parameter sampling (gui)
    this example shows how to use the sensitivity analyzer to explore the behavior of a pi controller for a dc motor.

  • use sensitivity analysis and response optimization and to evaluate how well a model satisfies design requirements and optimize design variables in the presence of uncertainties in model parameters.

steady-state evaluation


  • an operating point of a dynamic system defines the states and root-level input signals of the model at a specific time.

speed up evaluation

  • use parallel computing for sensitivity analysis
    specify model dependencies and use parallel computing for performing sensitivity analysis in the app, or at the command line.

  • this topic shows how to speed up sensitivity analysis using simulink fast restart.

  • simulink design optimization software supports normal and accelerator simulation modes.

  • how to speed up evaluation in the app by storing intermediate data.

sensitivity analyzer tasks


  • select model parameters for sensitivity analysis in the app.

  • specify time-domain requirements such as signal matching, amplitude bounds, step response bounds, reference signals, elliptical bounds, and custom bounds.

  • specify requirements such as monotonic, smoothness, property, and relational constraints on parameters in your model.

  • specify frequency-domain requirements in the sensitivity analyzer.

  • perform preprocessing operations such as removing offsets and filtering the data before you use it.

  • create linearization input/output sets in the response optimizer or sensitivity analyzer.

  • evaluate your design requirements in the sensitivity analyzer.

  • use the results generated in the sensitivity analyzer to configure parameter estimation or response optimization.

  • plot and interpret parameter set, requirement, result scatter, contour, and tornado plots.

code generation


  • this example shows how to automatically generate a matlab function to solve a sensitivity analysis evaluation problem.

  • this example shows how to automatically generate a matlab function to solve a sensitivity analysis statistics problem.
网站地图