What Is System Identification?

  About Dynamic Systems and Models

  System Identification Requires Measured Data

  Building Models from Data

  Black-Box Modeling

  Grey-Box Modeling

  Evaluating Model Quality

  Learn More

  When to Use the GUI Versus the Command Line

• About This Tutorial

  Data Description

  Loading Data into the MATLAB Workspace

  Opening the System Identification Tool

  Importing Data Arrays into the System Identification Tool

  Plotting and Processing Data

  How to Estimate Linear Models Using Quick Start

  Types of Quick Start Linear Models

  Validating the Quick Start Models

  Strategy for Estimating Accurate Models

  Estimating Possible Model Orders

  Identifying Transfer Function Models

  Identifying State-Space Models

  Identifying ARMAX Input-Output Polynomial Models

  Choosing the Best Model

  Viewing Model Parameter Values

  Viewing Parameter Uncertainties

• About This Tutorial

  Data Description

  Loading Data into the MATLAB Workspace

  Opening the System Identification Tool

  Importing Data Objects into the System Identification Tool

  Plotting and Processing Data

  Estimating a Second-Order Transfer Function Using Default Settings

  Tips for Specifying Known Parameters

  Validating the Model

  Estimating a Second-Order Process Model with Complex Poles

  Validating the Models

  Viewing Model Parameter Values

  Viewing Parameter Uncertainties

  Prerequisites for This Tutorial

  Preparing Input Data

  Building the Simulink Model

  Configuring Blocks and Simulation Parameters

  Running the Simulation

• About This Tutorial

  Data Description

  Loading Data into the MATLAB Workspace

  Plotting the Input/Output Data

  Removing Equilibrium Values from the Data

  Using Objects to Represent Data for System Identification

  Creating iddata Objects

  Plotting the Data in a Data Object

  Selecting a Subset of the Data

  Why Estimate Step- and Frequency-Response Models?

  Estimating the Frequency Response

  Estimating the Empirical Step Response

  Why Estimate Delays?

  Estimating Delays Using the ARX Model Structure

  Estimating Delays Using Alternative Methods

  Why Estimate Model Order?

  Commands for Estimating the Model Order

  Model Order for the First Input-Output Combination

  Model Order for the Second Input-Output Combination

  Specifying the Structure of the Transfer Function

  Validating the Process Model

  Residual Analysis

  Specifying the Structure of the Process Model

  Viewing the Model Structure and Parameter Values

  Specifying Initial Guesses for Time Delays

  Estimating Model Parameters Using procest

  Validating the Process Model

  Estimating a Transfer Function with a Noise Model

  Model Orders for Estimating Polynomial Models

  Estimating a Linear ARX Model

  Estimating a State-Space Model

  Estimating a Box-Jenkins Model

  Comparing Model Output to Measured Output

  Simulating the Model Output

  Predicting the Future Output

• About This Tutorial

  Data Description

  Types of Nonlinear Black-Box Models

  What Is a Nonlinear ARX Model?

  What Is a Hammerstein-Wiener Model?

  Loading Data into the MATLAB Workspace

  Creating iddata Objects

  Starting the System Identification Tool

  Importing Data Objects into the System Identification Tool

  Estimating a Nonlinear ARX Model with Default Settings

  Plotting Nonlinearity Cross-Sections for Nonlinear ARX Models

  Changing the Nonlinear ARX Model Structure

  Selecting a Subset of Regressors in the Nonlinear Block

  Specifying a Previously-Estimated Model with Different Nonlinearity

  Selecting the Best Model

  Estimating Hammerstein-Wiener Models with Default Settings

  Plotting the Nonlinearities and Linear Transfer Function

  Changing the Hammerstein-Wiener Model Input Delay

  Changing the Nonlinearity Estimator in a Hammerstein-Wiener Model

  Selecting the Best Model

  What Is a Frequency-Response Model?

  Data Supported by Frequency-Response Models

  How to Estimate Frequency-Response Models in the GUI

  How to Estimate Frequency-Response Models at the Command Line

  Selecting the Method for Computing Spectral Models

  Controlling Frequency Resolution of Spectral Models

  Spectrum Normalization

  What Is Time-Domain Correlation Analysis?

  Data Supported by Correlation Analysis

  How to Estimate Impulse Response Models Using the GUI

  How to Estimate Impulse Models at the Command Line

  How to Compute Response Values

  How to Identify Delay Using Transient-Response Plots

  Correlation Analysis Algorithm

  What Is a Process Model?

  Data Supported by Process Models

  How to Estimate Process Models Using the GUI

  How to Estimate Process Models at the Command Line

  Process Model Structure Specification

  Estimating Multiple-Input, Multi-Output Process Models

  Disturbance Model Structure for Process Models

  Assigning Estimation Weightings

  Specifying Initial Conditions for Iterative Estimation Algorithms

  What Are Polynomial Models?

  Data Supported by Polynomial Models

  Preliminary Step - Estimating Model Orders and Input Delays

  How to Estimate Polynomial Models in the GUI

  How to Estimate Polynomial Models at the Command Line

  Polynomial Sizes and Orders of Multi-Output Polynomial Models

  Assigning Estimation Weightings

  Specifying Initial States for Iterative Estimation Algorithms

  Polynomial Model Estimation Algorithms

  Example - Estimating Models Using armax

  What Are State-Space Models?

  Data Supported by State-Space Models

  Supported State-Space Parameterizations

  Preliminary Step - Estimating State-Space Model Orders

  How to Estimate State-Space Models in the GUI

  How to Estimate State-Space Models at the Command Line

  How to Estimate Free-Parameterization State-Space Models

  How to Estimate State-Space Models with Canonical Parameterization

  How to Estimate State-Space Models with Structured Parameterization

  How to Estimate the State-Space Equivalent of ARMAX and OE Models

  Assigning Estimation Weightings

  Specifying Initial States for Iterative Estimation Algorithms

  State-Space Model Estimation Algorithms

  What are Transfer Function Models?

  Data Supported by Transfer Function Models

  How to Estimate Transfer Function Models in the GUI

  How to Estimate Transfer Function Models at the Command Line

  Transfer Function Structure Specification

  Specifying Initial Conditions for Iterative Estimation Algorithms

  Estimating Transfer Functions with Delays

  When to Refine Models

  What You Specify to Refine a Model

  How to Refine Linear Parametric Models in the GUI

  How to Refine Linear Parametric Models at the Command Line

  Why Transform Between Continuous and Discrete Time?

  Using the c2d, d2c, and d2d Commands

  Specifying Intersample Behavior

  Effects on the Noise Model

  Choosing a Conversion Method

  Zero-Order Hold

  First-Order Hold

  Impulse-Invariant Mapping

  Tustin Approximation

  Zero-Pole Matching Equivalents

  Estimation Using Band-limited Data

  What Is Subreferencing?

  Limitation on Supported Models

  Subreferencing Specific Measured Channels

  Separation of Measured and Noise Components of Models

  Treating Noise Channels as Measured Inputs

  About Concatenating Models

  Limitation on Supported Models

  Horizontal Concatenation of Model Objects

  Vertical Concatenation of Model Objects

  Concatenating Noise Spectrum Data of idfrd Objects

  See Also