Systematic Experimental Determination of Discrete-Time Models for Nonlinear Systems
Abstract
Techniques are presented for experimentally computing discrete-time model equations from a finite set of samples observations of the system inputs and outputs. Existing modeling techniques typically consider simple model forms, and often make limiting assumptions and simplifications for mathematical convenience. This research extends these techniques to efficiently obtain a more accurate model equation. Four key points are examined: (1) form of the model equation, (2) choice of the error minimization technique, (3) efficiency of model determination and evaluation algorithms, and (4) interpretation of the obtained model equations in typical applications. A new algorithm for efficient model determination, the Search Indicator Growth Algorithm, is presented. This iterative algorithm efficiently evaluates a set of model terms and eliminates the undesired terms. The technique produces more accurate and robust model equations, and offers significant computational advantages over existing techniques.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1982
- Accession Number
- ADA120533
Entities
People
- Martin Mandelberg
Organizations
- Naval Postgraduate School