Adaptive Identification of Fluid-Dynamic Systems
Abstract
Fluid-dynamic systems are inherently nonlinear and are subject to a combination of coherent and random unsteady disturbances. As a result, accurate low-order dynamic models are difficult to obtain for real-time control of such systems. Therefore, controllers implementing adaptive on-line system identification are ideally suited to flow control problems. Adaptive linear and nonlinear filters for real-time system identification are presented in this paper. The linear models studied are traditional FIR and IIR filters, and the nonlinear models include a 2nd-order Volterra filter and the Bilinear filter. The coefficients of the adaptive filter models are calculated and updated using two of the most popular recursive methods, the normalized LMS and RLS algorithms. The adaptive filters are tested offline in software and then implemented on real-time DSP hardware. The focus of this study is on model accuracy and viability in real-time applications. The real-time performance is measured in terms of achievable sampling frequency. Specific applications to relevant nonlinear systems, a spring-mass damper model and a drag-law problem, are also considered in detail.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 14, 2001
- Accession Number
- ADA466403
Entities
People
- Aravind Pillarisetti
- Louis N. Cattafesta Iii
Organizations
- University of Florida