FEASIBILITY OF IDENTIFYING NON-LINEAR VIBRATORY SYSTEMS CONSISTING OF UNKNOWN POLYNOMIAL FORMS

✍ Scribed by C.M. Richards; R. Singh

• Publisher: Elsevier Science
• Year: 1999
• Tongue: English
• Weight: 440 KB
• Volume: 220
• Category: Article
• ISSN: 0022-460X
• DOI: 10.1006/jsvi.1998.1918

✦ Synopsis

System identification techniques for non-linear systems may require a priori knowledge of the nature and mathematical form of the non-linearities. However, for practical systems, this is not always possible. As a result, non-linearities are often approximated and questions remain as to whether a reasonably accurate model can be determined. Concurrently, under experimental conditions, some means of quantifying the amount of measurement noise present in the identification process must also be obtained. To resolve such issues, a discrete non-linear system problem is formulated in the presence of uncorrelated noise and critically examined from the standpoint of identification. Coherence functions are introduced which are based on a ''reverse path'' spectral approach recently developed by the authors for multi-degree-of-freedom systems. These coherence functions, as calculated from conditioned spectra, indicate the extent of uncorrelated noise present and the accuracy of assumed mathematical models employed for describing non-linear systems. Using several example simulation systems, including a system with a continuous non-linearity described by a non-integer exponent, both temporal and spectral identification techniques are employed to study the issues described above.