Friday, September 20
12:30 p.m. – 1:30 p.m.
2233 GG Brown
The Friday, September 20, Department of Mechanical Engineering Seminar features Brian Feeny of the Michigan State University Department of Mechanical Engineering. He will discuss “Some Problems in Vibration System Identification.” An abstract of his presentation is below.
For more information on future department seminars, visit the seminar Web site.
System identification can be as broadly based as merely determining the number active states needed for a model, to as specific as estimating a parameter in a differential equation. This presentation addresses two such situations. First, it deals with the identification of Coulomb and viscous damping parameters in simple oscillators. Then it centers on the usage of proper orthogonal decomposition (POD) in modeling of multi-degree-of-freedom and distributed parameter vibration systems.
For the damping problem, the interest is in the extraction of Coulomb and viscous friction from single degree of freedom systems, which are otherwise linear. One approach is to compare responses with analytically predicted ones, and establish ways of separating the effects of the two friction sources. We look at unforced and harmonically forced systems.
POD is based on the covariance matrix of several simultaneously measured signals. It yields proper orthogonal modes (POMs), sometimes referred to “energy modes,” as they represent the optimal distribution of signal energy in the measurement space. POD can be used to determine spatial coherence and system dimensionality. The POMs are used as “empirical modes” for Galerkin projections for reduced order modeling. For a limited class of linear and nonlinear vibration systems, the statistically generated POMs are tied to the natural modes.