Identification in Rotordynamics: Model-Based vs. Direct Measurements

Abstract

In a rotor-bearing system, there are usually some under- or unmodeled components, such as foundations and seals. Identifying the dynamic characteristics of these components often requires both an analytical model and test data due to the working conditions, such as running speed above the first bending mode and non-collocation measurements. The existing methods always identify the dynamic characteristics by solving the equations of motion at discrete frequencies of the measured frequency response functions (FRFs). They have two problems: first, the physical background of the identification is buried in the equation solving process, and second, there is no quality estimation of the identified result. This paper discusses the first problem which is the equation solving process. The second problem, quality estimation, is discussed in a subsequent paper [1]. This paper reveals that model-based identification is the interconnection of certain transfer functions. These transfer functions are either generated from an analytical model (the common model-based method), or directly measured (direct measurement method). The process of both these methods is then illustrated by use of experimental data. A novel seal test design is proposed based on the idea of the direct measurement method. Identification under complex situations is also considered as complementary to the main content, such as different input/output locations. The conditions for identifiability are given.

Model Validation for Damage Identification and Determination of Local Damage Dynamics

“…Finally, d 0 is solved for algebraically as shown in Eq. (1) where the uncertain model is partitioned into M 11 , M 12 , M 21 , and M 22 with the Damaged Data T [7]. Note that d 0 is assumed to be SISO corresponding to particular inputs and outputs of the uncertain model.…”

“…As a result, the virtual controller adopts the dynamics that is missing from the nominal model. Wang et al [6,7] further developed model reconciliation by including uncertainty analysis of the identified system to quantify the uncertainty of the identified unmodeled dynamics. Sawicki and Madden [8] applied model reconciliation with two experimentally tuned models of a rotordynamic system in order to extract the local change in dynamics induced by damage in the shaft of a magnetically levitated experimental rotor rig.…”