Frequency response sensitivity analysis to determine sensor placement for vibration-based damage detection in structural elements
Abstract
In structural health monitoring and numerous other applications, it is desirable to carefully select sensor locations, thereby minimizing the number of sensors and the cost associated with an on-board monitoring system. The best sensor locations in the present context are those with a high sensitivity to the damage of interest, robustness in the presence of changes in the excitation frequency and structural natural frequencies, and potentially other metrics as well. Sensor locations are commonly optimized using iterative genetic algorithms, a process which becomes lengthy. In this dissertation, a new method of sensor selection is introduced, using sensitivity functions that are calculated using frequency response function data from the healthy structure to maximize the changes in those frequency response functions due to damage. A modal description of the embedded sensitivity functions is developed, and the functions are applied to a lumped parameter system to investigate the additional insight gained from the use of this modal form of the sensitivity functions. The embedded sensitivity function is then applied to two finite element models (a rod with translational motion and a finite element beam with translational and rotational motion) and a test specimen (an aluminum I-beam with translational, rotational, and torsional motion). Embedded sensitivity functions are shown to be effective at identifying locations for sensors on structural beams that maximize the changes in measured frequency response functions within a range of forcing frequencies. Limitations of embedded sensitivity analysis are presented and discussed. Modal parameters of an aluminum I-beam are estimated, and are used to further investigate the modal form of the sensitivity function. From the modal form, modes of vibration most affected by damage in a specified location are identified. It is found that the sensitivity function combines modal vectors in a way that includes effects which are not visually evident.
Degree
Ph.D.
Advisors
Adams, Purdue University.
Subject Area
Mechanical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.