SEM Modal Analysis Technical Division presents

Output Only Modal Parameter Estimation Round Robin

IMAC XXIX 2011 Jacksonville, FL



 Organized By: Dr. Shashank Chauhan, Research Engineer, Bruel & Kjaer Sound and Vibration

Last year at IMAC XXVIII 2010, the Round Robin exercise concerned with estimation of modal parameters in traditional experimental modal analysis framework where information of input excitation forces is also available along with output responses. Typically, the information about excitation forces and corresponding output responses is processed to obtain frequency response functions which are then utilized for modal parameter estimation purposes.

The goal of this year’s round robin is to estimate modal parameters in the framework of Operational Modal Analysis, OMA. Operational modal analysis differs from traditional experimental modal analysis in that it only requires information of the output responses and modal parameters are estimated only the basis of this information.

OMA presents several advantages including the availability of modal parameters of structure in operation thus representing a closer picture of the structure and its boundary conditions (which are not that easy to realize in laboratory conditions). However, lack of input excitation force information presents several challenges as well, e.g. rigorous signal processing is needed to deal with low Signal-to-Noise ratio, proper estimation of damping, estimation of modal parameters in presence of harmonic excitation, etc.

OMA modal parameter estimation algorithms range from simple Single Degree of Freedom (SDOF) peak picking methods to modifications of existing traditional modal parameter estimation algorithms and state space model based algorithms, such as Stochastic Subspace Identification (SSI) algorithm. The purpose of the Output-Only Modal Parameter Estimation Round Robin is to compare modal parameters estimated by various participants (using commercially available and custom made OMA algorithms). It is expected that this study will help in better understanding of OMA algorithms and associated signal processing techniques, and aid in their overall application to various real life structures.

Four different datasets, an analytical example, an experimental structure and two real-life structures, are chosen for this exercise. For consistency, the datasets are presented in terms of raw time histories. The participants are free to process the raw time histories using signal processing techniques of their choice before estimating the modal parameters.

Details of the characteristic information provided by various datasets is as follows

Raw time histories as .mat files (Matlab format) as well as in universal file format,
Geometry information (point identification and 3-D coordinates, connectivity),
Descriptive Information of the system,
Data acquisition conditions and parameters

For comparing the results, participants are requested to submit the following

Natural Frequencies (in Hz), Damping (in % critical),
Unscaled Mode Shapes,
Description of Signal processing used, for e.g.

Time or Frequency domain
Block size, Window used, Number of Averages and Percent Overlap used
Other Techniques like Cyclic Averaging

Modal parameter estimation algorithm used,
Software used,
General comments about the user confidence in the results,
Descriptive Results

The organizers will compare the submitted results using several methods:

Modal Assurance Criteria (MAC) to establish mode correspondence
Statistics of natural frequencies for corresponding mode shapes
Statistics of damping values for corresponding mode shapes

At a special session of IMAC XXIX, participants will have the opportunity to present their work, and the organizers will present a comparison of results.

Several data sets are available on the Michigan Technological University web site ( www.me.mtu.edu/imac_oma). Users are kindly asked to analyze various data sets since they together represent the diversity normally encountered in OMA applications.

The data sets along with their respective description can be downloaded at following links.

Analytical Datasets

Calibration Datasets (General Moderately Damped System)
Calibration Datasets (General Lightly Damped System)
Calibration Datasets (Proportional Heavily Damped System)

Lab based setup

Scaled Model of a Wind Turbine Blade

Real Life Cases Studies

Participants are requested to kindly send their results to Shashank Chauhan at schauhan@bksv.com or Svend Gade at sgade@bksv.com