Using EFDD as a Robust Technique for Deterministic Excitation in Operational Modal Analysis

The algorithms used in Operational Modal Analysis assume that the input forces are stochastic in nature. While this is often the case for civil engineering structures, mechanical structures, in contrast, are subject inherently to deterministic forces due to the rotating parts in the machinery. These forces can be seen as harmonic families in the responses, and their influence should if possible be eliminated before extracting the modes in their vicinity.

This paper describes a new and automated method based on the Enhanced Frequency Domain Decomposition technique. First the deterministic signals are identified from kurtosis calculations; secondly performing a linear interpolation across their presence in the identified SDOF functions significantly reduces their influence. No prior knowledge of the deterministic signals is required.

For assessing the quality of the method, various experiments were carried out on a plate structure excited by respectively a pure stochastic signal and the same stochastic signal superimposed by a deterministic signal. Good agreement was found in terms of both natural frequencies, damping ratios and mode shapes. Even the influence of a deterministic signal located exactly at a natural frequency can be practically eliminated.