Output-only Modal Analysis (OMA) has found extensive use in the identification of dynamic properties of structures. This study aims to investigate the effect of excitation force on the accuracy of modal parameters. For this purpose, the modal parameters of a simply supported beam are obtained through the Experimental Modal Analysis (EMA) and the OMA method using three different types of artificial and natural excitations, namely a shaker, acoustic waves, and environmental noise. Frequency

more »

... Decomposition (FDD) technique is used to identify dynamic characteristics. Finally, these results are compared with those obtained by the analytical method and the EMA method. The results demonstrated the following: 1) Acoustic excitation presents the natural frequencies with the smallest errors in comparison with the analytical results. 2) Inaccuracy is observed at certain natural frequencies during the excitation with a shaker with respect to the connecting point between the shaker and the beam. 3) Modal Assurance Criterion (MAC) showed that the mode shapes extracted by the acoustic excitations are more similar to the analytical results. Experimental Study on the Effect of Excitation Type on the Output-Only Modal Analysis Results In the first method, known as Experimental Modal Analysis (EMA), the system is affected by a specific and controllable excitation. In order to achieve the dynamic characteristics of a system using this method, the measured response and excitation data of the test are processed in the time or the frequency domain based on different methods such as the Least Squares Complex Exponential (LSCE) method [4], the Ibrahim Time Domain method [5, 6] , and the Rational Fraction Polynomial (RFP) method [7] . The second method, called Output-only Modal Analysis (OMA), is a relatively new and improved approach which has recently been advanced. In this method, the system can be excited by various types of forces to be analysed, without the necessity to measure theses forces. This advantage of the method makes this method superior to the EMA method. According to the source of excitation, a system can be excited either artificially or naturally. In artificial excitation, the system can be affected by a specific excitation, just like in the EMA method, and modal parameters can then be approximated by the measured responses. However, in natural excitation, the vibration is investigated in the operating condition, such as measuring the vibrations of a turbine in operation or vibrations in the body or other parts of a vehicle moving on a rough road. In this type of excitation, the system can be affected by different unknown sources of excitation, and the approach is therefore called Operational Modal Analysis. The advantages of operational modal analysis are as follows: 1) Huge structures and constructions, such as bridges and buildings, which cannot easily be excited by artificial and controllable forces, can be excited by forces from natural sources such as wind, acoustic waves, and traffic loads. In mechanical systems such as turbines and vehicles, exciting forces can also be caused by aerodynamic and engine loads, as well as by road and rail traffic loads. 2) The operational modal analysis would achieve modal parameters during work. Even if the system is placed in laboratory conditions, the workplace excitation cannot be exactly simulated. One of the main disadvantages of the OMA method is the existence of high noise which can significantly affect the quality of the results. Estimation of modal parameters via OMA can be accomplished in the time or the frequency domain just like in EMA; different methods such as Natural Excitation Technique [8, 9] , Stochastic Sub-space Identification (SSI) [10], and Frequency Domain Decomposition (FDD) [11, 12] can be used for this purpose. The OMA methods in the frequency domain are defined based on the relation between the input and output power spectral densities (PSD) [13] . The peak picking technique is the most efficient method in the frequency domain; in this method, the natural frequencies are obtained directly by acquiring data from the peaks of the PSD diagram [14] . If the system contains separated modes, peak picking is an acceptable method for determining its modal parameters [15] . Brincker et al. offered the FDD method to identify the modes of a system, especially in a noisy condition or modes that are close to each other [11] . In this method, the singular values of PSD at different frequencies are used to determine the modal parameters. The Enhanced Frequency Domain Decomposition (EFDD) has also been defined to approximate the damping ratio [16, 17] . In recent years, some researches have been carried out using the OMA method and the FDD technique for extracting modal parameters. Zhang et al. made some changes to the FDD theory and described the method of Frequency Spatial Domain Decomposition (FSDD) [18]. Magalhães et al. compared modal damping ratios which were approximated based on free vibration and ambient tests [19]; the SSI and FDD techniques have been used to identify modal damping ratios in ambient tests. Yan et al. proposed the Power Spectrum Density Transmissibility (PSDT) for other excitations beside the white noise [20]. Araújo et al. implemented Singular Value Decomposition (SVD) on PSDT to reduce extra non-physical