4.2. Measurement Duration Due to their sizes and specific operation conditions, wind turbines cannot be tested dynamically by using standard experimental modal analysis tools. Although there have been some attempts to excite the turbine (at parked condition) by measurable excitation forces (24), (25), as the dimensions of the turbines increased the application of this method became impractical because of the large amplitudes of forces required to excite these structures efficiently. Several researchers also tried to excite a specific turbine mode by applying a harmonic force at its frequency. Simulations and feasibility tests showed that a harmonic force at the desired frequency can be generated by continuously changing blade pitch angles and generator torque or by installing a special rotating mass system on the nacelle. However, the field tests performed on large scale wind turbines showed that it is not possible to achieve the required pitch amplitudes to excite the modes with high modal frequency or high damping ratio due to the limited capacity of electrical pitch actuators. On the other hand, significant energy transfer between the modes having similar frequencies but different damping ratios (especially longitudinal and lateral tower modes) has been observed. The authors concluded that since the excited turbine vibrations are not pure modal vibrations, the estimated damping cannot be considered as the actual modal damping (26),(27). Therefore, dynamic testing of the turbines can only be performed by using the methods utilizing the response measurements only. Several types of these methods are currently in use and proven to provide very valuable information on the dynamic properties of the structures. However, these methods are based on some important assumptions such as time invariant system and stationary random excitation which are not always satisfied easily. Besides, these identification algorithms are mainly based on the correlations established between different response signals (28),(29),(30). Although in theory the calculation of the correlation function is founded on the use of infinite time series, for systems with low modal damping ratios, it was demonstrated to produce satisfactory results even with finite length short time series (30). However, if the system analyzed has high damping ratios, much longer response data should be used to generate correlation functions that enable to detect these high damping modes. The length of the data series is usually expressed in terms of the number of cycles of the lowest frequency included in the measurement block. Since for a rotating wind turbine, most of the modes have relatively high damping ratios (due to aerodynamic coupling), long time series are needed to calculate the required correlation functions. Carne and James, in a very recent article (31) reviewed the history and development of NExT (Natural Excitation Technique - a well known output only modal analysis method) by presenting the outcomes of the landmark tests performed on wind turbines. The results of the tests and analyses summarized in the article show that the appropriate measurement duration is directly related to the expected damping ratios. The authors tested the efficiency of the method by using an analytical model. For this purpose, they generated data series of 82 seconds (40 cycles of the lowest frequency) to compute correlation functions and compared the extracted system parameters with the real known values. Since the damping ratios were relatively low (maximum 0.6 %) such a short duration enabled correlation functions to represent the system response 264
RkJQdWJsaXNoZXIy MTMzNzEzMQ==