4 OMA of a High-Rise TV Tower Using the Novel Poly-reference Complex. . . 45 Fig . 4. 2 Sensors’ locations and directions: measured DOFs (a ) and position of each dataset (b). The blue arrows indicate the reference sensors, whereas the red ones denote the moving sensors thirty minutes due to the very low fundamental frequency suited around 0.2 Hz. The acquisition of such responses was done with a sampling rate, . fs, of 10Hz. Due to the dimensions of the tower, two independent (un-synchronized) acquisition systems were used to record the vibration responses: one used as reference and the other as roving system, being both comprising two 3D nodes with each node consisting of three geophone sensors. The sensors’ locations and directions of each dataset are shown in Fig. 4.2. Figure 4.2a shows all Degree s O f Freedom s (DOFs) measured in the test campaign, whereas Fig. 4.2b indicates the locations and measurement directions of the sensors in each measured dataset, with the blue arrows designating the reference and red arrows the moving sensors. During the testing campaign, a total of 6 datasets were measured along the height of the tower, being the 1st five measured at the tower itself, whereas the sixth and last dataset was measured at the bottom part of the antenna, as indicated in Fig. 4.2b. Since the reference and roving acceleration responses were acquired with two independent systems, the acceleration time series recorded with the roving system had to be synchronized with regard to those of the reference system. The synchronization of each dataset was accomplished by, first, calculating the average phase angle of the transfer function between the 1st reference and 1st roving channel and subsequently by removing the time shift corresponding this phase from all the roving channels. Afterward, the synchronized times series were de-trended and decimated with a factor of 4, leading to time series with a sampling frequency of 2.5 Hz. Once synchronized, filtered, de-trended, and down-sampled, the time series of each dataset was used to compute the HS matrix with a resolution of 513 frequency lines. Afterward, the HS matrix of each dataset was prescaled and stacked on the top of each other to form a global spectrum following the approach described in [11, 10]. Finally, the formed global HS matrix was used as primary data by the pCF-MM technique from which the natural frequencies, damping rations, and operational factor vectors were extracted with the aid of a stabilization diagram. Figure 4.3 shows the stabilization diagram constructed with pCF-MM by identifying models with order ranging from 1 to 70. The vertical red lines in such diagram indicate the natural frequencies of the physical vibration modes automatically identified by means of a Hierarchical Cluster Algorithm [10]. It is worth highlighting that the peak suited to around 0.7 Hz corresponds to poorly excited modes and, therefore, could not be clearly identified as physical. The details of the closed–spaced modes around 0.2, 0.5, and 0.82 Hz are shown in Fig. 4.3b, c, and d, respectively.
RkJQdWJsaXNoZXIy MTMzNzEzMQ==