186 E. Papatheou et al. Fig. 17.1 General framework for population based SHM its varying states. Essentially, anything recorded or derived from an experimental test can be used. In order to accomplish the feature exploration in the population based SHM concept, a full modal testing was performed in all of the structures in their normal state, as well as in a ‘damaged’ state—when an added mass was applied. Features are selected by visually scanning through the FRFs and by using the effect of the added mass as a guidance. The suitability of the features is then tested with the help of outlier analysis. The layout of this paper is as follows, the next section briefly describes the structures which were tested and the test setup. The following section describes the results of the comparison of the structures, firstly in the mode shapes and secondly in the FRFs with the help of outlier analysis. Finally the paper is rounded off with some overall conclusions and some discussion about future work. 17.2 Test Structures and Data Acquisition The experimental process was described in detail in [3]. The main points are repeated here for the convenience of the reader. The structures of interest are the horizontal tail wing sections of a Piper PA-28 ‘Arrow’ and of a Piper PA-28 ‘Cherokee’. The two wings were cut in half in order to create two pairs of nominally identical structures. The two new sections originating from the ‘Cherokee’ were named as A1, A2 and the other two sections created from the ‘Arrow’ as B1 and B2. All of the structures can be seen in Figs. 17.2 and 17.3 and a closer look can reveal that the ‘Arrow’ (B1–B2) sections are a subset of the ‘Cherokee’ wing (A1 and A2) as both originate from a variant of the same aircraft, the PA-28. Both wing sections have the same airfoil NACA0012 and the same chord (0.76 m), but different span: tail A is 3.62 m long where tail B is 2.74 m. So, sections B1 and B2 are in theory similar to A1 and A2, but shorter by 0.44 m each. A full modal test was performed for each of the structures. The acquisition system used throughout the tests was a DIFA Scadas III controlled by LMS software running on a DELL desktop PC, and the sensors used were PCB piezoelectric uniaxial accelerometers. In total, there were 65 FRFs recorded for sections B1 and B2, and 75 FRFs for sections A1 and A2. The sensors were rotated in rows as was described in [3]. The locations of the sensors can be seen in Fig. 17.3, where one row of accelerometers is visible. The sensor mapping between the nominally similar structures is identical, but mirrored. Since B1 and B2 are subsets of A1 and A2, there is also an identical subset of sensor locations among all of them with their number being nominally 55, but because of an inconsistency to the test of A2 there are 44 which can be used for comparison. The wings were suspended by springs in order to approximate free-free boundary conditions, and they were excited using a Data Physics (dp) electrodynamic shaker attached directly to their lower surface. A narrowband Gaussian excitation
RkJQdWJsaXNoZXIy MTMzNzEzMQ==