Table 1 – FE model characteristics. Fe model # nodes element type # elements Crankshaft 70570 Tetha10 309817 Figure 4 – Crankshaft test rig. COMPARISON BETWEEN EI AND MOGESEC NODES To measure the selection performance in the reduced models, two indexes are taken into account and the master nodes location is compared with respect to the spatial distribution of the model. The first index is the illconditioning parameter of matrices, i.e. the ratio min max O O between the maximum and minimum eigenvalues of both mass and stiffness matrices. The second one is the modal error estimation between experimentally identified and numerical results both on eigenfrequency and mode-shape by means of MAC index. The sensor location for the crankshaft is evaluated in Figure 5, where the selected nodes of MoGeSeC (connected through blue lines) and EI (connected through red lines) are compared. EI computational time is a key point, due to factorial increase of determinant evaluation and iterative nature of the approach, so the greater the number of sensor locations, the bigger, in a non linear manner, computational efforts become. MoGeSeC, instead, evaluated the sensor location in a direct computation and it is not iterative. Globally the MoGeSeC approach demonstrates really good performance with very low computational cost, also comparing with other two selection approaches (Genetic Algorithm and Monte Carlo Algorithm) that are under investigation by the authors. EIGENFREQUENCIES AND MODE-SHAPE COMPARISON LMS TestLab software based on polyreference least-squares complex frequency-domain method is used in both MoGeSeC and EI technique [12]. Table 2 shows the comparison between numerical FE models and identified results. The relative error in eigenfrequencies between actual structure and FE model are minor than 8% for the first 7 modes and the mean values are similar and minor than 4% for both MoGeSeC and EI cases. The first 7 identified and numerical mode shapes are reported for the crankshaft in Figure 7-12 respectively for MoGeSeC and EI approach with respect to FE results: 287
RkJQdWJsaXNoZXIy MTMzNzEzMQ==