164 S. S. Christensen et al. Fig. 15.1 Stabilization diagram overlaid by its probability mass function based on modal parameter estimates derived using the PLSCF algorithm on simulated data. circle – bins with pole estimates that are below the horizontal dashed black line and outside the white patched area; square – poor MAC valued pole estimates; triangle – stable pole estimates with similar MAC values Fig. 15.2 Stabilization diagram overlaid by its probability mass function based on modal parameter estimates derived using the MITD method on simulated data. circle – bins with pole estimates that are below the horizontal dashed black line and outside the white patched area; square – poor MAC valued pole estimates; triangle – stable pole estimates with similar MAC values as red circles does not satisfy step (4) and step (7), while modal parameter estimates denoted as blue squares does not satisfy step (5), see also Sect. 15.2. The remaining modal parameter estimates (and the bin they belong to), marked as green triangles, are strongly frequency stable and have high MAC similarity. Mean values and standard deviations are computed for the estimates in each of the combined bins. When looking at Fig. 15.2 the amount of spurious information is abundant, and in direct contrast to what was seen Figure in 15.1, where barely any spurious information was present. It is however seen that almost no spurious information is present at the first four modes. It should be mentioned that the same nine modes that was found when using the PLSCF algorithm were also found using the MITD algorithm. When zooming onto the first two modes, it is observed that below a model order of 50 the frequency component of the pole estimates are more stable than those found when using the PLSCF algorithm. At model orders above 50 it appears that the two algorithms produced similar results.
RkJQdWJsaXNoZXIy MTMzNzEzMQ==