Machine learning in Al2TiO5 flexible ceramics with microcrcaks strengthened for damage mode automatic identification

The damage modes of Al2TiO5 flexible ceramics (AT) with microcracks strengthened were studied by machine leaning for damage mode identification. Results show that the optimal number of clusters increases from 3 to 4 with the increase of ZrO2 content. Meanwhile, the concentrated region of cluster with the same label gradually shifts to the direction of low amplitude with increase of ZrO2. In addition, the quantity of cluster is easy saturated since the AT grain boundary microcracks are strengthened by the formation of ZrTiO4. For achieving the automatic identification of different damage modes, support vector machine based on Kennard-stone training set selection was used to distinguish the damage modes of other samples in the same sample set. Almost the same evolution trend of corresponding clusters in new samples and training samples proves high efficiencies of this method.