The anti-oxidation mechanism of SiCf/SiC-B4C modified with Al2O3 in wet atmosphere based on machine learning

To study the anti-oxidation mechanism of SiCf/SiC-B4C modified with Al2O3 in wet atmosphere, the damage evolution of composites after oxidation was explored by unsupervised machine learning technology (k-means). Results display that the mean feature values of cluster-1 (some small cracks in oxidation layer and matrix as well as fiber debonding) in composites modified with Al2O3 are larger than that in virgin after oxidation. Meanwhile, as the oxidation time increases, the concentrated area of cluster-2 (fiber breakage, cracks in axial and transverse yarns) in both composites gradually shifts toward the direction of high mean feature values. Because Al2O3 can protect the BN from oxidation, the increase of peak frequency and risetime of composites modified with Al2O3 is less than that of virgin. Moreover, the proportion of cluster-3 (big transverse matrix cracks) in composites modified with Al2O3 is always more than that of virgin. Based on the research results, the corresponding relationship between each damage behavior and each cluster is thoroughly built.