Application of a priori knowledge-enhanced fuzzy clustering to acoustic emission-based damage identification of composite laminates

Abstract

Acoustic emission (AE) technology has been widely used in the researches on composite damage identification. Nevertheless, traditional classification and clustering models usually ignore the underlying physical mechanisms of the complex failure process of composites, limiting the comprehensive understanding and analysis of damage mechanisms. In this paper, a Prior Knowledge-enhanced Fuzzy C-Means (PK-FCM) is developed and validated by open-hole tension and compression experiments on plain-weave glass fiber-cyanate composite laminates. The experiments successfully subdivided the multiple stages of composite damage development with the help of AE monitoring, fracture morphology observation and in-situ penetration flaw detection techniques. The PK-FCM algorithm uses the experimental prior knowledge to guide the clustering, and specifically solves the problem of damage accumulation and evolution characteristics of composite materials. By dynamically adjusting the membership matrix, the cumulative effect and evolution order between damage modes are accurately described. Compared with the traditional K-mean and fuzzy C-mean (FCM) clustering methods, PK-FCM reveals the core features of the damage evolution of composite materials, significantly improves the accuracy and prediction ability of damage analysis, significantly improving the reliability of damage identification and advancing our understanding on the damage mechanisms of composite materials.