An adaptive damage monitoring method based on transfer features mapped for advanced composite structures

Monitoring damage in advanced composite structures proves challenging due to insufficient transfer and reuse of diagnostic models. Such limitations fail to meet practical monitoring requirements. This paper proposes an innovative method to monitoring damage via Lamb waves based on transfer learning. Firstly, the proposed method utilizes algorithm-centric transfer learning to extract transferable features from the source domain and a limited amount of data from the target domain, using the domain adaptive feature mapping. The aim of this method is to achieve feature adaptation in both the source and target domains by obtaining approximate distribution patterns in the feature space. Subsequently, a fine-tuning method is presented based on data-centric transfer learning for adaptive damage identification. A screening mechanism utilizing a coarse-to-fine strategy is employed to select suitable samples for diagnosing and evaluating damage in the target domain. The proposed method is verified by targeted monitoring cases of advanced composite structures under four monitoring stages, namely, detection of the damage existence, identification of damage types, damage localization and quantification. The results show that the proposed method can present higher monitoring ac-curacy in general damages for advanced composite structures than nine state-of-the-art methods.