Abstract: Damage to glass fiber epoxy resin reinforced composite（GFRP）,which is the main material of wind turbine blades,will lead to a reduction in blade life and even fracture,so damage detection must be carried out. The acoustic emission method can be used to analyze the damage characteristics of GFRP composite materials and realize online monitoring of blade performance. In this paper,the acoustic emission b-value feature is used to characterize the damage degree of GFRP,and the acoustic emission parameters are clustered by principal component analysis and the k-means++ algorithm to analyze the damage mode of GFRP. The results are as follows:1)The b-value analysis of the whole acoustic emission event shows that the b-value of the GFRP sample shows a significant downward trend before fracture,indicating that the change rate of the b-value can be used as an early warning signal of material fracture. 2)According to the proportion of nine acoustic emission parameters in the principal component analysis,the amplitude and peak frequency of the AE signal are selected. Through the k-means++ algorithm to cluster the characteristic parameters,it is found that the acoustic emission signals in the stretching process are divided into four clusters,and the characteristic frequency of each cluster is found. 3)The tensile fracture of the GFRP specimen is observed by scanning electron microscope（SEM）,and four damage modes corresponding to matrix cracking,fiber/matrix debonding,delamination and fiber fracture are obtained. The method based on acoustic emission b-value analysis is applicable to research on GFRP damage characteristics and can be popularized in the field of wind turbine blade damage detection.