Abstract: U HMWPE/ LDPE composites were tested to collect acoustic emission (AE) signals under a quasi2static
tensile load. The unsupervised pat tern recognition ( PR) techniques were used to classify AE signals af ter their
preprocessing. Damage mechanisms in several specimens (0° 、90°and [ + 45° / - 45° ]) were investigated according to the classifying result s. The study shows that the PR technique is able to identify damage modes in the specimens such as mat rix cracking , fiber2mat rix debonding , fiber pullout , fiber breakage , etc. The identification result s are the same with the observation result s of the f racture surfaces of the specimens observed by a scanning elect ron microscope (SEM) . For U HMWPE/ LDPE composites , the characteristics of the AE signal are not affected by the type of specimen but are dependent on the failure mode. The PR technique can classify the AE signals issued f rom different modes effectively. By the cumulative AE hit s of each damage mode vs st rain curves , the damage process of the specimens can be reviewed clearly. Using the PR technique on AE signals gives nicety evidence for the damage mechanisms analysis of composites.