This work presents a coupling between three multivariable analysis techniques (Principal Component Analysis (PCA), K-means and Kohonen Self-Organizing Map (KSOM)) applied to the acoustic emission data recorded on Glass Fiber-Reinforced Polymer (GFRP) composite materials in order to monitor and identify, in real-time, their damage mechanisms: matrix cracking, interfacial debonding, fiber breakage and delamination between layers. Two mechanical loadings were used during this study: a monotonic tensile test until the failure and a step-wise tensile test of 50 MPa each time (7 ramps and 6 levels of 4 min holding time). The first loading, applied to the specimens in pure epoxy resin, unidirectional (UD) [0]4 and [90]4 GFRP, as well as the laminates [0/90]S, allowed to evaluate the acoustic signature of each damage mechanism and establish a physical learning basis. The obtained physical data were employed for the learning operation of the Kohonen map which will be used for the identification of the damage mechanisms according to the level of the applied loading in the gradual tensile test. Post-mortem inspections conducted on the fracture facies of tested specimens under SEM confirmed the relevance of this {multivariable statistical analysis/acoustic emission} coupling for the detection and identification of GFRP damage mechanisms. Thus, the results of this study showed the relevance to identifying the damage mechanisms generated in a GFRP material by using multivariable acoustic emission analysis and provided a real potential for damage identification that would be developed in composite structures, made with the same material, under in-service loadings.