“…These models aim to represent stress ields near single ibre-breaks accurately, although most assume regular packing [5, 17-25, 27, 30]; many models [16,18,25,26] also neglect the increase in the stress-recovery length near broken clusters, which was shown to overestimate bundle strength [30]; moreover, all models [5,[16][17][18][19][20][21][22][23][24][25][26][27]30] consider simpli ied matrix constitutive laws (e.g. neglecting progressive matrix fracture), neglect dynamic stress concentrations, and assume perfectly-aligned ibres (the only exceptions [28,29] considered very small representative volume elements). In addition, simulation results are often not compared against experiments [16-20, 26, 28, 29], and when compared they tend to overpredict the average strength of composites [5,21,22,24,25]; they also signi icantly underpredict the variability of strength distributions [25], and underpredict (or fail to predict) M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT size-effects for increasing ilament counts in bundles [22,26].…”