Abstract A numerical technique based on the Monte-Carlo method in a shear-lag model is developed to simulate the tensile strength and fracture process for a unidirectional carbon-fiber-reinforced-plastic (CFRP) composite. The technique improves on the conventional approach by using an r min method that can determine the stresses working on the fiber and matrix elements as the damage progresses. The r min method is based on tracking the incremental ratio of the strength of an element to its stress. The present model includes the effect of the sliding frictional forces around fiber breaks caused by debonding between the fiber and matrix. Statistical properties of the tensile strength were obtained through simulation runs involving 100 samples for each value of the frictional force parameter. Also studied was the size effect in composite strength with increasing numbers of fibers, N, where it was found numerically that mean strength varies linearly as 1 [l n( N)] 1 2 and coefficient of variation varies linearly as 1 l n( N) , as suggested from a simple theory.