Experimental approach and statistical theory are applied to broaden knowledge of materials failure by fracture, from initial service surface conditions and during damage accumulation, using PMMA beams with repeated random surfaces, which were developed by computer controlled laser ablation, and then bent to fracture. The Central Limit Theorem was used to simulate dynamic degradation. Replicas allowed the study of probability of failure location (PFL) versus parameters of the surface topography. Results show that, depending on the auto correlation length (ACL), even at early stages of degradation, there is high correlation of PFL with RMS and average roughness, and two other proposed parameters. This correlation shows an asymptotic limit behavior that agrees with fracture mechanics results for brittle materials. As important, a Sensitivity Transitional Region (STR), dependent on the ACL, was observed, below which the PFL is uncorrelated to the parameters studied; The STR's dependence on temperature is currently being studied. (C) 2012 Elsevier Ltd. All rights reserved.