Abstract

Evaluating the stress distribution in structures under temporal loads is being carry out by many of the engineering applications such as: impacts, cracks, bending, thermal-transient and other. In those cases, conventional photoelasticity techniques are more complex to evaluate the stress field because of their complicated and expensive experiments, quantity of computational procedures, and their time by time analysis. However, dynamic photoelasticity experiments produce temporal information, such as color variations, which could be analyzed, described, and classified in order to perform a whole stress field evaluation. In this paper, the one-dimensional local binary patterns (1D-LBP) are used to describe such color variations and use them to identify the stress values they belong. For different experimental configurations, this proposal achieved an accuracy of 98% when evaluating the stress field of cases with similar light sources than with a reference experiment, and 92% for experiments with other light conditions. These results make this descriptor able to determine categorical stress maps from a photoelasticity video itself, which significantly opens new opportunities to simplify the experimental and computational operations that limit the stress evaluation process in line with the dynamic experiment.