Aplicação de modelos mistos na análise das deflexões de pavimentos flexíveis em CBUQ.

Abstract

Furthermore, the distribution of thousands of kilometers of roads that connect the country, as well as the heterogeneity of geotechnical and climatic factors make road surface management difficult, perpetuating failures in maintenance and repair, which maintain their quality and increase their service life. In this scenario, prediction models that make it possible to obtain important characteristics for the analysis of road surface condition become essential. Several methods and different characteristics have been used for this task, from functional, structural, traffic and even climatic aspects. Thus, this paper seeks to develop a prediction model for road surface deflection based on easily obtainable characteristics, such as the Global Gravity Index (GGI) - a functional parameter obtained by visual inspection and field measurements - and project characteristics, such as paving thickness, as well as climatic and traffic factors. The data were obtained by observing homogeneous sections of rehabilitation projects of five federal highways and one state highway. By analyzing the database, it was possible to infer that its configuration suggested the implementation of a mixed model.Thus, the model was developed by deciding which variables would be significant for its random intercept and fixed effects. Furthermore, logarithmic transformations and data standardization were tested, in addition to the combination of several structures to identify the best model, which was chosen by the Akaike Information Criterion (AIC) and the Deviance Information Criterion (DIC).Thus, the model was developed by deciding which variables would be significant for its random intercept and fixed effects. Furthermore, logarithmic transformations and data standardization were tested, in addition to the combination of several structures to identify the best model, which was chosen by the Akaike information criterion (AIC) and the deviance criterion (DIC).Finally, it was possible to notice that the GGI positively affected the response variable, showing significant influence on its prediction. As the model generated satisfies the assumptions on which its adjustment was based, further studies can be developed for its implementation in management systems.