A probabilistic approach to evaluate the relationship between visual and quantified pavement distress data using logistic regression

摘要

Detailed measurements along with visual ratings of major pavement distresses assist highway authorities in maintenance decision makings. The ratings are allocated by professional pavement engineers whereas the objectively collected measured data are collected through electronic or automated devices by trained personnel who may have a lack of experience. Therefore, data quality discrepancy from both types of surveys has gained attention in pavement maintenance management to find the reliability of pavement distress data to predict the overall pavement condition, both at the project and network level. This research employs probabilistic logistic modeling to evaluate the consistency in two types of survey data at the network level. The measured distress used in developing the logit models include crack (% area involved), rut depth (mm), and loss of surface texture (left wheel path %). Developed logistic models predict visual crack and deformation conditions from quantified distress data with a medium success rate (55% to 61%). However, deformation (sprayed sealed network) and texture loss (both asphalt surfaced and sprayed sealed network) data cannot be validated due to the failure of the logistic models. The gradual deterioration process of the pavement surface associated with loss of texture makes it difficult to detect visually. In the case of deformation ratings, assessors evaluate both longitudinal and local depressions. It appears that other local depressions dominate longitudinal depression (rutting) in the sprayed sealed network, and hence the data from both types of surveys are not related statistically significantly in this logistic approach. Data collection and synchronization error in the objective survey have potential influences as well, in creating this disagreement. The approach used in this study would help the state road authorities to ensure the data integrity in developing overall pavement condition models for the bituminous road network.