A Pavement Management System (PMS) becomes an effective decision-making tool when its performance models are accurate, and its trigger points (values) and benefit weight factors are precisely calibrated. Since 2011, NCDOT began collecting pavement distress data using automated methods. Once new performance models are developed, they can present different deterioration rates, benefit curves, and decision trees. Such impacts on trigger points and benefit weight factors need to be evaluated, and new values need to be determined if necessary. This study was conducted to meet these needs.
In this study, automated data collected in 2013, 2014, and 2015 were analyzed in order to determine maximum allowable extent (MAE) input and threshold values. Then these raw data were cleaned and used to develop distress and performance models for 10 asphalt pavement (ASP) families and concrete (JCP) pavements. Cost-benefit analysis (CBA) was performed to determine the benefit weight factors on decision trees.
Primary findings of this study include:
- The newly developed distress models can be implemented into the NCDOT PMS with preliminary trigger points determined.
- MAE input values are essential in obtaining correct distress index values. Percentiles of distress index values and input from NCDOT engineers are the key information to derive appropriate MAE input values.
- Collecting importance scores of JCP distresses from NCDOT engineers and researchers is an imperative step in calculating PCR values of JCP pavements. The scores are accurate as attested by the robustness of JCP models.
- For ASP pavements, newly developed PCR curves are visually comparable to PCR curves developed using manual data.
A new set of Weight factors were determined by performing CBA analysis and statistical regression. They are: 2.68 for Interstate, 1.26 for US, 1.16 for NC, and 1.0 for SR.
Recommendations for further avenues of research are:
- Pretreatment condition can have significantly impact on treatment performance. It is recommended to include pretreatment condition as a grouping factor when develop performance models. For example, the Interstate 0-50k family can be divided into three sub-families based on Good/Fair/Poor pretreatment condition: Interstate 0-50k/Good, Interstate 0-50k/Fair, and Interstate 0-50k/Poor, and three family models can be developed to more accurately predict pavement performance.
- More advanced image recognition techniques are recommended to be used to improve the quality of raw performance data. One such technique is deep learning, which has proven to be superior to traditional computer vision algorithms and if trained appropriately can improve the quality over time.
- To transition the NCDOT PMS to full-use of automated data, the following tasks are recommended:
- Step 1. Redefining roadway families by adjusting AADT breakpoints for more consistent performance within families.
- Step 2. Developing new distress and performance models once more automated data become available.
- Step 3. Loading automated data and newly developed models (Step 2) into the NCDOT PMS.
- Step 4. Determining a new set of benefit weight factors using cost-benefit analysis (CBA).
