1. Home
  2. RIP
  3. View Record
https://www.nationalacademies.org/trb/events

Pavement Conditions Assessment and Prediction (PCAP): A geospatial machine learning approach to inform decision-making

The Maine Department of Transportation (MaineDOT) continues to observe an increased rate of pavement deterioration on its 8,800 mile roadway network, which is the largest and most heavily used component of the transportation system under the MaineDOT’s jurisdiction. Pavement deterioration is governed by a variety of factors, including traffic load, quality and design of the pavement structure, increased frequency of climatic events like freeze-thaw cycles, topographic influences and drainage, and geologic considerations like the native subgrade soils. While these factors have been identified individually as potential attributes to pavement degradation and distress, it is likely the confluence of several attributes that impute the greatest rate of degradation on pavement systems. However, the combination(s) of attributes linked to varying degrees of the pavement degradation rate remain poorly understood and must be identified to make informed decisions regarding resource allocation. This project seeks to identify and link the combination(s) of attributes described in the preceding section (e.g. pavement design/structure/quality, traffic loading, environmental stressors) to temporal and spatial differences in the rate of pavement degradation on MaineDOT’s highway network; i.e. to understand the relative influence of attributes imputing pavement distress. By working with the MaineDOT, UMaine will use existing and/or collect new pavement quality data (geo-located cracking index values) using the Automatic Road Analyzer (ARAN) to quantify the degree of pavement distress. ARAN data surveyed across the state will allow an assessment of variations in pavement quality across pavement types (e.g. new construction, rehabilitation, spot improvements, LCP, preservation paving), regions/space (i.e. for consideration of climate, geology, drainage, wetness, soil, and traffic loading) and epochs (time since last paving or improvement). The project is expected to consist of three components: Phase 1a (3-6 months): A literature review of existing studies and methods that incorporate data-driven analyses of spatial and/or temporal differences in the rate of pavement degradation. Phase 1b (18-21 months): Data-collection and integration, mapping & visualization, and predictor selection and attribution of factors influencing pavement degradation rates via machine learning. Phase 2: (18 months): Extension and refinement of Phase 1b to develop a tractable forecasting model to predict the degradation rate of pavement systems.