Implementing Machine Learning with Highway Datasets

Every year Maryland Department of Transportation State Highway Administration (MDOT SHA) invests millions of dollars into testing geomaterials to optimize engineering designs. There is a significant cost savings opportunity, by leveraging the historic material testing data with predicative machine learning models to provide estimated values for the engineering characteristics for newly proposed projects. The Office of Materials Technology has developed a deep learning Neural Network to predict drilling data, but this is just a first step. The purpose of this study is to write algorithms to improve, refine and optimize the accuracy of this neural network, and to develop other machine learning models by continually updating, retraining, and optimizing the neural network geometries. The study will evaluate and test additional data sets such as laboratory soil test data, in-situ and geophysics data, as well as others. It will also investigate the opportunity to implement machine learning for other large datasets in OMT such as pavement data, construction history, slope stability, geologic risk, as well as others.

    Language

    • English

    Project

    • Status: Active
    • Funding: $151468
    • Contract Numbers:

      SPR20B4E

    • Sponsor Organizations:

      Maryland Department of Transportation State Highway Administration

      707 North Calvert Street
      Baltimore, MD  United States  21202
    • Project Managers:

      Xiang, Hua

    • Performing Organizations:

      University of Maryland, College Park

      Department of Civil and Environmental Engineering
      College Park, MD  United States  20742
    • Principal Investigators:

      Zhang, Yunfeng

    • Start Date: 20191023
    • Expected Completion Date: 20210430
    • Actual Completion Date: 0

    Subject/Index Terms

    Filing Info

    • Accession Number: 01743424
    • Record Type: Research project
    • Source Agency: Maryland State Highway Administration
    • Contract Numbers: SPR20B4E
    • Files: RiP, STATEDOT
    • Created Date: Jun 22 2020 11:48AM