PI: Burcu Akinci
Co-PI(s): Silvio Maeta
University: Carnegie Mellon University
The Pennsylvania oil boom launched the American commercial oil industry when the first well was drilled in 1859 near Oil Creek, Venango County. Pennsylvania was one of the first states to have oil exploration fields in the country, starting around the 1860’s until the early 1900’s. Now, most of those oil exploration fields are covered by dense forests, and the abandoned wells present an environmental and societal hazard since they can leak gas and/or they can collapse, causing damage to property and endangering people. To eliminate such hazards, it is important to identify all of the wells in the region and seal them.
However, this is a challenging task, given the absence of records that show where those wells were drilled and the absence of visual marks that indicate the location of those wells. Any indications of man-made structures are mostly gone after one century. To address this problem, we propose to team up with a NETL team and utilize data from aerial lidar surveys to develop an automated processing approach to detect abandoned wells. Other sources of information (e.g. magnetic field variations caused by buried metal pipes, old pictures of Oil Creek, and the GPS position of already-identified wells) will be combined to improve detection accuracy. This project will evaluate different automated machine learning and deep learning approaches to assess their performance and feasibility in detecting abandoned wells.
The research team will conduct a detailed field case study with NETL in which they will process 3D imagery data and generate a list of candidate locations for abandoned wells. They will verify if the detected well candidates are valid against existing ground truth data and measure the confidence level. The team is well equipped to perform this research bringing knowledge on aerial robots, 3D imaging, and data analytics applied to civil engineering problems. Teaming up with NETL will provide unprecedented access to data and resources to perform a detailed case study at Oil Creek.