The students will work on industry inspired research project utilizing data science skills in solving energy and infrastructure domain. Although the actual title and the scope of the industry project could vary slightly upon meeting with UTP team and industry sponsors, this section provides an overview of the student research projects.
For 2021 cohort, the focus of student projects will be centered on the theme of improving productivity and efficiency of energy (oil and gas) distribution system by assessing reliability of the pipeline and related infrastructure; assessing impact and risks of failures caused by extreme weather conditions, random events, and natural degradation processes; and finally, formulating maintenance strategy based on reliability and conditions of the infrastructures and equipment. Overall scope of the research can be classified in three different categories: risk assessment, reliability analysis, and maintenance planning.
Under overall theme/scope, there will be three projects: one for each team of three students from three participating institutions: NDSU, TAMU, UNLV, and one student from UTP. Different students will be assigned to lead on different components of the projects that are not aligned with their expertise and need to seek help from their group member(s). For example, industrial engineering team member will be assigned to work on environmental aspects while civil/environmental engineering team members will be in charge to big data tools. This approach will allow cross-training within the team.
1. Risk assessment of oil spills
Background: The research will study an oil spill case in the U.S. or Malaysia. The goal of this project is to gather different types of data in the past ten years including oil spills in the state/province such as spill volume, frequency, and cleanup time (including environmental cleanup). The research team will approach provincial, county, and state offices including entire region for other data such as weather, infrastructure/utilities, health, environmental, and economics/business. The student team will use these data for the first five years to predict risk levels in the last five years for oil leakage for different durations. Associated with oil leakage for different durations, the student team will calculate risks of environmental contamination (mainly water and soil) that will violate the legal standards, damage on crop and agriculture land, income loss below the poverty level and unemployment above 75% percentile of the nation, and population loss of more than 3% per year. For example, population loss might not be because of leakage directly but through unemployment.
Research plan: The team will determine these risks using @Risk which is a Monte Carlo simulation-based software. Then, using the first five-year data, the team will prepare a big data analytics model to predict the stated risks through data mining and machine learning algorithms which will be specifically created for the case. The team will use the predictive analytics model to quantify the risks for the last five years. The predictive analytics model developed will be modified to be more holistic and elaborative for environmental risks as oil and associated contaminants can travel far from the release point and can affect terrestrial and aquatic organisms and human through food chain. Therefore, the student team will have to incorporate toxicity to organisms including humans in the model. Contaminated sites can affect property value in the area and make people move away from the town resulting in population reduction. These are examples of aspects/issues that will be considered in the model preparation/modification. Eventually, the faculty mentor(s) will select two sensitive oil pipeline locations in the studied state/province and the student team will apply the predictive analytics model to assess the risks.
2. Reliability assessment of pipeline network
Background: A reliable pipeline network in extreme weather conditions requires pipeline operators to have fast and reliable access to at least two sets of big heterogeneous data from various sources. These data include but not limited to a) Real-time pipeline condition/operation data from the pipeline integrated management system (PIMS), which are collected from field pressure and temperature sensors including geographic information systems data, gas/oil operational pressure, and temperature. The abnormal changes in gas/oil pressure can indicate possible incident/leakage or any abnormal demands on a pipeline system, which is a key parameter that operators would continuously monitor to ensure the pipeline operates appropriately. The PIMS usually is backed up with radio communication and a separate on-site power generator to ensure its function during an extreme event; and b) the pipeline historical inspection and maintenance data such as wall thickness, recorded corrosion or damage, and cleaning/dewatering history. The PIs from US universities and faculty mentors from UTP will identify and work closely with the project sponsoring company to ensure these data are easily available.
Research Plan: The students will use these historical data to model degradation processes of the pipeline using available models such as regression or stochastic model such as gamma distribution to estimate model parameters and finally assess reliability. Subsequently, students will develop a machine learning algorithm based on the theory of networks to capture the degradation pattern. The deep machine learning algorithm will be trained to distinguish between normal and abnormal patterns to predict the probability of failures.
3. Maintenance planning
Research Background: To minimize failures in pipeline and improve availability and productivity of energy infrastructure, students will study the theory on strategic condition-based maintenance (CBM) planning1. This project on maintenance planning will employ a data-driven approach with different machine learning methods to (1) extract various degradation patterns, (2) predict the remaining useful life, and (3) determine when to execute the appropriate maintenance plan.
Research plan: For data collection purpose, student team working on this project will work closely with the team that will be working on project 2 (reliability assessment) as both teams will be working on similar data sets. To some extent, the outcome of project team 2 will be feeding project team 3 and hence these two teams will be required to work closely. This CBM platform will be able to provide emergency maintenance and services to the most needed infrastructure locations and obtain an optimized emergency response in extreme nature force conditions. Students will investigate at least two dynamic strategic CBM planning algorithms. In consultation with company engineers, faculty mentors and their own analysis, the team will decide the threshold level of degradation indicator that will be used as a criterion for decision making on maintenance or part replacement. If time permits, students will be encouraged to develop a maintenance optimization model considering cost minimization as objective function and degradation threshold level (also called as on-condition threshold) and inspection interval as decision variables.
Note: Given above are few sample projects that are covered by the overall theme of this IRES program. As noted above, the actual project may slightly vary in background and scope based on the industry need and availability of the data at the time of the project execution.
Overall Student outcome: The students will become familiar with the advancement in the field of big data analytics, machine learning, and optimization; and be able to prepare and apply a holistic tool (predictive analytics model) that is robust and can accurately predict risks of energy infrastructure failure and cascading risks on the environment, economics, and society. The students will develop an ability to function on an interdisciplinary global team and realize the value of other STEM and non-STEM disciplines.
1) Jain, N., Yadav, O.P., Rathore, A., and Jain, R. (2018). Maintenance planning based on reliability assessment of multi-state multi-component systems, Proceeding of International Conference on Industrial Engineering and Engineering Management, Bangkok (Thailand), Dec. 16-19.