Machine Learning for Pipeline Integrity Management

Event Summary

Machine learning is emerging as a fundamental practice in support of pipeline integrity and risk management. As the amount and complexity of data is ever expanding, machine learning is making it possible to efficiently identify risks and manage what is important to integrity objectives. This course presents the fundamentals of machine learning through interactive discussion and presentation of real-world use cases.

Attendees will receive 16 Professional Development Hours (PDH) upon completion.

Course Objectives

This course is a 2-day highly interactive hands-on workshop using provided example or attendee data to take the mystery out of machine learning and associated analytics in the context of pipeline integrity. Through a secure web-based application, the attendee will work through key fundamentals and use cases to personally experience how models are learned and validated, complex patterns revealed, outliers identified, and data quality assessed and mitigated.

Day 1 of the course focuses on the machine learning process and fundamentals including learning types, learning models, model validation, data quality and mitigation, and data preparation in support of learning.

Day 2 focuses on high-value use cases including learning threat models, predicting mitigation costs, predicting missing data, assessing rare threats, threat model validation and using machine learning to support risk analysis. Special emphasis will be placed on how machine learning practices can support new and existing PHMSA based requirements.

The expectation is that the attendee will leave with a sound understanding of the process, the ability to converse in the context of machine learning, and access to resources to continue building their expertise.

Learning Objectives

This course will present the basics of machine learning and how the process is used to support integrity and risk management. You will learn:

• Data sampling, preparation & quality assurance methods
  • Feature analysis & engineering
  • Classification & cluster learning methods
  • Regression learning methods
  • Basics of inferential statistics & sampling
  • Outlier detection
  • Model validation
  • Machine learning based risk & QRA
  • Where to go to learn & perform your own machine learning

Who Should Attend

• Pipeline Integrity Managers – Managers who are responsible for integrity management programs will gain a basic understanding of machine learning methods and how the model performance indicators of the model can be used to support allocation of resources and budget.
• Pipeline Integrity Engineers – Engineers who are responsible for identifying potential risk will understand how to develop machine learned models with a specific target of interest (e.g. external corrosion predictions).
• Data Science Analysts – Data specialists will learn Exploratory Data Analytics (EDA) fundamentals and the best practices for collecting the data required for machine learning models.

Course Materials

Will be shared with you digitally the week prior to the workshop.

Meals

Continental breakfast & lunch will be provided daily.

Meet Your Instructors