Analytics for PHM Course
Course Presenter: Dr. Neil Eklund, Oak Grove Analytics
Course Administrator: Jeff Bird, TECnos and PHM Society Education & Professional Development Chair, jeffbird@rogers.com
Overview
This course is intended for engineers, scientists, and managers who are interested in data-driven methods for asset health management. You will learn how to identify potential data-driven projects, visualize data, screen data, construct and select appropriate features, build models of assets from data, evaluate and select models, and deploy asset monitoring systems. By the end of the course, you will have learned the essential skills of processing, manipulating, and analyzing data of various types, creating advanced visualizations, detecting anomalous behavior, diagnosing faults, and estimating remaining useful life. Note that this course is an advanced course with only a brief, high-level overview of PHM presented – students are expected to know the basics of PHM already. New practitioners are encouraged to take the fundamentals course or contact the course leader to examine their background and skills.
The content has been completely updated for the 2024 conference, and includes coverage of emerging trends in natural language processing and large language models, ensuring participants are equipped with the latest tools and insights to drive innovation in modern industries.
The course is predominantly lecture with and an optional hands-on lab. Students who elect to take the lab will be expected to bring a laptop to use either a Jupyter Notebook or Google Colab.
Topics include (most with embedded case studies and examples):
- Data: Types, Sources, Quality Issues, and Challenges
- Data Visualization
- Data Preprocessing and Feature Engineering
- Time-Series Data Handling
- Dimensionality Reduction
- Machine Learning Techniques for PHM
- Supervised Learning
- Unsupervised Learning
- Introduction to Neural Networks
- Deep Learning Techniques
- Anomaly Detection
- Diagnosis
- Remaining Useful Life (RUL) Estimation
- Physics-Based vs. Data-Driven Approaches
- Model Evaluation, Validation, and Interpretation
- Natural Language Processing and Large Language Models (LLMs) in PHM
- Case Studies and Real-World Applications
- Challenges, Solutions, and Lessons Learned
- Hands-On Sessions
We will go around the room on the first day to have short introductions from each participant to know their name, organization, and what they would like to get out of the course. We usually have a great mix of organizations and nations! Paper copies of the slides will be available when you arrive the first morning. No soft copies are provided.
Technical Labs
The optional (but encouraged) labs correspond to some of the major topics from the lectures. The labs can be done either in Jupyter Notebooks on your laptop or on Google’s Colab, a free notebook environment that runs in the cloud (you will need a Google account for Colab). Either platform will allow you to run all the necessary Python libraries. There is a Colab tutorial HERE.
If you intend to participate in the lab portion of the course, we will send you a link with the code and data to ensure you can open and execute it before the course begins. Data, code, and libraries are already included in the notebooks, so you will be able to explore the tools, tweak the code, and review the results.
We look forward to meeting you.
Dr. Neil Eklund, Oak Grove Analytics
Jeff Bird, TECnos and PHM Society Education and Professional Development Chair