Artificial Intelligence and Machine Learning for Nondestructive Evaluation
July 25-26, 2020
9:00 AM – 4:00 PM daily (light refreshments & lunch included in registration fee)
Instructor: Dr. Joel B. Harley, University of Florida, Electrical and Computer Engineering
This short course is designed to be a crash course on the fundamentals of machine learning and artificial intelligence. Its goal is to teach fundamental terminology and concepts (e.g., when machine learning succeeds and when it fails), foundational mathematics (i.e., linear algebra and optimization), applied algorithms (i.e., neural networks), and how they can be applied to nondestructive evaluation (through demonstrations). This is an aggressive course structure that is designed so that by the end of the course, students will be able to design machine learning and artificial intelligence solutions for a given nondestructive evaluation problem.
Note that this class focuses on teaching concepts and mathematical foundations rather than teaching how to code since participants will have variable coding experience and significantly different coding needs. In addition, there are a large number of well-designed coding tutorials on the internet. We believe that we can provide much more value through teaching foundational material and their relationships and uses within nondestructive evaluation, which is material that is not available anywhere on the internet. That being said, each part of the course will be accompanied by code demonstrations that will be freely available to course participants. The code will be available on CodeOcean, which is a reproducible code archive that allows students to run and modify code online (without the need for any specialized software). There exists no online course similar to this.
This course is a shortened form of a 4-day course taught at the Air Force Research Laboratory in September 2019. As a result, most of the material already exists. Relative to the previous course, the shortened course will remove some initial data science material that relates less to machine learning and artificial intelligence as well as remove more mathematically rigorous sections (such as mathematical properties of convex optimization) that are not necessary for the given audience. A larger component of the course will focus on concepts than mathematics. We will further polish and add new (NDE focused) coding demonstrations to this version of the course.
At the conclusion of this course, you should be able to:
- Discuss machine learning with terminology common within machine learning communities
- Understand the relationships between linear algebra, optimization, and machine learning
- Design a machine learning solution for given nondestructive evaluation problem
- Identify the purpose of specific neural network designs
Dr. Harley is the director of the SmartDATA (Smart Diagnostics, Acoustics, and Time-series Analysis) Laboratory at the University of Florida. Dr. Harley’s research integrates advanced signal processing, machine learning, and artificial intelligence into ultrasonics, nondestructive evaluation, and industrial applications. In the last six years as a professor, Dr. Harley’s lab has worked on funded projects for the National Science Foundation, the Air Force Office of Scientific Research (with a Young Investigator Award), the Department of Energy, and the Florida Department of Transportation. His projects address a broad array of engineering challenges, from standardizing ultrasonic inspection of ultra-high performance concrete to designing digital surrogates for improving aircraft structural monitoring to formulating new theories that integrate wave physics into deep learning.
Dr. Harley and his lab have published more than 26 refereed journal papers and 74 conference papers, including four best student papers. He has given a four-day course at the Air Force Research Lab on the fundamentals of machine learning for mechanical engineers and material scientists. He is the co-founder of the Machine Learning and Statistical Methods in Nondestructive Evaluation session at the Review of Quantitative Nondestructive Evaluation (QNDE) conference. He earned a University of Utah (where he was formerly a professor) teaching award and a mentorship award. He is an associate editor of Structural Health Monitoring. In 2020, he was awarded the IEEE Ultrasonics, Ferroelectrics, and Frequency Control Society Star Ambassador Award to allow him to present his work on machine learning in ultrasonics across the country. Dr. Harley has been regularly attended QNDE for since 2011.