Introduction
This bi-monthly seminar series explores real-world applications of physics-informed machine learning (Φ-ML) methods to the engineering practice. They cover a wide range of topics, offering a cross-sectional view of the state of the art on Φ-ML research, worldwide.
Participants have the opportunity to hear from leading researchers and learn about the latest developments in this emerging field. These seminars also offer the chance to identify and spark collaboration opportunities.
About the event
Non-Destructive Evaluation (NDE) underpins the safe manufacture and operation of safety-critical engineering components. NDE techniques include ultrasonic, electromagnetic, optical and radiographic modalities, each with numerous variants tailored for specific applications.
While data-driven artificial intelligence in the form of deep machine learning is enabling step changes in areas ranging from self-driving cars to medical image analysis, NDE is lagging behind. NDE is a conservative industry and heavily reliant on skilled human operators to perform inspections and analyse data. However, the increasing volume and complexity of data coupled with the anticipated trajectory of the fourth industrial revolution mean that this has to change.
This talk will present ultrasonic examples of how machine learning can be applied to the three basic NDE activities: property measurement, defect characterisation and defect detection. The key to the first two activities is the availability of labelled, high-fidelity training datasets in the large volumes (10,000+) necessary for training machine learning algorithms. This requires scalable simulation tools and it is the increasing availability of such tools that will be one of the enablers for widespread machine learning in NDE.
Conversely, defect detection algorithms can be trained on defect-free data which is generally much more readily available; however, the testing of detection algorithms still requires labelled, true-positive data so the need for large-scale simulations remains. The examples presented will be used to show some of the benefits and potential pitfalls of applying machine learning to NDE data. The talk will conclude with some thoughts on where research efforts are most needed.
Watch on demand:
