top of page

Acerca de

Motor Neuron under the microscope in Lab._.jpg

Machine Learning

Background on this research

When we think about artificial intelligence or machine learning, we may visualize self-driving cars, like Telsa, or even the robots, like Spot, from Boston Dynamics. However, machine learning techniques have emerged as powerful tools for predicting damage within composite structures in recent years. Composite damage is complex and often very difficult to predict, but when introducing various machine learning algorithms, they can identify patterns and relationships that may be otherwise impossible.

​

An example of one of these algorithms is a neural network, also known as an artificial neural network (ANN). Neural networks are at the heart of deep learning and are inspired by the human brain, mimicking how biological neurons signal to one another. Similar to how we require data to learn, neural networks rely on training data to learn. Once the neural network is trained and tuned for accuracy, it can be an extremely powerful tool, allowing us to predict behavior based on a high-dimensional cluster of data.

​

An application for such a model would be composite and adhesive joints. Used in many structural applications like wind turbines, vehicles, and aircraft, these joints often hold the structure together. Yet, their behavior is very non-linear, making it hard to predict their strength and failure. The parameters that affect the performance of these joints are sometimes not clear either, whether it be temperature, surface preparation, material properties, or geometrical configurations. As a result, this work focuses on building a neural network that will predict composite and adhesive joint behavior.

bottom of page