A Comparative Study of Multiple-Instance Regression Techniques With Data Augmentation for Cold Spray Powder Hall Flow Rate Prediction

Schuliger, Ashley

Etd

A Comparative Study of Multiple-Instance Regression Techniques With Data Augmentation for Cold Spray Powder Hall Flow Rate Prediction

Public

Cold spray additive manufacturing is an emerging technology that has been widely adopted for various applications, including the repair of military vehicles and equipment, to improve efficiency and save resources. However, the behavior of a powder during cold spray processing varies depending upon its properties, particularly its flowability. One common measure for flowability is the Hall flow rate, which measures the time taken for 50 grams of a material to flow through a funnel. Currently, domain experts do not fully understand the quantitative impact of a powder’s properties on its Hall Flow rate. As a result, we attempt to use classification to predict the flowability of a powder based on its various properties. However, each powder is composed of multiple particles, each with their own set of features. Current machine learning algorithms do not accurately model the relationship between multiple-instance data and their labels. Thus, we propose the use of a Multiple-Instance Regression framework to process multiple-instance powder data so that it can be fed into common machine learning models, such as a Decision Tree classifier. In this work, we compare and contrast the performance of three Multiple-Instance Regression frameworks in the scope of predicting the Hall flow of a powder. We also propose a novel strategy for augmenting multiple-instance data to offset the lack of resources and data available in the Materials Science field and the lack of augmentation techniques for multiple-instance data.

Creator