Date of Award

5-2025

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Mechanical and Civil Engineering

First Advisor

Seong Hyeon Hong

Second Advisor

Ryan T. White

Third Advisor

Soheil Saedi

Fourth Advisor

Troy V. Nguyen

Abstract

Additive manufacturing (AM) is the process of creating a product by applying layers of material to build up to the final product. AM has revolutionized the field of manufacturing and has allowed more complex products to be manufactured at reasonable cost, quicker manufacturing time, and optimized material usage. Traditional forms of manufacturing were based on subtractive manufacturing, where by a product would be created by removing material from its source. AM is a new manufacturing process and faces significant challenges when it comes to manufacturing due to defects. Defects can vary from type and severity and can overall weaken or destroy a final product. Once a final product has received enough errors and defects in its production, the product will not meet the requirements and will need to be disposed. This disposal of material is wasteful and costs time, money, and material. If there was a way to actively monitor the AM process, we can prevent these defects from occurring and save a user time, money, and material. With my process of using neural networks (NN) to create a real time monitoring and fault detection, we can provide reliable and effective defect correction. My research determined there is a connection with a trained generalized autoencoder NN recognizing manufacturing defects of bowing, stringing, and gapping against non-defects for generalized types of shapes. I used an autoencoder with Visual Geometry Group 16 Layers (VGG16) backbone to create a generalized semi-supervised NN that is capable of differentiating manufacturing defects from non-defects.

Comments

Copyright is held by author.

Download

Available for download on Monday, November 10, 2025

Included in

Manufacturing Commons

Share

COinS