Machine Learning Applications in Advanced Additive Manufacturing: Process Modeling, Microstructure Analysis, and Defect Detection

Aug 5, 2026

Upload by Patrick

Non-destructive evaluation (NDE) techniques are essential for verifying material integrity, health, and mechanical properties in additive manufacturing. While high-fidelity NDE supports quality control, it often produces large data volumes that cannot be processed manually, creating a bottleneck for process engineers. This dissertation applies machine learning models to learn patterns, extract features, and uncover hidden relationships across diverse NDE datasets. Neural networks are used for prediction enhancement, anomaly detection, anomaly classification, image segmentation, material health estimation, and direct behavior modeling. The work also includes a grain image generation method for improved microstructure segmentation, CNN-based powder health qualification for stainless steel, and a feasibility study of binder jetting using Martian and Lunar regolith with a simplified binder, accelerating analysis and enabling faster advanced manufacturing development.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages170

Reading Time7 h 8 min

AccessFree to download

Table of Contents

List of Figures

List of Tables

Chapter 1: Introduction

1.1 Motivation

1.2 Intellectual Merit

1.3 Dissertation Outline

1.4 List of Publications

Chapter 2:

University of Central Florida STARS Electronic Theses and Dissertations, 2020- 2023 Machine Learning Applications in Advanced Additive Manufacturing: Process Modeling, Microstructure Analysis, and Defect Detection Peter Warren University of Central Florida Part of the Manufacturing Commons Find similar works at: [https://stars.library.ucf.edu/etd2020](https://stars.library.ucf.edu/etd2020) University of Central Florida Libraries [http://library.ucf.edu](http://library.ucf.edu) This Doctoral Dissertation (Open Access) is brought to you for free and open access by STARS. It has been accepted for inclusion in Electronic Theses and Dissertations, 2020-by an authorized administrator of STARS. For more information, please [contact](contact STARS@ucf.edu)[ STARS@ucf.edu](contact STARS@ucf.edu). STARS Citation Warren, Peter, "Machine Learning Applications in Advanced Additive Manufacturing: Process Modeling, Microstructure Analysis, and Defect Detection" (2023) . Electronic Theses and Dissertations, 2020-. 1693. [https://stars.library.ucf.edu/etd2020/1693](https://stars.library.ucf.edu/etd2020/1693) MACHINE LEARNING APPLICATIONS IN ADVANCED ADDITIVE MANUFACTURING: PROCESS MODELING, MICROSTRUCTURE ANALYSIS, AND DEFECT DETECTION by PETER WARREN B.S. University of Central Florida, 2017 M.S. University of Central Florida, 2018 A dissertation submitted in partial fulfilment of the requirements for the degree of Doctor of Philosophy in the Department of Mechanical and Aerospace Engineering in the College of Engineering and Computer Science at the University of Central Florida Orlando, Florida Spring Term 2023 Major Professor: Ranajay Ghosh © 2023 Peter Warren ii ABSTRACT Non-destructive evaluation (NDE) techniques are critical for assessing the integrity, health, and mechanical properties of materials manufactured from various methods. High fidelity NDE techniques are essential for quality control but often lead to massive data generation. Such a vast data load cannot be manually processed, this leads to a severe bottleneck for process engineers. Machine learning (ML) offers a solution to this problem by providing powerful and adaptable algorithms capable of learning patterns, identifying features, and finding hidden relationships in large sets of data. Various ML models are used in this work to improve predictions, improve measurements, detect anomalies, classify anomalies, segment images, determine material health, and directly model behavior. These neural network or ML models are implemented to perform these tasks by utilizing data gathered through various NDE techniques. Additive manufacturing enables the production of complex geometries and customized parts with reduced waste and lead times. The development of new material printing capability and techniques is necessary to expand its capabilities to produce high-performance parts with unique properties and functionality. Contributions to advanced additive manufacturing are made via the application of customized machine learning algorithms in this work. The development of a novel grain image generation method was completed to improve grain and grain boundary image segmentation methods on microstructure images. Convolutional Neural Networks (CNNs) were also applied to datasets of Stainless Steel Powder to help identify, qualify, and classify the health of the powder prior to print application. A feasibility study of the implementation of Binder Jetting (BJT) is conducted on Martian and Lunar regolith using a simplistic binder in this work. The need for efficient techniques to process data gathered from NDE methods is crucial to enhance the accuracy, efficiency, and speed of the analysis of this data. This will lead to faster development and implementation of advanced manufacturing techniques. This work is dedicated to my mother, Susan With Perszyk, who showed me what it means to be caring, kind, consistent, loving, determined,

FAQs

为什么在无损检测（NDE）中需要机器学习？

高保真 NDE 会产生海量数据，人工难以处理，导致过程工程师的分析瓶颈。机器学习能自动学习数据模式、特征并挖掘隐藏关系。

本研究中机器学习模型主要完成哪些任务？

用于提升预测与测量、检测与分类异常、对图像进行分割、评估材料健康，并实现对行为的直接建模。

论文在微结构与材料健康评估方面有哪些具体贡献？

提出用于生成晶粒图像的新方法以改进晶粒与晶界分割；并将卷积神经网络应用于不锈钢粉末数据，用于在打印前识别、限定与分类粉末健康状态。

Aug 5, 2026

Upload by Patrick