Georgia Institute of Technology
A Data-Driven Approach to Machining Process Planning: Learning Sequences from 3D Geometry Using Transformers
Abstract
dc:description.abstractEfficient and intelligent machining process sequencing is a key challenge in computer-aided process planning (CAPP). Traditional automated approaches often rely on manually defined rules and explicit feature recognition, making them difficult to generalize across diverse parts and evolving manufacturing environments. Recent advances in deep learning (DL), particularly in sequence modeling with transformers, offer a promising alternative by enabling systems to learn directly from data without the need to explicitly model complex dependencies. This thesis proposes a novel DL-based framework that predicts machining sequences directly from the 3D geometry of final parts. The approach eliminates the need for rule-based logic or feature extraction modules by operating on a voxelized representation of the final part and predicting an ordered sequence of operations, each associated with a volumetric shape representing the material removed from raw stock. The framework integrates three key components: an autoencoder based on 3D convolutional neural networks (CNN) for compressing high-dimensional voxelized data, a transformer-based sequence autoencoder for modeling operation sequences, and a 3D CNN encoder that maps final part geometry to latent representations of valid machining plans. The model is trained on a large synthetic dataset of prismatic parts governed by realistic manufacturing constraints, including precedence rules based on geometric interactions. After training, the model achieves a sequence prediction accuracy of 99.48% and reconstructs final part geometry with a volumetric intersection-over-union (IoU) of 97.33%. Results demonstrate the framework’s ability to generalize sequencing logic and shape transformations directly from data, offering a flexible and scalable approach to process planning. This study provides a foundation for extending DL in manufacturing toward more complex, real-world scenarios.
Degree
thesis:*- Level thesis:degree_level
- Masters
- Department dc:contributor.department
- Mechanical Engineering
- Grantor dc:publisher
- Georgia Institute of Technology
- Year dc:date.issued
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Maqueda Gomez, Jaime
- Advisor dc:contributor.advisor
-
- Melkote, Shreyes N.
- Committee members dc:contributor.committeemember
-
- Kurfess, Thomas
- Rosen, David W.
Subjects
dc:subject × 3Rights
- Language dc:language.iso
- en_US
Identifiers
dc:identifier.*- Handle dc:identifier.uri
- https://hdl.handle.net/1853/82021