Automatic extraction of STL features

Published: 4 September 2024| Version 1 | DOI: 10.17632/jbkpf5p97b.1
Contributors:
Sonali Sagar Patil, Sandip Thite, Yogesh Deshpande, Dattatraya Parle

Description

The dataset consists of the following attributes, each representing different aspects of 3D printed models: 1. File Name: The name of the STL file representing the 3D model. 2. Volume: The overall volume of the 3D model in cubic units, indicating the amount of space the object occupies. 3. dX, dY, dZ: The dimensions of the bounding box that encloses the entire 3D model, along the X, Y, and Z axes, respectively. These dimensions help understand the spatial extent of the model. 4. Number of Layers: The total number of layers that make up the 3D printed object. This reflects the complexity and granularity of the model. 5. Area of First Layer: The surface area of the first layer of the model. This is crucial for assessing the initial contact with the build platform, which affects adhesion. 6. Area of Last Layer: The surface area of the last layer of the model, which might differ from the first layer, providing insights into the model’s geometry and build process. 7. comFLx, comFLy, comFLz: The coordinates of the center of mass of the first layer along the X, Y, and Z axes. These values are important for understanding the balance and distribution of mass at the initial layer. 8. comLLx, comLLy, comLLz: The coordinates of the center of mass of the last layer along the X, Y, and Z axes, providing similar insights as the first layer but for the completion of the model. 9. Center of Mass Difference: A tuple representing the difference in the center of mass between the first and last layers, which helps in analyzing stability and structural integrity throughout the build. 10. Build Adhesion: A binary or categorical attribute indicating whether the initial layer adhered properly to the build platform (1 for successful adhesion, 0 for failure). This is a critical quality metric in 3D printing. This dataset is designed to analyze the relationships between these attributes and to predict the likelihood of successful build adhesion in 3D printing processes.

Files

Categories

Machine Learning, Three Dimensional Printing, Automated Machine Learning

Licence