A Learning-Based Framework for Error Compensation in 3D Printing.

Area of Science:

• Cyber-Physical Systems
• Additive Manufacturing
• Artificial Intelligence

Background:

• 3D printing offers mass customization but suffers from low accuracy compared to traditional methods.
• Arbitrary object shapes and small batch sizes hinder universal error compensation techniques.
• Manual error compensation is inefficient and costly for 3D printing applications.

Purpose of the Study:

• To develop an automated framework for error compensation in 3D printing.
• To leverage 3D deep learning for accurate deformation prediction and correction.
• To enhance the precision of 3D printed objects, particularly for custom applications like dental crowns.

Main Methods:

• Utilized 3D scanning to capture object geometry.
• Trained a deep neural network using a large dataset for deformation function learning.
• Proposed and implemented an "inverse function network" for error compensation.
• Employed a convolutional AutoEncoder for end-to-end learning.

Main Results:

• The deep learning model successfully learned deformation functions for specific tasks like dental crown printing.
• The inverse function network effectively predicted and compensated for translation, scaling, and rotation errors.
• Experimental results demonstrate significant improvements in 3D print accuracy.

Conclusions:

• The proposed 3D deep learning framework offers an effective solution for automatic error compensation in 3D printing.
• This method enhances accuracy with minimal impact on hardware expenses.
• Represents the first application of deep neural networks for error compensation in 3D printing.