Deep Learning Used to Predict Stress in SLA 3D Printed Structures

In a thesis entitled “Deep Learning Based Stress Prediction for Bottom-Up Stereo-lithography (SLA) 3D Printing Process,” a University at Buffalo student named Aditya Pramod Khadlikar describes a method of predicting stress distribution in SLA 3D printed parts using a Deep Learning framework. The framework consists of a new 3D model database that captures a variety of geometric features that can be found in real 3D parts as well as “FE simulation on the 3D models present in the database that is used to create inputs and corresponding labels (outputs) to train the DL network.”

As Khadlikar points out, part deformation and failure during the separation process are common problems encountered in bottom-up SLA 3D printing.

“Cohesive Zone Models have been successfully used to model the separation process in bottom-up SLA printing process,” Khadlikar says. “However, the Finite Element (FE) simulation of the separation process is prohibitively computationally expensive and thus cannot be used for online monitoring of the SLA printing process.”

Therefore, Khadlikar created an alternate method of predicting stress. A convolutional neural network (CNN) was used to develop a deep learning framework that could calculate the stress induced in any layer of a CAD model in real time to assist in online monitoring of the bottom-up SLA 3D printing process. To train the network, a dataset was created using Autodesk Inventor API, and ABAQUS python script was used to carry out FE
simulations on the generated dataset.

Experiments were carried out on multiple samples using the CNN. Several parts with similar cross-sections at a particular layer were examined to see the stress distribution on that layer for a given part. Khadlikar and colleagues discovered that different parts with the same cross-section at a particular layer had different stress distribution at that layer.

“This shows that for non-uniform 3D parts, along with given layer information we need information from the previous layers as well,” Khadlikar says. “This motivated us to develop a new architecture where the stress information of the previous layer is also used for stress prediction for a given layer.”

Stress distribution on the cured layer

An important conclusion reached was that CNN is drastically faster than FEA simulation. The created dataset worked effectively, helping to determine parameters such as peak stress and dependence on previous layer information to determine the stress distribution on a layer. The deep learning model, overall, outperformed the simple neural network model previously used for stress prediction.

“This framework can also be further used for training a larger dataset of 3D parts with varying heights as well,” Khadlikar says. “This framework cannot be used to predict stress on all the layers in a 3D part. This is due to the fact that previous layer stress information to predict current layer stress. Using a prediction of the previous layer to predict current layer stress induces more error due to compounding. Future work will be stress prediction on each layer of the 3D part…A good direction for future research can be incorporating more parameters like the height of slice and pull-up velocity to mimic the 3D printing process more realistically and to get better control over the process.”

Discuss this and other 3D printing topics at 3DPrintBoard.com or share your thoughts below.