Amir Abbas Kazemzadeh Farizhandi^1*, Mahmood Mamivand^2*

1 Computer Science Department, Boise State University, 777 W Main St, Boise, ID 83702, US

2 Department of Mechanical and Biomedical Engineering, Boise State University, 1910 W University Dr, Boise, ID 83725, US

* Corresponding authors emails: amirabbaskazemza@u.boisestate.edu, mahmoodmamivand@boisestate.edu

DOI10.24435/materialscloud:es-a4 [version v1]

Publication date: Nov 21, 2022

How to cite this record

Amir Abbas Kazemzadeh Farizhandi, Mahmood Mamivand, Spatiotemporal prediction of microstructure evolution with predictive recurrent neural network, Materials Cloud Archive 2022.156 (2022), doi: 10.24435/materialscloud:es-a4.

Description

Prediction of microstructure evolution during material processing is essential to control the material properties. Simulation tools for microstructure evolution prediction based on physical concepts are computationally expensive and time-consuming. Therefore, they are not practical when either there is an urgent need for microstructure morphology during the process, or there is a need to generate big microstructure datasets. Essentially, microstructure evolution prediction is a spatiotemporal sequence prediction problem, where the prediction of material microstructure is difficult due to different process histories and chemistry. We propose a Predictive Recurrent Neural Network (PredRNN) model for the microstructure prediction, which extends the inner-layer transition function of memory states in LSTMs to spatiotemporal memory flow. As a case study, we used a dataset from spinodal decomposition simulation of FeCrCo alloy created by the phase-field method for training and predicting future microstructures by previous observations. The results show that the trained network predicts quantitatively accurate microstructure morphologies while it is several orders of magnitude faster than the phase field method. The trained model aims to generate future material microstructures by learning from the historical microstructures, where the visual dynamics are believed to have modular structures that can be learned with compositional subsystems. A PredRNN (https://github.com/thuml/predrnn-pytorch) model was trained to predict Fe-Cr-Co microstructures evolution during the spinodal decomposition.

Materials Cloud sections using this data

Files

File name	Size	Description
Spatiotemporal Prediction of Microstructure Evolution with Predictive Recurrent Neural Network.zip MD5md5:3874706f908102a14eabbc71097d450b	356.3 MiB	The Zip file contains the training and validation microstructure sequences, bash script for PredRNN (https://github.com/thuml/predrnn-pytorch) training, trained hyper-parameters,and the prediction of the trained model after 80000 iterations for 10 testing sequences. After downloading the PredRNN, the "train.npz" and "valid.npz" can be used for training the PredRNN by running the "predrnn_microstructure_evolution_train.sh" file for results reproduction.

License

Files and data are licensed under the terms of the following license: Creative Commons Attribution 4.0 International.
Metadata, except for email addresses, are licensed under the Creative Commons Attribution Share-Alike 4.0 International license.

External references

Keywords

Deep learning, Predictive Recurrent Neural Network Spatiotemporal Prediction, Material Microstructure Phase-field, Spinodal Decomposition