TEP Predictor

This website provides a machine learning model to predict thermoelectric properties of PbTe-based materials (PbTeNaAg). The current version is a prototype model; it will be updated in the near future. The detailed information on the material synthesis process is omitted because the purpose of this website is the introduction of our machine learning model, not real materials.
You can also find a thermoelectric power generation simulator on our web simulator.

Please follow the steps below to predict the material properties.

Step 1: Choose Material and Composition

The meaning of the composition (x, y) is the following:

Pb_1-xTeNa_xAg_y

Scatter Plot: Experiments (Expt)

Plot 1: Machine Prediction (Pred 1)

Plot 2: Machine Prediction (Pred 2)

Show Options

Step 2: Check the Thermoelectric Properties

Regression/Interpolation Methods for Thermoelectric Properties

In the plot, the filled region means 95% CI (confidence interval) of the mean. However, because the synthesis is rarely repeated in this example, the CI has no significant meaning.

Thermoelectric Properties

Methods

Machine learning model: We use an artificial neural network. We designed a special architecture for the prediction of physical property curves. The details will be made public in the near future.
CI prediction: We use a method that estimates the mean and the variance of the material properties, assuming that electrical resistivity, Seebeck coefficient, and thermal conductivity have an error of normal distribution. For details, refer to Nix and Weigend (1994).

Contacts

Dr. Jaywan Chung: jchung at keri.re.kr | Ext. 1631
Dr. Byungki Ryu: byungkiryu at keri.re.kr

Powered by

Google Fonts: Poppins font
Google Charts: Graphs

Machine Learning Model PredictingThermoelectric Properties ofPbTe-based materials ver.0.1a