GlassNet: a multitask deep neural network for predicting many glass properties

Daniel R. Cassar

DOI: https://doi.org/10.1016/j.ceramint.2023.08.281

2023-11-21

Abstract:A multitask deep neural network model was trained on more than 218k different glass compositions. This model, called GlassNet, can predict 85 different properties (such as optical, electrical, dielectric, mechanical, and thermal properties, as well as density, viscosity/relaxation, crystallization, surface tension, and liquidus temperature) of glasses and glass-forming liquids of different chemistries (such as oxides, chalcogenides, halides, and others). The model and the data used to train it are available in the GlassPy Python module as free and open source software for the community to use and build upon. As a proof of concept, GlassNet was used with the MYEGA viscosity equation to predict the temperature dependence of viscosity and outperformed another general purpose viscosity model available in the literature (ViscNet) on unseen data. An explainable AI algorithm (SHAP) was used to extract knowledge correlating the input (physicochemical information) and output (glass properties) of the model, providing valuable insights for glass manufacturing and design. It is hoped that GlassNet, with its free and open source nature, can be used to enable faster and better computer-aided design of new technological glasses.

Soft Condensed Matter,Materials Science

What problem does this paper attempt to address?

The main goal of this paper is to develop a multi-task deep neural network model capable of predicting various properties of glass. Specifically, the researchers trained a multi-task deep neural network named GlassNet, aiming to predict up to 85 different properties based on different glass chemical compositions, including optical, electrical, dielectric, mechanical, and thermal properties, as well as density, viscosity/relaxation behavior, crystallization, surface tension, and liquidus temperature.

The main contributions of the paper can be summarized as follows:

1. **Model and Data**:
- GlassNet is a multi-task deep neural network built on a training dataset of over 218,000 different glass compositions.
- The model can predict 85 different glass properties and is applicable to various types of glass chemistry (e.g., oxides, chalcogenides, halides, etc.).
- The paper provides the model and training data as part of the free open-source software GlassPy module for use and further development by the research community.

2. **Performance Validation**:
- By combining GlassNet with the MYEGA viscosity equation, the viscosity variation with temperature was predicted, and it outperformed the existing general viscosity model ViscNet on unseen data.
- Using interpretable artificial intelligence algorithms (SHAP), the researchers extracted knowledge about the relationships between inputs (physicochemical information) and outputs (glass properties), providing valuable insights for glass manufacturing and design.

3. **Methodological Innovations**:
- The study adopted a multi-task learning approach to improve prediction performance, where a single model can predict multiple related outputs.
- Experimental results supported the hypothesis that multi-task learning can bring better performance when expanding the model's scope (inputs and outputs).
- A multi-head neural network architecture was proposed, which includes a shared multi-layer perceptron and multiple heads specialized for specific property predictions to further enhance the model's performance.

4. **Practical Application Value**:
- It is expected that GlassNet, through its free open-source nature, can accelerate the computer-aided design process of new technological glass materials.

Overall, this paper demonstrates how machine learning techniques (particularly multi-task deep learning) can be effectively used to predict various properties of complex materials like glass, opening up new possibilities for materials science research.