GeoDict-Cloud

We support researchers who envision the next generation energy materials on the microscale by establishing a seamless interface between local and Cloud computing for digital materials design, by generating and processing big data fast and so, reducing time to market. The current speed of materials R&D will multiply by linking an online material repository, solutions for batch operation in the Cloud, training, and usage of artificial neural networks (ANNs) for material property prediction.

For the next generation of energy materials, such as SOFC electrodes, understanding and optimizing the performance on the microscale is essential. Digital material design is currently restricted by the lack of computational power, but AI and unlimited computational power in the Cloud on demand will fully unleash its potential. Bringing the design of materials into the Cloud is necessary for the digitalization of R&D and to reduce time to market of new materials.

In the current joint venture between KaleidoSim (KS) and Math2Market GmbH (M2M), GeoDict users run simulations in the Cloud, by uploading simulation cases via a web interface. The main software development goal of this project is a new module called GeoDict-Cloud, which provides a direct GeoDict interface for Cloud computing and seamlessly integrates the KS Cloud with GeoDict. Without leaving the GeoDict GUI, the user is able to handle data pre- and post-processing, data transfer and management as well as batch operation routines in the Cloud.

GeoDict-Cloud contains GeoApps to run hundreds of simulation cases simultaneously in the Cloud, enabling fast and reliable digital materials design based on large data analysis. The user will be able to

• predict effective material properties based on given structural parameters, or
• obtain optimized structural parameters for the newly developed material to present certain effective properties.

During the project and supported by KS and M2M, the ZHAW will investigate, develop, and validate these batch operations and AI approaches to develop the next generation of energy materials with optimized microstructure. The developed concepts and approaches will be implemented into the GeoDict-Cloud module by M2M as GeoApps. In addition, the ZHAW will establish a public online material repository with standardized format for automatic exchange with the GeoDict-Cloud. M2M will implement a direct GeoDict interface to the repository data as GeoApp.

Starting with SOFC electrode materials, these approaches may be adapted to other fuel cell concepts, batteries, CO2-storage, composites, filter materials, and many more. Math2Market started to offer the use of convolutional neural networks (CNNs) for 3D image segmentation and object identification already in 2019. To train and use CNNs in the context of Cloud batch operation, e.g., to predict physical properties, will be revolutionary.

• ZHAW Zürcher Hochschule für Angewandte Wissenschaften, ICP Institute of Computational Physics (https://www.zhaw.ch/de/forschung/forschungsdatenbank/projektdetail/projektid/5010/)
• KaleidoSim Technologies AG

Start: October 1, 2021
End: March 31, 2024