HiPerClust Technology for Automated and Scalable Atom Probe Tomography (APT) Data Analysis

Summary

HiPerClust is a machine learning–powered software platform designed to automate and accelerate the analysis of atom probe tomography (APT) data. By leveraging transfer learning, the software identifies patterns and optimizes clustering parameters without manual intervention. Trained on synthetic datasets, HiPerClust predicts the optimal number of clusters and parameter settings for a given dataset—eliminating time-consuming trial-and-error. Integrated with high-performance computing (HPC) systems, it enables scalable, reproducible, and efficient processing of large APT datasets, making high-resolution materials analysis faster and more reliable.

Solution

HiPerClust automates the APT clustering process by using transfer learning models trained on synthetic data to predict the most effective clustering parameters—such as minimum cluster size and number of points for algorithms like HDBSCAN. The tool’s integration with HPC environments allows researchers to test, validate, and apply these predictions rapidly across large datasets, improving both accuracy and speed.

Key Advantages

• Automated Parameter Optimization: Predicts optimal clustering parameters, removing manual tuning and human variability.
• Reproducible Analysis: Ensures consistent clustering outcomes under the same conditions.
• High-Performance Scalability: Optimized for HPC systems, enabling rapid processing of massive datasets.
• Transfer Learning Efficiency: Applies trained models to new materials without retraining, accelerating deployment.
• Reduced Research Time: Frees scientists from repetitive computational tasks, allowing focus on interpretation and discovery.

Market Applications

• Materials Science Research: Automated APT data clustering for alloy, semiconductor, and nanomaterial development.
• Nuclear Materials Analysis: Reliable clustering of nuclear fuel and structural materials for degradation studies.
• Advanced Manufacturing: Rapid assessment of microstructural properties for additive manufacturing and advanced alloys.
• Research Institutions and National Labs: High-throughput, reproducible data analysis for collaborative materials programs.

Access

This software is available as open-source and can be accessed through the software's GitHub Repository.