Natural Hazards Research Summit 2022: Efficient supercomputer resource utilization for machine learning: a case study on identifying damage in a five-story reinforced concrete structure

In this poster, presented at the NHERI Natural Hazards Summit 2022, we present an efficient approach for high-performance computing (HPC) resource utilization for machine learning (ML). HPC resources are applied to integrate the learning capacity of ML-based approaches with experimentally-measured data from a small model to estimate damage to a reinforced concrete building using transfer learning. Benefits of transfer learning include saving training time, improving performance, and reducing required training data. The data in this project can be reused by researchers who want to apply ML-based approaches for damage identification of in situ structures or who want to learn about HPC resources and their efficient use for ML. This project is unique because it uses experimental data from a 5-story small model to perform transfer learning for damage identification to a 5-story full-scale reinforced concrete structure. The audience for this project includes researchers and practitioners in the fields of HPC, ML, and structural engineering who are interested in developing ML-based solutions for damage identification of in situ structures.