OpenAE in action: From open data sets to shareable machine learning models

Data-based analysis of Acoustic Emission (AE) data typically relies on features extracted from the time and frequency domains. These features form the input vector for training machine learning models. For a trained model to be used by other parties or in different environments, it is essential that the input vector can be reproduced accurately. This requires the feature extraction computations to be consistent and standardized. When this standardization is achieved, models can be shared, reproduced, and deployed seamlessly across various applications and environments. OpenAE is an initiative to address these challenges by establishing open standards for data-based AE analysis. It is designed as a collaborative platform based solely on GitHub repositories for easy contribution. It provides standardized definitions and implementations of feature extraction algorithms as well as open data sets. In the future, OpenAE should also serve as a platform for publishing and sharing trained models, further promoting transparency and collaboration. This talk will showcase an end-to-end application of OpenAE, starting with an open dataset of AE data, which will be published on the OpenAE platform. We will demonstrate how OpenAE’s standardized feature extraction algorithms transform raw AE data into meaningful features. Using these features, a machine learning model is trained that can be easily reused and deployed across different environments.