Predicting Concrete Conductivity with Neural Networks

This study demonstrates that integrating Generative Adversarial Networks (GANs) for data augmentation significantly enhances the predictive accuracy of machine learning models in construction materials science. By using a conditional GAN to generate synthetic data, researchers overcame the common challenge of limited training datasets, allowing an Artificial Neural Network to more reliably predict the thermal conductivity of various concrete types. This approach proved more effective than training on real data alone, offering a robust solution for scenarios where experimental data is scarce or expensive to obtain. The relevance to open data lies in the methodological framework for handling data scarcity through synthetic generation. Open data initiatives in engineering and materials science often face limitations regarding the volume and diversity of available datasets. This research highlights how synthetic data can mimic the statistical properties of real-world observations, thereby enriching public repositories and enabling more sophisticated analysis without requiring additional physical experiments. It suggests a pathway for enhancing the utility of open data by artificially expanding dataset size and variability while maintaining structural integrity. Ultimately, this innovation supports broader goals of sustainability and energy efficiency in building design. Accurate prediction models enable engineers to select optimal material compositions, reducing trial-and-error experimentation and conserving resources. By validating the effectiveness of GAN-augmented models, the study encourages the adoption of advanced computational techniques within open data ecosystems, fostering more efficient and data-driven advancements in sustainable construction practices.

Source: azobuild.com
Published on 2024-09-05

Por proyectos del RIGI, la Casa Rosada espera la llegada de más de US$17.000 millones »

Open Data

Related news