Compatibility Review of Publicly Available Experimental Data on Common Building Material Temperature Dependent Properties for use with ML Algorithms
International Conference ReShaping the Built Environment through Sustainability and Circularity, ReSBE 2026, Thessaloniki, Yunanistan, 20 - 23 Ekim 2026, cilt.905 LNCE, ss.1424-1431, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Cilt numarası: 905 LNCE
- Doi Numarası: 10.1007/978-3-032-34930-9_166
- Basıldığı Şehir: Thessaloniki
- Basıldığı Ülke: Yunanistan
- Sayfa Sayıları: ss.1424-1431
- Anahtar Kelimeler: building materials, data analysis, information circularity, machine learning
- Açık Arşiv Koleksiyonu: AVESİS Açık Erişim Koleksiyonu
- Kocaeli Üniversitesi Adresli: Evet
Özet
The analysis and predictive capabilities of Machine Learning (ML) algorithms are becoming an increasingly invaluable asset in the research domains of heat transfer and building materials. Precision and high performance of such models are heavily reliant on the availability of substantial amounts of data for training and testing purposes. This study concentrates on reviewing available sources of experimental data and assesses compatibility implications with ML algorithm requirements. Focus is given on common building materials with temperature-dependent properties assessed against elevated temperatures or fire scenarios. The structure and consistency of the available information are discussed and compared with the fundamental training data requirements of ML algorithms, focusing on Artificial Neural Networks. The study concludes with a brief discussion on the financial and environmental benefits of data and material circularity and proposes methods of reusing available experimental results in cutting-edge Artificial Intelligence applications.