Applying the machine learning method to improve calibration quality of TDR measuring technique
Artykuł w czasopiśmie
MNiSW
100
Lista 2024
| Status: | |
| Autorzy: | Mikušová Dominika, Suchorab Zbigniew, Paśnikowska-Łukaszuk Magdalena, Zaburko Jacek, Trník Anton |
| Dyscypliny: | |
| Aby zobaczyć szczegóły należy się zalogować. | |
| Rok wydania: | 2024 |
| Wersja dokumentu: | Elektroniczna |
| Język: | angielski |
| Numer czasopisma: | 3 |
| Wolumen/Tom: | 18 |
| Bazy: | BazTech |
| Efekt badań statutowych | NIE |
| Materiał konferencyjny: | NIE |
| Publikacja OA: | TAK |
| Licencja: | |
| Sposób udostępnienia: | Otwarte czasopismo |
| Wersja tekstu: | Ostateczna wersja opublikowana |
| Czas opublikowania: | W momencie opublikowania |
| Data opublikowania w OA: | 12 kwietnia 2024 |
| Abstrakty: | angielski |
| The article presents the application of the time domain reflectometry (TDR) technique for measuring the moisture of porous building materials used in construction. The work is focused on using the potential of artificial intelligence to improve the quality of TDR measurements through a new approach to the interpretation of data obtained from the TDR readings. Machine learning is a data analysis technique, used nowadays in many scientific disciplines. The authors performed a measurement data analysis using the artificial intelligence algorithms to assess moisture of aerated concrete samples tested with a TDR multimeter using two non-invasive sensors which differ in thickness. Data analysis was carried out using supervised machine learning to analyse a series of reflectograms obtained during the measurement. For the data achieved by the classical and machine learning method interpretation, correlation analysis was conducted to confirm the potential of artificial intelligence to improve the quality of TDR measurement. The summary of the work discusses the obtained analytical results and highlights the effectiveness of moisture assessment using the Gaussian Process Regression method, which allowed achieving a level of 0.2 - 0.3% of the RMSE errors value, which is about 10 times lower than the traditional approach. |
