Automated Air Void Parameter Evaluation in Hardened Concrete using Confocal Laser Scanning Microscopy and Deep Learning
Dokumenttyp:
Konferenzbeitrag
Autor(en):
Kostic, Viktor; Kotsev, Viktor; Khan, Qadeer; Cremers, Daniel; Timothy, Jithender; Kränkel, Thomas; Gehlen, Christoph
Abstract:
The current standardized method to measure air void parameters to assess the freeze-thaw resistance of concrete is time-consuming and prone to uncertainty due to the manual counting and measuring of air void chords. This research presents a new advanced approach by automated pairing of data from confocal laser scanning microscopy with the U-net model, reliably predicting the locations of air voids, aggregates and cement paste within image data without extensive specimen preparation. Automated air void chord analysis was conducted and compared to the human results of a round-robin test, yielding promising results of evaluating micro air void (<300 μm) content that are in good agreement with manual human testing methods.
Stichworte:
GNI
Herausgeber:
Briffaut, Matthieu; Torrenti, \Jean Michel\
Kongress- / Buchtitel:
Proceedings of the 2025 fib International Symposium - Concrete Structures