The use of artificial neural networks in adiabatic curves modeling

Trtnik, Gregor (author), Kavčič, Franci (author), Turk, Goran (author)

Adiabatic hydration curves are the most suitable data for temperature calculations in concrete hardening structures. However, it is very difficult to predict the adiabatic hydration curve of an arbitrary concrete mixture. The idea of modeling adiabatic temperature rise during concrete hydration with the use of artificial neural networks was introduced in order to describe the adiabatic hydration of an arbitrary concrete mixture, depending on factors which influence the hydration process of cement in concrete. The influence of these factors was determined by our own experiments. A comparison between experimentally determined adiabatic curves and adiabatic curves, evaluated by proposed numerical model shows that artificial neural networks can be used to predict adiabatic hydration curves effectively. This model can be easily incorporated in the computer programs for prediction of the thermal fields in young concrete structures, implemented in the finite element or finite difference codes. New adiabatic hydration curves with some other initial parameters of the concrete mixture can be easily included in this model in order to expand the range of suitability of artificial neural networks to predict the adiabatic hydration curves. (C) 2008 Elsevier B.V. All rights reserved.

• Language: English
• Keywords: sveži beton, adiabatne krivulje, hidratacije betona, eksperimenti, umetne nevronske mreže
• Typology: 1.01 - Original Scientific Article
• Organization: FGG - Faculty of Civil and Geodetic Engineering
• Publisher: Elsevier
• Year: 2008
• Number of pages: Str. 10-15
• Numbering: Letn. 18, št. 1
• PID: 20.500.12556/RUL-32112
• UDC: 624.012.4:004
• ISSN on article: 0926-5805
• DOI: 10.1016/j.autcon.2008.04.001
• COBISS.SI-ID: 4044385
• Publication date in RUL: 10.07.2015

Record is a part of a journal

• Title: Automation in construction
• Shortened title: Autom. constr.
• Publisher: Elsevier
• ISSN: 0926-5805
• COBISS.SI-ID: 14103045