Akademik Veri Yönetim Sistemi
4th International Civil Engineering & Architecture Conference (ICEARC'25), Trabzon, Türkiye, 17 - 19 Mayıs 2025, ss.2048-2054, (Tam Metin Bildiri)
This study estimated the first frequency values of masonry stone arches
subjected to high-temperature effects by classical regression analysis (CRA).
Within the scope of the study, 450 heat transfer and 5850 modal analysis
results of masonry stone arches with different dimensions and temperature
histories were used. The first mode frequencies of the arch members were
obtained with 6300 finite element analysis. Then, the first mode frequencies of
the arch members were estimated by CRA using the geometrical parameters of the
arch member (span, height, width, thickness) and the temperature history
condition (0, 15, 30, 45, 60, 75, 90, 105, 120, 135, 150, 165, 180 min) to
which the arch members were exposed. In the CRA method, four different
functions, namely linear function (LF), power function (PF), exponential
function (EF), and quadratic function (QF), were applied, and the coefficients
of these functions were calculated. While developing the prediction models, %70
of the data was used for training and the remaining %30 for testing the models.
Various performance statistics (square root of mean squared error, mean
absolute error, scatter index, and Nash Sutcliffe (NS) efficiency coefficient)
were used to evaluate the prediction performance of the developed models. The
NS performances of the models developed using LF, PF, EF, and QF functions in
the CRA method are calculated as 0.448, 0.923, 0.975, and 0.772 for the
training data set and 0.449, 0.922, 0.975, and 0.722 for the test data set,
respectively. When the performance statistics obtained are compared, it is seen
that the most successful prediction results are obtained from EF. The
performance statistics obtained from the modeling studies showed that the first
mode frequency values of masonry stone arches with different dimensions and
temperature histories could quickly be predicted with high performance, with
less model and cost.