ABSTRACT
A reversed cyclic loading test was conducted on a RC column specimen to examine the effectiveness of vibration measuring and machine learning for damage evaluation of RC columns. In the loading test, 11 acceleration sensors were placed on the surface of the column to obtain frequency responses at concrete cracking and spalling of the cover concrete. It was determined that an unsupervised learning model can evaluate the damage levels and locations, based on intact data measured before the loading test. Moreover, a following supervised learning method indicated the possibility of creating crack images on RC columns, from frequency responses measured with sensors.
