[V-549] A feasibility study on application of machine-learning to impact-echo inspection method
Keywords:impact-echo inspection method, non-destractive inspection, spectrogram, machine learning
In Japan, the importance of maintainance and management of infrastructure is increasing. Impact-echo method is put to practical use as a relatively simple degradation diagnosis method for concrete structures, but its accuracy depends on the skill of the diagnostician, and the accuracy is also limited. Here, the feasibility of the method of estimating the state of the defect with machine learning of the spectrogram image of the impact sound was examined by an indoor test. As a result, it was shown that according to this method, it is possible to accurately estimate the condition of the defect in concrete without depending on the skill of the diagnostician.
Abstract password authentication.
Password is required to view the abstract. Please enter a password to authenticate.