Over the past century, infrastructure has undergone rapid progress, particularly in the construction of reinforced concrete structures. Consequently, health monitoring plays a significant role in assessing structural conditions. Visual inspection and tapping technique (hammering test) became widely trendy due to its simplicity and economy. Nonetheless, several challenges exist, especially in the difficulty of interpreting test results. Hammering tests can determine the size or location of material discontinuities. However, when the structure becomes more diverse and complex, such as concrete structures with steel bars inside, an acoustic wave of tapping sound by hammering test produces chaotic reflections and resonances. This study endeavors to conduct research and development on the evaluation of results in tapping sound by hammering methods. The objective is to utilize Artificial Intelligence (AI) to learn and diagnose the data, thereby compensating for the shortage of experts and minimizing errors. Reinforced concrete beams 2 m in length with a square cross-section of 0.2 m were subjected to electrochemical acceleration technique to induce corrosion of steel bars. Unsupervised deep learning was implemented to learn the complex dimensional hammering test results based on various perspectives of characteristic acoustic wave sounds. The results demonstrate that in comparison to the intact stage, this technique effectively provides an index of abnormality in echo wave patterns correlated with crack propagation and reinforcement damage levels.