In order to monitor the long-term condition of structures in nuclear waste disposal system and evaluate the degree of damage, it is necessary to secure quantitative monitoring, diagnosis, and prediction technology. However, at present, only simple monitoring or deterioration evaluation of the structure is being performed. Recently, there is a trend to develop monitoring systems using artificial intelligence algorithms, such as to introduce artificial intelligence-based failure diagnosis technology in nuclear power plant facilities. An artificial intelligence algorithm was applied to distinguish the noise signal and the destructive signal collected in the field. This can minimize false alarms in the monitoring system. However, it is difficult to apply artificial intelligence to industrial sites only by learning through laboratory data. Therefore, a database of noise signals and destructive signals was constructed through laboratory data, and signals effective for quantitative soundness determination of structures were separated and learned. In addition, an adaptive artificial intelligence algorithm was developed to enable additional learning and adaptive learning using field data, and its performance was verified through experiments.