Abstract
This paper proposes a new neural network system that differentiates the multi-pattern mixed data into multiple pattern fitting functions, and its application for structural monitoring system to detect the change of structural characteristics. All parameters dominating cause of output monitoring signals are not necessarily monitored in real monitoring system. In such case, the relationship between input and output monitoring data are seemed to be relating to multiple functions controlled by hidden parameter. The proposed neural network systems can differentiate those data into multiple pattern-fitting functions. This network system can detect the appearance of new pattern of input-output relationship of structural monitoring data caused by a damage or deterioration.
