2020 年 106 巻 2 号 p. 80-90
The equipment and the structures in steel works are used in a very long term. Therefore, the inspections and the maintenance are indispensable. To make this more effective, much effort has been devoted to develop methods of fault detection and diagnosis. However, the effectiveness of those methods may be limited due to the tradeoff between false positive and false negative reactions. To mitigate the tradeoff, this paper considers to use model sets involving parameters and disturbances that are expected to change. More specifically, we first propose a method to estimate the parameters and the disturbance so as to minimize the deviation between the real output and the output generated by the model sets. The resulting residual enables us to conduct the health diagnosis. The effectiveness of the proposed framework is examined by using numerical examples.