Diagnosis Algorithm of Immunity-Based System with Continuous Variable of Mutual Test Results among Sensors

The fault diagnosis for high availability system is required that it detects failure points as quick as possible. Moreover, it can realize the much higher availability, if it detects the point that is more likely to break down while the system operates normally. To meet these requirements, we propose a new algorithm for the immunity-based system diagnosis, which can detect unknown fault modes. The immunity-based system diagnosis has mutual test among sensors in the system. Proposed algorithm has continuous variables of its mutual test results among sensors. The variables are calculated from the deviation from typical relationship among output of sensors. Our experimental results demonstrate that proposed algorithm needs much less number of times of sampling than conventional algorithm for fault detection, and it can detect the sensor that has the largest error in the system while system operates normally.