Data based modeling for PEM fuel cell monitoring — a test bench application, Part II

In part II of this paper two important methods for the monitoring procedure of the fuel cell (part I) are presented. First, a time delay estimation method, which allows a better data based modeling of the underlying fuel cell system. Second, the sequential probability ratio test (SPRT) which improves the fault detection results concerning over-detection and reliability. For the time delay estimation method using mutual information a minimum search algorithm is discussed, because the easy approach of a global minimum does not always guarantee a real time delay. The challenge for the SPR test is the design of a probability density function without having prior knowledge about the process.