Abstract
Fault detection is essential for monitoring of various biological processes and becomes even more important in that context. This paper, therefore, presents an enhanced tool that merges state estimation with fault detection (FD) methods to improve monitoring of biological processes. The proposed technique, so-called particle filter (PF)-based maximum double adaptive exponential weighted moving average (EWMA) chart, involves two steps. First, the states of the biological processes are estimated using the PF method. In the second step, the faults are detected using the maximum double adaptive EWMA chart. The proposed method is based on the maximum of the absolute values of the EWMA statistics, one monitoring adaptively the variance and the other controlling the mean. The FD performance is studied utilising a wastewater treatment model. The detection performances are assessed in terms of missed detection rate, false alarm rate, detection speed, sensibility to fault sizes and robustness to noise levels.