A new fault detection and diagnosis approach for a distillation column based on a combined PCA and ANFIS scheme

In this paper, a new approach is introduced for fault detection and diagnostic. The method uses integration of PCA (Principal Component Analysis) and ANFIS (Adaptive Neuro -Fuzzy Inference System). PCA is employed to reduce the recorded data dimension and yet extract informative features for fault detection purpose. The reduced data is then fed to ANFIS to discriminate the occurred fault. Resolution of the multiple ANFISs is enhanced through adequate selection of the utilized membership function (MF) numbers to compensate for the large number of possible created rules. This approach naturally removes extra pressure on each ANFIS to yield good responses only on close neighborhood of faulty data in training process. The combination of boundary models in the extra number of MFs provides fault isolation of the faulty plant section even when novel faults. The key point of this approach is the ability to detect and diagnose any novel fault with the same time-response pattern but different severities. The efficacy of the proposed FDD approach has been demonstrated via extensive conducted tests in a distillation column benchmark.