An overview of fault diagnosis techniques

Generally speaking, the main objective of FDD is to convert the process observations into a few measures, indicating the anomalies of the process behavior and thereby help the plant engineer in determining the status of the plant. An intuitive approach might be limit sensing, i.e. defining thresholds for signals and raising alarms when signals cross the thresholds [96]. Despite its simplicity, limit sensing has several drawbacks, e.g. it does not consider the interaction between different components of the plant and the correlation between the observations. Moreover, setting thresholds and monitoring the measurements in a large scale process which e.g. consists of several thousands of variables is not feasible.