Fault tolerant system in a process measurement system based on the PCA method

Better process control is an important step towards increasing the efficiency of production facility. More complex control systems are introduced which require a more complex process measuring systems. Efficient process control is based upon quality and reliable process variable measurement. Process equipment failure can significantly deteriorate the product quality and even cause production outage, resulting in high additional costs. This paper analyzes automatic fault detection and identification of process measurement equipment or sensors. Different statistical methods can be used for this purpose. PCA based statistical process monitoring algorithms are applied on selected examples. For the purpose of fault detection and identification, the PCA method is used to model the correlation among process variables in the input space. Hotelling's (T
²
) and Q (SPE) statistics are used for fault detection because they provide an indication of unusual variability within and outside normal workspace. Contribution plots are used for fault identification. This paper also presents the estimation (reconstruction) of the value of faulty sensor process variable, which allows the continuation of the process, although the fault might have occurred. Results of all considered examples are compared and discussed.