Fault diagnosis in industrial processes via parallel processing

The model-based fault detection/isolation (FDI) methods are used in many areas of industry. In most of the applications the systems are distributed spatially and temporarily. As a result of this condition the systems can be divided into subsystems. They are described by a subset of signals. These subsets of signals can be used in state estimator schemes to detect and evaluate the faults. When using this approach the parity relations and the observer Kalman filters can be realised by different processes in parallel processing. Partitioning the task of FDI and allocation of these tasks to transputers help to satisfy real-time requirements. The partitioning and task allocation methods must be examined by MATLAB and EPICAS software tools to define the efficiency of this approach.