Advances in fault diagnosis using analytical redundancy

This paper provides an overview of the subject of model-based fault diagnosis (fault detection and isolation) using analytical redundancy. Most of the recent advances and developments in this subject have been outlined, with a particular emphasis on robust design problems. The main principle is to use a residual, a symptoms of process faults to facilitate the diagnosis tasks. The residual is generated using the mathematical model of supervised process. A crucial goal of the residual generator design is to make the residual insensitive to modelling errors, whilst sensitive to faults. The best solution to this problem is to combine together disturbance decoupling design and the adaptive threshold decision-making. The methods outlined are very well suited to process control systems, particularly when robustness techniques are included