Fault Detection Of Sample System Based On Parity Space

With the development of computer science and the automation technology, more and more sample control systems are widely applied in the areas such as electric power system, petrochemical et al. These control systems are very large and expensive, once faults occur, the results are catastrophic. The high security and reliability of fault detections are required. This paper mainly discussed the fault detection based on parity space in sample control system:The three main fault diagnosis methods, namely analytical redundancy model-based methods, knowledge-based methods and data-driven technique, were illustrated separately. The cruxes and the latest states of art of fault diagnosis were addressed.The research was done on the characteristics of analytical redundancy parity space approach to fault detection in terms of observability indexes, including the effect of different selections of parity vector on fault detection, the relationship between the parity relation and the dimension of parity space, the robustness of fault detection et al. The above theoretical analyses were proved through the simulation of a two water order control system.For fault detection of single rate sample control systems, firstly, the direct fault detection method was adopted through the numerical operator; secondly, principals of the choice of sample period under disturbance, the conditions of perfect decoupling were studyed respectively. Simulations proved that the direct approach possesses better fault detectability than indirect one, but the selection of sample period had obvious affection on effectiveness of direct fault detection.Fault detection of multirate output control system was studied. Firstly, familiar with single rate sample control system, the numerical operator was used to design direct fault detection system. Secondly, the lifting technique was introduced to realize the refection from continuous time systems to sample systems, which transforms the MIMO control system with different sample period into single sample rate multivariate problems where the sample period was equal to the inputs, so the direct fault detection systems was achived accordingly. At last, the effectiveness of two methods was elaborated by simulations, and a concise comparison was concluded.