Fault detection in nonlinear systems: an observer-based approach

摘要: An un-permitted deviation of at least one characteristic property or parameter a system from standard condition is referred as fault. Faults result in reduced efficiency the system, quality product, and sometimes complete breakdown process. This not only causes economic losses but may also fatalities. early detection faults can assist to avert these losses. Therefore, fault process monitoring becoming an essential part modern control systems. Fault linear dynamical systems has been extensively studied well established techniques exist literature. However, for nonlinear yet active field research. work motivated by fact that most real are nature there need develop Observer-based methods have proven be among capable approaches, therefore, this research focused towards methods. The first step observer-based generate symptom signal, called residual which carries information faults. done comparing measurements their estimates generated observer (filter). It desired signal sensitive robust against disturbances. presents new designing (filter) signal which Three types filters proposed dissertation; include filter, disturbance attenuating filter achieve simultaneous attenuation disturbances amplification faults. Despite filters, completely decoupled effect uncertainties. threshold needed care Selection plays important role performance system. If it selected too high, some will detected. Conversely, if low, uncertainties false alarms. method determine avoid false-alarms minimize missed-detections. A generator itself dynamic produces variable threshold. changes with effects fits more tightly fault-free and, hence, improved. In addition generation stage, optimized post-filtering. further contribution proposing post-filter operates on modified signal. simultaneously Together post-filter, strategy adopted select maximizes detectability minimizes number false-alarms.