Detection and estimation of randomly occurring deterministic disturbances

The term randomly occurring deterministic disturbances refers to a class of process disturbances that occur randomly and infrequently in time and have a known effect on the behavior of the process. Since these disturbances occur infrequently in time, traditional filtering methods which assume identically distributed noise terms may not yield acceptable performance. The reason for this is that, when tuning the filter, a compromise must be made between sensitivity to noise and the ability to track these disturbances when they do occur. The stochastic model of these disturbances leads to a multi-filter approach for the state estimation. Since the number of filters grows exponentially with the data length, a suboptimal algorithm is required. The performance of this approach is evaluated for state/parameter estimation of a continuous polymerization reactor