Identification of non-uniformly sampled multirate systems with application to process data compression

The bizarre properties of non-uniformly sampled multirate data prevent the data from being directly adopted in conventional identification methods. Here, a novel numeric integration method is proposed, which can conveniently estimate all orders of definite integration of the non-uniformly sampled data. A continuous-time identification method based on subspace estimation is established on the integral data. Furthermore, this identification method is applied to the data compression in industrial processes, where the compressed data by conventional swinging door trending method are twice compressed by modelling through identification. The model parameters are stored instead of the identification outputs. Numeric results on the data from Monte Carlo simulation as well as a station of the long-distance gas pipeline are presented and analysed, showing the feasibility of the identification method and higher compression ratio achieved by the new data compression scheme.