Functionally Pooled models for the global identification of stochastic systems under different pseudo-static operating conditions

The problem of identifying a single global model for stochastic dynamical systems operating under different conditions is considered within a novel Functionally Pooled (FP) identification framework. Within it a specific value of a measurable scheduling variable characterizes each operating condition that has pseudo-static effects on the dynamics. The FP framework incorporates parsimonious FP models capable of fully accounting for cross correlations among the operating conditions, functional pooling for the simultaneous treatment of all data records, and statistically optimal estimation. Unlike seemingly related Linear Parameter Varying (LPV) model identification leading to suboptimal accuracy in this context, the postulated FP model estimators are shown to achieve optimal statistical accuracy. An application case study based on a simulated railway vehicle under various mass loading conditions serves to illustrate the high achievable accuracy of FP modelling and the improvements over local models employed within LPV-type identification.