Evolutionary Parametric Identification of Dynamic Systems

In this paper a new method for the estimation of SISO ARMAX models was presented. The proposed methodology lies in the context of Evolutionary system identification. It consists of a hybrid optimization algorithm, which interconnects the advantages of its deterministic and stochastic components, providing superior performance in PEM, as well as a two-stage estimation procedure, which yields only stable models. The method's main characteristics can be summarized as follows: ? improvement of PEM is implemented through the use of a hybrid optimization algorithm, ? initial ``guess'' is not necessary for good performance, ? convergence in local minima is avoided, ? computational complexity is sufficiently decreased, compared to similar methods for Evolutionary system identification. Furthermore, the method has competitive convergence rate to conventional gradient-based techniques, ? stability is guaranteed in the resulted models. The unstable ones are penalized through the objective function, ? it is successive, even in the presence of noise-corrupted measurement. The encouraging results suggest further research in the field of Evolutionary system identification. Specifically, efforts to design more flexible constraints are taking place, while the implementation of the method to Multiple Input-Multiple Output structures is also a topic of current research. Furthermore, the extraction of system's valid modal characteristics (natural frequencies, damping ratios), by means of the proposed methodology, is an additive problem of crucial importance. Evolutionary system identification is an growing scientific domain and presents an ongoing impact in the modelling of dynamic systems. Yet, many issues have to be taken under consideration, while the knowledge of classical system identification techniques and, additionally, signal processing and statistics methods, is necessary. Besides, system identification is a problem-specific modelling methodology, and any possible knowledge o