Czech digital mathematics library · 1984 · 28 citations · 3 references
Open access
EngineeringSystems EngineeringDynamical SystemsModel-based Systems EngineeringSystems ModelingBayesian FilteringBayesian InferenceBayesian ModelingBayesian StatisticsPrior SpecificationSystem IdentificationBayesian System IdentificationEliciting Prior KnowledgeIdentification Of Dynamical SystemsLinear System IdentificationPulse Response AnalysisPulse Response DynamicsState IdentificationSystem Identification In The Time Domain
Bayesian approach to system identification requires from the user to collect and express his prior information about the identified system.The paper presents a way in which a global prior knowledge can be quantified.Two simple cases, pulse response smoothness and knowledge of the static gain of a system modelled by mixed autoregressive-regressive model, are elaborated in detail.These cases which are of practical importance illustrate main steps in the quantification.The significant and favourable effect of this built-in prior information on the start of selftuning control is demonstrated on an example.
3