Robust global identifiability theory using potentials--Application to compartmental models.
Math Biosci
; 262: 182-97, 2015 Apr.
Article
em En
| MEDLINE
| ID: mdl-25660327
ABSTRACT
This paper presents a global practical identifiability theory for analyzing and identifying linear and nonlinear compartmental models. The compartmental system is prolonged onto the potential jet space to formulate a set of input-output equations that are integrals in terms of the measured data, which allows for robust identification of parameters without requiring any simulation of the model differential equations. Two classes of linear and non-linear compartmental models are considered. The theory is first applied to analyze the linear nitrous oxide (N2O) uptake model. The fitting accuracy of the identified models from differential jet space and potential jet space identifiability theories is compared with a realistic noise level of 3% which is derived from sensor noise data in the literature. The potential jet space approach gave a match that was well within the coefficient of variation. The differential jet space formulation was unstable and not suitable for parameter identification. The proposed theory is then applied to a nonlinear immunological model for mastitis in cows. In addition, the model formulation is extended to include an iterative method which allows initial conditions to be accurately identified. With up to 10% noise, the potential jet space theory predicts the normalized population concentration infected with pathogens, to within 9% of the true curve.
Palavras-chave
Texto completo:
1
Coleções:
01-internacional
Base de dados:
MEDLINE
Assunto principal:
Modelos Lineares
/
Dinâmica não Linear
/
Modelos Biológicos
Tipo de estudo:
Prognostic_studies
Limite:
Animals
/
Female
/
Humans
Idioma:
En
Revista:
Math Biosci
Ano de publicação:
2015
Tipo de documento:
Article