STRUCTURAL IDENTIFIABILITY OF BOND GRAPH MODELS

The global identifiability is a structural property of models, which associates a unique set of parameters with given input/output response. The translation of this property into bond graph modelling language allows the combination of the physically meaningful language of bond graph methodology and the numerical accuracy of identified transfer function models. Based on the building mechanisms of a transfer function from a bond graph model, the paper develops and explains why a bond graph can be not identifiable. Both internal and input/output dynamics can be written with the Mason's rule, using causal loops and action chains. Then the way the combination of causal loops and action chains influences the identifiability of models is discussed. As a result a criterion is given, which decides whether a bond graph model is structurally globally identifiable or not. This is a crucial issue in order to guarantee the reliability of identification processes