Model synthesis weighting effects on model tuning in system identification

System identification is the process of deriving dynamic equations from observed system behavior, the inverse of the common problem of deriving solutions to a given set of dynamics. The system identification process generally consists of two steps, a model synthesis step followed by a model tuning step. For complex systems, standard system identification tools often fail to provide satisfactory results without extensive manipulation by an experienced engineer. Input, output, and frequency weightings are often used to adjust the properties of the identified model in model tuning. In this effort, we examine the impact of model synthesis weightings on model tuning results. Model synthesis weightings are shown to improve the initial models used for model tuning. However, it is shown that an improved initial model for model tuning does not necessarily lead to faster model tuning or more accurate identified models.