An input signal, such as M series signal, used in the least square method is required to satisfy P.E. condition, which often results in resonance, and consequently large noise and vibration in a mechanical object, as in an oil hydraulic system. On the other hand, a step signal is practical because it causes less noise and vibration. Conventional off-line identification methods restrict the order of an identifiable model if a step signal is used for it.
In this paper, we propose a new off-line identification method, which can be applied to an arbitrary order object and performed with data generating constant steady state output response. The feature of the method is that a reduced order model can be obtained directly in the way that the intermediate parameters, from which the system parameters (coefficients of rational transfer function) are derived, are determined uniquely irrelevant to the unmodelled dynamics, so that 0∼
N-tuple integral values of the output error lead to zero.
Numerical simulations indicate the effectiveness of the proposed method.
View full abstract