An Identification Technique for Linear Systems: Application on a Hydraulic Testbed

Practitioner engineers in both academic and industrial areas, are often faced with the challenge of identifying the model of a given system or process in order to setup a controller or to extract some useful information. Among the existing identification algorithms, those being numerically simple and stable are more attractive for practitioners. This paper deals with identification of state-space models, i.e., the state space matrices A, B, C and D for multivariable dynamic systems directly from test data (data-driven). In order to guarantee numerical reliability and modest computational complexity compared with other identification techniques, in this paper, we propose a synergistic identification technique based on the principal components analysis (PCA) and subspace identification method (SIM) under white noise assumptions. The proposed technique identifies the parity space - PS (or null space) from input/output data, and from there, the matrices related to the system through the extended observability matrix and a block triangular Toeplitz matrix. In order to show its capability, the proposed identification technique is applied to an academic test bed that is related to a hydraulic process.