Multichannel Processing of Noisy Vibration Responses: A Constrained Subspace Approach

In this paper we discuss multiple input/multiple output (MIMO) constrained subspace identification techniques that will be evaluated when applied to the fundamental problem of identification of a dynamic structural model from noisy vibrational response measurements. The structure is a cylindrical copper pipe consisting of three (3) accelerometers positioned along its length as well as a single transducer measuring the excitation function (chirp or white noise) totaling ten (10) sensor channels (9 outputs and a single input). Here we discuss a suite of "constrained" subspace identification techniques and evaluate their performance on experimental data.