Investigation of Nonlinear System Identification Techniques

Note: This study is part of a broader project “Riser Interaction Model: A Combined T/F Domain Model” (MMS Project 360)

This research study builds upon the reverse multiple input/single output identification technique as presently reported in the open literature. The reverse system identification technique, which is able to extract information about the key system parameters from excitation/response data, is modified to deal with realistic structural systems.

First, the identification technique is extended to address distributed-parameter multi-degree-of-freedom systems with general damping-restoring types of nonlinearities subject to random excitation. Generalized coordinates and force function are used correspondingly as inputs and output of the reversed system. The value of the system parameters can be recovered for each mode of vibration. Analytical expressions as functions of the modes are derived and numerical simulations of a marine riser are used to explore the adequacy of the methodology and the benefits of using modal analysis in the system identification procedure.

The second adaptation of the system identification technique is devised to illustrate the technique’s inherent ability to identify the potential frequency dependence of certain parameters. The data analysis of a compliant deepwater offshore platform, which is modeled as a nonlinear multi-degree-of-freedom system, makes no preemptive assumption about the form of the system parameters. The system identification results for the frequency-dependent hydrodynamic added-mass and damping coefficients are compared with those obtained using an industry standard radiation-diffraction software package.

Finally, the method used in the second application to recover independently the value of parameters associated with fully correlated inputs is explained in details and an attempt is made to explain in simple terms the theoretical justification establishing the validity of the methodology.

Collectively, the improvements reported in this investigation expand significantly the range of applications of the reverse multiple input/single output identification technique. The results presented demonstrate the methodology as modified in this study yields accurate values of the identified system parameters.