Hybrid frequency–time domain models for dynamic response analysis of marine structures
Introduction
Dynamic response of marine structures to wave excitation is commonly analyzed in the frequency domain by a first-order potential theory approach and by assuming the wave process as Gaussian, see for instance Newman (1977) and Faltinsen, 1990, Faltinsen, 2005. The response statistics is then obtained using the well-established theory for Gaussian processes. This approach is based on linear theory, which implies that the wave steepness is small and also that the response due to wave excitation is proportional to the wave amplitude (Faltinsen, 2005). The frequency response functions of the structure can be obtained by using the standard hydrodynamic codes (e.g. WAMIT, 2006). This method is particularly efficient since for each frequency considered, the motion response is obtained by solving a set of simultaneous linear equations with constant coefficients.
When nonlinear effects such as viscous forces, water entry and exit are considered, however, the linearity assumption is no longer valid. One approach to overcome the difficulties of full nonlinear time-domain analysis is to apply a higher order frequency-domain approach, e.g. by using simplified bilinear and trilinear frequency response functions based on Volterra functional representations (Bendat, 1998). However, frequency-domain approaches are limited to steady-state processes. Besides this fact, higher order frequency-domain methods can be cumbersome to implement and computationally inefficient.
A different approach consists of using a linear time-domain model based on the Cummins equation which will be referred to as a hybrid frequency–time domain model throughout this paper. The resulting linear model is a vector integro-differential equation which involves convolution terms. Nonlinear effects can be introduced to this model at a later stage. In this regards, Wu and Moan (1996) introduced a hybrid frequency time domain approach by first solving the linear problem in the frequency domain and then transforming the input–output into time domain and accounting for nonlinear effects as “additional” loads.
The convolution terms in a time-domain model are not convenient for analysis and design of motion control systems (Fossen, 2002, Perez, 2002). Moreover, time-domain simulations of linear transient or nonlinear problems with convolution terms are computationally demanding and their implementation in standard simulation packages is inconvenient (Kashiwagi, 2004). For these reasons, different methods have been proposed as approximate alternative representations of the convolutions. Because the convolution is a linear operation, different approaches can be followed to obtain an approximately equivalent linear system in the form of either transfer function and state-space models. This process involves the use of system identification, and several options are available depending on the way in which the identification problem is posed. This raises the question whether one method is better than the others.
Hence, the main purpose of this paper is to investigate efficient alternative formulations of the dynamic equations of motion that avoid convolution terms. 2 The Cummins equation and the frequency response model, 3 Relation between time and frequency domain models, 4 Alternative representations of the radiation force of this paper deal with description of the Cummins equation and the convolution integrals. Different representations of the linear systems, with focus on the state-space models, is the subject of Section 5. Convolution replacement alternatives are investigated in Section 6. Sections 7 and 8 deal with different identification approaches that are commonly used to obtain the state-space models. Application and discussion regarding the use of these methods for a simple memory function is carried out in Section 9. Dynamic response simulations for a container vessel is performed in Section 10 by using state space models. Finally, comments and conclusions are given in Section 11.
Section snippets
The Cummins equation and the frequency response model
Cummins (1962) considered the behavior of the fluid and the structure in time domain. He assumed linear behavior and considered impulses in the components of motion. This resulted in a boundary value problem in which the potential was separated into two parts: one valid during the duration of the impulses and the other valid afterwards. By expressing the pressure as a function of these potentials and integrating it over the wetted surface of the vessel, he obtained a vector integro-differential
Ogilvie relations
The relationship between the parameters of Eq. (1) and those of Eq. (2) were established by Ogilvie (1964):from which it follows thatandUsing these relations, Eq. (1) can be written asA related form of the above by Wehausen (1967) iswhereThe velocity formulation of Eq. (10) is more convenient for numerical
Alternative representations of the radiation force
Within the scope of fluid memory effects, the identification of state-space models can be posed for not only the retardation functions, but also for the other representatives of the radiation force. Therefore, we devote this section to clarify these alternatives before explaining different convolution replacement approaches in Section 6.
The total hydrodynamic radiation force vector in the frequency domain can be expressed asReplacing the acceleration vector by
State-space models for linear systems
Before proceeding with different representations of the convolution terms and of the Cummins equation (Eq. (10)), linear system representations are briefly revisited. This will then be used throughout the rest of the paper.
Convolution replacement
Time-domain simulations of the dynamic response of marine structures by direct evaluation of the convolution integrals may be computationally demanding depending on the time step, simulation length and degrees of freedom of the model (Holappa and Falzarano, 1999, Kashiwagi, 2004). The state-space representation provides an attractive alternative for simulation due to the simple form of its solution i.e. Eq. (32). For zero input, , this indicates that dynamic variables, in this
Identification methods for convolution replacement
Fig. 1 shows the components defining an identification method in the context of this paper. Every identification method consists of selecting a series of data, choosing a model structure (model type and order) and a fitting criterion. The unknown parameters in the prospective model are obtained by a parameter estimation method and using the specified fitting criteria (Ljung, 1999). The objective of system identification is to obtain the lowest order model possible that is able to reproduce the
Parameter estimation methods
In this section, we will describe three methods by which the unknown model parameters resulting from the identification methods are obtained:
- (i)
Impulse response curve fitting.
- (ii)
Realization theory.
- (iii)
Regression in the frequency domain.
Application example
Herein, we consider a simple retardation function as an example to compare the performance of the different identification methods. The example is based on pure analytical relations and satisfies all the properties of the convolution terms given in Section 3. It is standard practice in system identification to compare different estimators with scenarios for which the values of the true quantities to be estimated are known and this is the motivation for this example. We should mention, however,
Specification of the case
Numerical studies are performed in this section for a container vessel at zero forward speed. The specifications of the ship are given in Table 1. Fig. 6 shows the body plan of the hull.
Two different scenarios are analyzed:
- (i)
The objective of the first one is to examine the steady-state part of the time-domain response of the container vessel by using two different state-space models. The focus is on the models obtained by identifying the retardation functions based on the realization theory and
Concluding remarks
In this paper, different alternatives to replace the convolution terms in the Cummins equation have been reviewed. By this way, the integro-differential equations of motion have been written in the form of differential formulations. The emphasis has been on the methods that replace the convolutions by the state-space models. It was discussed how to obtain the replacing models via system identification techniques based on the hydrodynamic data from the standard software. Identification methods
Acknowledgements
We are grateful to Professor Johannes Falnes for his invaluable comments. We also thank Mr. Ingo Drummen at Centre for Ships and Ocean Structures for providing the data of the container ship. The authors appreciate the financial support of the Norwegian Research Council which has been granted through Centre for Ships and Ocean Structures.
References (52)
Time-domain floating body dynamics by rational approximations of the radiation impedance and diffraction mapping
Ocean Engineering
(2000)Simulation of wave power devices
Applied Ocean Research
(1984)- et al.
Optimal identification of potential-radiation hydrodynamics for moored floating structures: a new general approach in state space
Ocean Engineering
(2004) - et al.
State-space representation of radiation forces in time-domain vessel models
Ocean Engineering
(2005) - et al.
Nonlinear wave-loads and ship responses by a time-domain strip theory
Marine Structures
(1998) - et al.
State-space modelling of a vertical cylinder in heave
Applied Ocean Research
(1995) - et al.
An error bound for a discrete reduced order model of a linear multivariable system
IEEE Transactions on Automatic Control
(1987) Nonlinear Systems Techniques and Applications
(1998)Linear System Theory and Design
(1999)The impulse response function and ship motions
Schiffstechnik
(1962)
Absorption of outgoing waves in a numerical wave tank using a self-adaptive boundary condition
International Journal of Offshore and Polar Engineering
Modeling and Simulation for Automatic Control
Sea Loads on Ships and Offshore Structures
Hydrodynamic of High-speed Marine Vehicles
Marine Control Systems: Guidance, Navigation and Control of Ships, Rigs and Underwater Vehicles
A nonlinear unified state-space model for ship maneuvering and control in a seaway
International Journal of Bifurcation and Chaos
Effective reconstruction of linear state-variable models from input/output functions
Regelungstechnik
Application of extended state space to nonlinear ship roling
Ocean Engineering
Time domain models from frequency domain descriptions: application to marine structures
International Journal of Offshore and Polar Engineering
Linear Systems
Transient response of a VLFS during landing and take-off of an airplane
Journal of Marine Science and Technology
Complex curve fitting
IRE Transactions on Automatic Control
Cited by (351)
Application of a coupled linear-bistable system in point absorber wave energy converter
2024, Ocean Engineering