Abstract
A rigid body model for the dynamics of a marine vessel, used in simulations of offshore pipe-lay operations, gives rise to a set of ordinary differential equations. This system of equations consists of terms modelling Coriolis effects, damping, restoring forces and moments, environmental forces, and controls. We propose splitting methods tailored to the numerical solution of this problem. The suggested splitting is into the conservative part, and the remaining part containing damping and control. The conservative part is completely integrable, even though it is nonlinear. The damping and control part is linear, and therefore easily integrable. We investigate stability and energy behaviour of the methods in numerical experiments.
The Challenge of Simulating Offshore Operations
Simulating offshore operations, such as laying pipelines on the seabed, requires highly accurate models of marine vessels. These vessels are modeled as rigid bodies influenced by Coriolis effects, damping, environmental forces (wind/waves), and active control systems like PID controllers.
The primary mathematical challenge in these simulations is maintaining passivity. In physics, a passive system is one that does not create energy; it only stores or dissipates it. While the continuous physical equations are passive, standard numerical discretization methods (such as the ubiquitous Runge-Kutta) often fail to preserve this property. This "numerical leakage" can lead to artificial energy growth, causing simulations to become unstable or physically nonsensical, especially over long durations or with large time steps.
Geometric Numerical Integration and Splitting Methods
To address these issues, the authors employ Geometric Numerical Integration, specifically focusing on Splitting Methods. The core idea is to decompose the complex system of Ordinary Differential Equations (ODEs) into two distinct, more manageable parts:
- $S_1$ (Conservative): This part models the free-rigid-body flow. Although nonlinear, it is completely integrable using Jacobi elliptic functions.
- $S_2$ (Dissipative/Control): This part models damping, gravity, and control inputs. It is linear and can be solved exactly using the "variation of constants" formula.
By composing the exact flows of these two subsystems, the resulting numerical scheme remains "input-output passive." Furthermore, to avoid mathematical singularities like "gimbal lock" associated with traditional Euler angles, the vessel's attitude is reformulated directly on the Lie group $SO(3)$ using rotation matrices and Magnus series expansions. The authors extend basic Lie-Trotter (1st order) and Strang (2nd order) splittings to 4th and 6th-order symmetric compositions to achieve high precision.
Key Insights and Performance
The fundamental insight is that passivity can be maintained if the system is split into a skew-symmetric (conservative) part and a dissipative part. By choosing a splitting where each sub-step has an exact analytical solution, the method becomes both faster and more robust than "black-box" solvers that approximate the whole system at once. Integrating on $SO(3)$ ensures the vessel's orientation always stays physically valid (i.e., the rotation matrix remains orthogonal).
In numerical experiments using physical parameters for a supply vessel, the splitting methods (SP2, SP4) demonstrated significant advantages:
- Stability at Scale: They remained stable and accurate at large time steps where standard explicit Runge-Kutta methods became completely unstable.
- Energy Fidelity: The methods followed the exact energy and torque curves of the physical system almost perfectly. In contrast, Improved Euler and RK4 methods showed significant drift in energy behavior.
- Convergence: Numerical tests confirmed that the SP2, SP4, and SP6 methods achieved their theoretical 2nd, 4th, and 6th-order accuracy, respectively.
Conclusion and Future Directions
While these splitting methods offer a robust framework for marine vessel simulation, several areas for future work remain. The methods have not yet been fully optimized for computational efficiency regarding the number of stages composed. Additionally, while the vessel dynamics are handled on $SO(3)$, the PID controllers used still rely on local Euler angle parametrizations; a future "pure" $SO(3)$ controller design would likely be more robust. Finally, achieving passivity preservation for orders higher than two may require the investigation of complex coefficients.
Summary updated