Hybrid physics-data driven spectral forecasts of semisubmersible response
I.A. Milne, L.A. Astfalck, M. Zed, J. Lee-Kopij, E. Cripps · 2026
Read Hybrid physics-data driven spectral forecasts of semisubmersible response by I.A. Milne, L.A. Astfalck, M. Zed, J. Lee-Kopij, E. Cripps (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-26026.txt.
Introduction
Milne and co-authors build a hybrid forecaster for how a semisubmersible vessel moves in swell. Physics models predict heave response from wave spectra, but they carry bias. The team wraps those predictions in a Bayesian model fit to measured motion. Diagnostics showed the error model needed uneven variance and time-linked errors. Tests focus on periods when heave resonance and cancellation bands were excited. The hybrid gives honest uncertainty ranges and corrects the physics bias. The authors say this can make offshore work safer and more efficient.
Catalog record
| CREATOR | I.A. Milne, L.A. Astfalck, M. Zed, J. Lee-Kopij, E. Cripps |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 49424 |
| LICENSE | http://creativecommons.org/licenses/by-sa/4.0/ |
| RIGHTS | CC BY SA |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.26026
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Hybrid physics-data driven spectral forecasts of semisubmersible response I.A. Milne, L.A. Astfalck, M. Zed, J. Lee-Kopij, E. Cripps Abstract A framework for probabilistic forecasting of vessel motion is developed and validated for a semisubmersible operating in long period swell. Bayesian statistical methods are applied to predictions of the heave response from a physics model using numerical wave spectra and measured motion data. Model diagnoses motivate an additional level of complexity required for the error structure in the Bayesian model, specifically to account for heteroskedasticity and time-correlated errors. The hybrid model forecasts were evaluated during periods where the heave resonance and cancellation frequencies were excited. The method is demonstrated to be effective for providing reliable quantification of uncertainty and correcting bias in the raw physics model predictions. This justifies its value for improving the efficiency and safety of offshore operations. keywords Response-based forecasting ,Bayesian statistics ,Semisubmersible ,Hydrodynamics ,Weather-forecasting † † corresponding: Corresponding author 1 Introduction Offshore operations have historically relied on forecasts of the seastate for assessing weather related risk. Most commonly these forecasts are of integrated spectral parameters such as the significant wave height ( H_{s} ), with respect to the time horizon. Many offshore operations involving floating bodies are, however, fundamentally restricted by thresholds associated with wave induced hydrodynamic response. For instance, drilling may be limited by the vertical travel of the drillstring and a heavy-lift by the velocity of the crane tip. The hydrodynamic response to surface waves can vary appreciably between the different hull forms and displacements (i.e. mass) of the vessels employed across the offshore industry. Notably for regions where long period swell is prevalent, the hydrodynamic heave resonance may be linearly excited and result in large motions. Guidance based only on the seastate can therefore compromise efficiency, safety and cost of operations. Response-based forecasts can inherently facilitate superior maritime support for offshore operations. While wave-by-wave forecasts remain challenging, phase-averaged stochastic or spectral-based response forecasts can be readily operationalised. These require knowledge of only the directional wave energy spectra, S\left(f,\theta\right) , and the Response Amplitude Operators (RAOs), the latter describing the hydrodynamic response of the particular vessel to the wave amplitude. Statistics associated with the short term vessel response can subsequently be readily estimated. Such an approach was recommended by regulation agencies two decades ago ( 13 , see) . Since then studies such as 20 ; 21 ; 18 have demonstrated the benefits for a Floating Production Offshore and Storage (FPSO) and drillships operating in a complex swell dominated seastates. Apart from ship-shaped vessels, semisubmersibles have been widely utilised as substructures since the 1960s to support offshore drilling, oil and gas production, heavy-lift cranes and more recently, floating wind turbines. Semisubmersibles have particularly attractive hydrodynamic motion characteristics; notably the attenuation of vertical motion at shorter wave periods. However, they are weakly damped systems and can be linearly excited in heave by long period swell. In contrast to the Gulf of Mexico and North Sea where numerous semis operate, for regions affected by long period swell such as North West Australia and West Africa, the heave resonance can therefore present a criticality. The near directionality invariant heave response of a typical semisubmersible suggests that formulating a spectral-based forecast of the response would be simpler than for a ship-shaped vessel. In particular, one could work directly with the 1-D wave (frequency) spectra without the need to resolve directionality or spreading. However, despite these simplifications the form of the heave RAO at long wave periods is characterised by a cancellation point and resonance, at which the response amplitudes are seastate dependent ( 9 ) . Since swells often originate at considerable distance away they are typically not well resolved by either wave models or wave buoys in frequency space and forecasting the wave energy spectra is also inherently challenging. Further compounding the challenge is that in contrast to parameterised forecasts, the wave spectra available for vessel motion prediction are typically sourced as raw outputs from regional numerical models. These factors combined can lead to potentially significant uncertainties in the heave response forecasts. This issue is addressed here by updating raw model, physics-based forecasts of heave response at a location where a semisubmersible rig is operating and measured heave motions are available. The spectra induced RAO forecasts are the state-of-the-art, industry deployed methods described in 21 . The measured heave motions are particularly novel in that excitation at the heave cancellation and resonant frequencies from long period swell is regularly observed and provide strong empirical evidence of the system at the location of interest. We demonstrate a hybrid physics-data driven approach, whereby hybrid approach we mean embedding raw physics-based heave motion forecasts within a data-driven Bayesian statistical model which is informed by the measured heave motions. The result is a method that updates and refines the forecast of the physical model in accordance with recent observations and provides a probabilistic measure of our confidence in the updated forecast to inform operational decisions. Extending the work of 15 , who consider a hybrid approach to forecasting sea surface winds for short forecast horizons of 0 and 6 hours, we consider the more challenging task of forecasting vessel heave motions, at forecast horizons of between 0 and 96 hours. We show that the hybrid approach outperforms the stand alone spectra induced RAO forecasts in terms of the proper scoring rules designed for probabilistic forecasts ( 12 ; 3 , see,) , and provide well calibrated probabilistic forecasts for both longer term (24, 72 and 96 hour) and shorter term (0, 6 and 12 hour) time horizons. The remainder of the paper is set out as follows. Section 2 outlines the physics-based approach, reviewing the fundamental hydrodynamics associated with heave response of semisubmersibles, and surface wave forecasts. Section 3 describes the Bayesian hybrid forecast approach and reviews proper scoring rules for probabilistic forecasts. Section 4 evaluates the performance of the model against the measured motions and demonstrates specific approaches to improve forecasts of the heave response at long wave periods. Section 5 contextualizes the study and discusses possible data-driven approaches to reduce inherent uncertainty in both underlying the hydrodynamic response model and forecasts of swell energy, with conclusions presented in Section 6. 2 Physics-based approach In this section, we describe the physics based approach to compute the motion statistics from wave spectra and the vessel RAOs which serve to elucidate various potential sources of uncertainty. We first introduce the preliminary hydrodynamics for submersibles and surface wave forecasting required to discuss the current state of the art vessel and industry employed semisubmersible response forecasts as described in 21 . We do so in the context of a semisubmersible operating in the North West Shelf of Australia. 2.1 Semisubmersible hydrodynamics Conventional semisubmersibles used in offshore drilling and production comprise multiple vertical columns and horizontal pontoons to provide the required stability and buoyancy. Figure 1 shows the frequency dependent characteristics of the amplitude of the hydrodynamic heave RAO ( =s_{3a}/{\zeta_{a}} , where \zeta_{a} is the incident wave amplitude associated with frequency component \omega ) for a typical twin pontoon, four column, semi. This demonstrates how the resonant ( \omega_{R} ) and cancellation ( \omega_{C} ) frequencies can be linearly excited by wave energy associated with