Search

This paper presents a comparison of three different machine learning frameworks applied in the wave buoy analogy used for estimating the sea state from measured ship responses. The three frameworks output and characterise the sea state in different ways: Model 1 outputs integral parameters, Model 2 outputs a point wave spectrum and the wave direction, Model 3 outputs the full directional wave spectrum. The assessment of the models is based on simulated motion measurements, i.e. synthetic data. In the particular investigations made, the performance of Model 2, relying on a novel framework, is generally superior. However, the central take-away from the study, is the importance of considering thorough and well-prepared training data encompassing many, not to say all, possible parameter combinations and shapes in the studied wave spectra forming the training data; any machine learning model is no better than the data upon which it is trained.

[Abstract]: This work presents a comparative study of two signal processing methods for the estimation of the roll natural frequency towards the real-time transverse stability monitoring of fishing vessels. The first method is based on sequential application of the Fast Fourier Transform (FFT); the second method combines the Empirical Mode Decomposition (EMD) and the Hilbert-Huang Transform (HHT). The performance of the two methods is analysed using roll motion data of a stern trawler. Simulated time series from a one degree-of-freedom nonlinear model, and experimental time series obtained from towing tank tests are utilized for the evaluation. In both cases, beam waves are considered but, while irregular waves are adopted in the simulated data, the towing tank tests are made in regular waves. Based on the available data the performance of both estimation methods is comparable, but the EMD-HHT method turns out slightly better than the sequential FFT. Finally, the use of a statistical change detector, together with the EMD-HHT methodology, is proposed as a possible approach for the practical implementation of an onboard stability monitoring system.

Optimisation of energy efficiency and operational performance as well as assessment of safety levels and emissions of marine operations require detailed information about the acting wave system. It is possible-with an analogy to classical wave buoys-to estimate the directional wave spectrum by processing sensor measurements of wave-induced responses (e.g., motions and structural responses) from a ship. Compared to other sources of wave data (e.g., buoys, satellites, third-generation wave models), estimation concepts using the ship itself as a buoy provide the wave spectrum at the exact spatio-temporal point, potentially increasing accuracy and with minimal associated cost. This paper gives an overview of the technology, discusses associated uncertainties, and highlights new developments made for estimating waves via measured ship responses.

This paper deals with a machine learning methodology for decomposing hull and propeller performance. This is facilitated by an ensemble approach predicting both propeller revolutions and shaft power under consistent simulated reference conditions. The derived performance estimates are compared to ISO 19030 and a semi-empirical framework aimed at isolating propeller performance. The case ship is a >300 m cruise vessel sailing in the Caribbean Sea. The auto-logged sensor data covers seven years with three dry-docking intervals and numerous in-water cleaning events. It is concluded that the separation of propeller performance is subject to multiple uncertainty sources and can thus not be reliably assessed. The practical relevance and potential shortcomings of the method are discussed.

The real-time provision of high-quality estimates of the ocean wave parameters at appropriate spatial resolutions are essential for the sustainable operations of marine structures. Machine learning affords considerable opportunity for providing additional value from sensor networks, fusing metocean data collected by various platforms. Exploiting the ship-as-a-wave-buoy concept, this article proposes the integration of vessel-based observations into a wave-nowcasting framework. Surrogate models are trained using a high-fidelity physics-based nearshore wave model to learn the spatial correlations between grid points within a computational domain. The performance of these different models are evaluated in a case study to assess how well wave parameters estimated through the spectral analysis of ship motions can perform as inputs to the surrogate system, to replace or complement traditional wave buoy measurements. The benchmark study identifies the advantages and limitations inherent in the methodology incorporating ship-based wave estimates to improve the reliability and availability of regional sea state information.

Performance data from ships is subject to distributional shifts, sometimes referred to as concept drift. In this study, synthetic monitoring data is simulated for the KVLCC2, considering publicly available reference data and a semi-empirical simulation framework. Neural networks are trained to predict the required shaft power and to overcome the deterioration in model accuracy due to concept drift, three methods of incremental learning are applied and compared: (1) Layer freezing, (2) regularization, and (3) elastic weight consolidation. Furthermore, an implicit methodology for quantifying the changing hull and propeller performance is presented. In addition, a generic feature engineering framework is used for eliminating insignificant features. In two investigations, sudden and incremental concept drift scenarios are examined, and the effect of different uncertainty categories on model performance is studied in parallel based on three different datasets. As a main finding, it is confirmed that data quality is of great importance for accurate machine learning-driven performance monitoring — even in simulated environments. Furthermore, the study shows that freezing layers during incremental learning proves to be most robust and accurate, but it will be part of future work to examine this on actual sensor data.

In vessel performance analysis, reliable information about speed-through-water (STW) is key for realistic modeling of the single ship’s fuel efficiency. It is paramount that STW measurements are reliable such that they can be adopted as input in fuel consumption forecast models. This paper presents a study where three variations of a Dynamic Factor Model are used to estimate the STW. A simulation study is presented to demonstrate the estimation and smoothing techniques for the Dynamic Factor Model. The model variations are then applied to real vessel data, including STW and propeller measurements, in combination with metocean data (sea currents). The present study suggests that the developed model can minimize the systematic STW measurement error even under highly non-stationary conditions.

Somatic embryogenesis (SE) cloning techniques for Nordmann fir Christmas tree production have been pursued and refined in Denmark, and three SE clonal trials with 201 SE clones from seeds of 27 Nordmann fir trees and seedlings from bulk harvest seeds in Georgia, Caucasus, were established from 2014 to 2015. This study was the first to evaluate the genetic performance of these SE clones in different sites. In addition, the performance of SE clones and (zygotic) normal seedlings in the field was compared in this study. This study was based on three key groups of traits for Christmas tree production, measured 5 and 6 years after planting: growth, commercial tree quality, and disorder. Seedlings grew faster and had better Christmas tree quality than SE clones. There were significant family effects for all traits, indicating that a good family also produced good clones on average. Growth and disorder traits were under moderate genetic control, with estimated broad-sense heritability (H-2) from 0.19 to 0.31 and from 0.20 to 0.28, respectively. Quality traits had low to moderate H-2, ranging from 0.09 to 0.24. Significant genotype-by-environment interactions were shown for Christmas tree quality, post-harvest needle retention, and branch angle, suggesting the importance of choosing the right clone at each site. There was no significant interaction in height among the years. Superior clones, compared to the unbred seed source, were identified by combining several important traits. Overall, the somatic clones were well established but had slower growth compared to standard seedlings.

A practical estimation methodology of the mean added resistance in irregular waves is shown, and the present paper provides statistical analyses of estimates for ships in actual conditions. The study merges telemetry data of more than 200 in-service container vessels with ocean re-analysis data from ERA5. Theoretical estimates relying on spectral calculations of added resistance are made for both long- and short-crested waves and are based on a combination of a parametric expression for the wave spectrum and a semi-empirical formula for the added resistance transfer function. The theoretical estimates are compared to predictions from an indirect calculation of added resistance relying on shaft power measurements and empirical estimates of the remaining resistance components. Overall, the comparison reveals a bias in bow oblique waves and higher sea states of the spectral estimates as well as the large variance of the empirically derived predictions — particularly in beam-to-following waves. One of the study’s main findings, confirming previous studies but based on a much larger dataset than in earlier similar studies, is that added resistance assessment based on in-service data is complex due to significant associated uncertainties.

Despite a much later inception of somatic embryogenesis (SE) propagation protocols for gymnosperms than for angiosperm species, SE is becoming increasingly important due to its applications for commercial forestry. For many conifers, there are however still major bottlenecks in the SE plant production process limiting the use of SE for forestry operations, Christmas tree production and research projects. In the present case study, the effects on plant growth from different cultural factors applied during the SE developmental process were studied in two conifer species of high value for Christmas tree production. Seven clones of Abies nordmanniana and two clones of Abies bornmuelleriana were included in the study. Accumulated effects from cultural treatments were recorded from the start of germination of mature embryos of different quality scores through development into plants in the third growing period. Experimental factors of the cultural treatments included were: germination temperature, germination time, light conditions, survival ex vitro and traits for plant growth and vitality. The results reveal that most of the studied experimental factors influenced plant growth during the first three years however their relative importance was different. Plant survival rate at end of the nursery stage was strongly impacted by germination temperature (p<0.001), initial embryo score (p=0.007), clone (p<0.001) and to a lesser extend week of germination (p=0.017). This case-study highlights and quantifies the strong interrelation between the developmental steps of somatic embryogenesis and show the importance of considering all cultural steps when optimizing SE plant production protocols.

This study assesses the relationship between speed and power of ships, and is based on operational data derived from more than 50,000 noon reports retrieved from 88 tankers. The analysis is made with a data-driven model based on principles from naval architecture and an econometric framework. The model facilitates the determination of a draught- and speed-dependent exponent. It is shown that the exponent is significantly lower than 3 at speed intervals below the design speed. The finding of a (much) smaller exponent, relative to 3, at low speed intervals is important in the slow steaming debate, since slow steaming will not be as good as often stated when the cubic speed-power relationship does not hold. On a practical level, the developed method can be used to set a more reliable benchmark in the performance monitoring of ships.

In the present study, model drift – mostly driven by biofouling – is mitigated by applying continual learning and the hydrodynamic performance of a large LNG tanker trading in worldwide service is assessed in parallel. The available auto-logged sensor data spans five years including two dry docking intervals. An adaptive training methodology of an artificial neural network is established, and the drift score is derived from simulated data under sea trial conditions. In addition, hindcast metocean data from ERA5 is included in the model’s feature space for capturing environmental conditions. Crucially, it was found that the obtained drift estimate is in general accordance to ISO 19030 results, proving the method’s validity. Still, the methodology itself is subject to considerable uncertainty. Finally, limitations and possible extensions of the proposed methodology are presented.

The reliable and accurate determination of prevalent sea state conditions during a vessel’s passage is of high scientific and practical importance not only due to crew and cargo safety, but also due to increasing energy efficiency demands. The present paper deals with the prediction of integral sea state parameters using 5-minute time series samples of wave-induced ship motions and builds upon deep neural networks. The study is based on synthetic motion histories of heave, pitch and roll of a container vessel generated from transfer functions and a parametric wave spectrum. The multi-dimensional search space of the four free variables: significant wave height, zero-upcrossing period, relative wave direction and advance speed, is discretized by Latin Hypercube Sampling. In order to comply with the assumption of moderate wave steepness, an empirical wave breaking constraint is incorporated into the spectral simulation procedure. Both a residual network and a multichannel convolutional LSTM (Long Short Term Memory) network are employed and compared to a state of the art architecture. In the frequency domain, a multilayer perceptron is used for comparison and trained on features obtained from cross-spectral analysis. The results of the simulation-based study show sufficient accuracy and generalization for all models in both time and frequency domain. Lastly, limitations of the present study are pointed out and it stressed that the applicability of the presented methodology to real in-situ measurement data is an important part of future work.

The analysis of wave-ship interactions is highly relevant for the safety – as well as the energy efficiency – of the maritime operations. One application is the onboard estimation of the ship’s responses in incoming seaways, which requires regular and accurate updates of the vessel’s seakeeping model, accounting for possible changes in the operational conditions. This paper presents a simple approach for fast estimation of the wave-to-motion transfer functions of vessels. Prior information of the wave spectrum, characterizing the sea state, and ship motion measurements, i.e. time series sequences from onboard sensors, are supposed to be available. Semi-empirical closed-form expressions derived for a box-shaped vessel define Parameterized Response Amplitude Operators (P-RAOs) for the heave and pitch motions. The five input parameters, namely the ship speed, length, breadth, draught and block coefficient, are regarded as optimization variables. An optimization problem is established to minimize the spectral discrepancy between, on one hand, the measured responses, and, on the other hand, the theoretical responses computed with the wave spectrum and P-RAOs. Numerical simulations of measured motions in a predefined long-crested sea state are carried out in the frequency domain for two different ships, a small research vessel and a container ship, using a set of RAOs obtained by a commercial potential flow code. The simulated measurements are considered as the ground truth. Tuning of the P-RAOs is carried out, and the results show a fairly good agreement between the tuned P-RAOs and the true RAOs over a wide portion of the frequency range. Moreover, the normalized error between the true and estimated response spectra is significantly decreased after tuning the P-RAOs.

The article presents a study investigating the level of variation in sea state parameters encountered by sailing ships crossing the oceans. The sea state parameters have been obtained from a reanalysis, in this case the ERA5. The study is based on the use of different interpolation schemes to compute parameters in geographical positions off the fixed grid. It is shown that the variation in sea state parameters can be significant. Consequently, in case of sailing ships, covering relatively long distances in a short time (30-60 minutes), it is recommended to rely on bilinear interpolation rather than nearest neighbour. The variation in the sea state parameters is, in fact, at a level which means that the normal assumption of a stationary seaway in periods up to 3 hours likely is violated for ships sailing the typical service speed (15 – 20+ knots).

The ever-changing sea state constitutes an integral part of situational awareness for vessels at sea. In this paper, a computationally efficient sea state estimation algorithm is extended by (1) automatic gain and tolerance calculation, and (2) trust measure for wave filtering condition identification. The resulting algorithm can run online in a control system without the need of user input, producing estimates of sea state parameters (significant wave height, peak wave frequency and main propagation direction), while simultaneously monitoring the quality of the estimates produced by identifying when wave filtering occurs. The algorithm is developed for application in dynamic positioning (DP) control systems, where the vessel has zero (or low) forward speed. The estimation method is tested with realistic values of measurement noise and sampling in terms of comprehensive time series simulations based on both long- and short-crested wave conditions. The sea state estimates correspond well with a state-of-the-art Bayesian estimation approach. This indicates that the simplifications made in order to enhance the computational speed, has not affected the quality of the estimates to any significant extent. The trust measu

This paper presents a study focused on wave spectrum estimation in practical scenarios where multiple ships operate in the same geographical area, potentially forming a network of wave recorders. A novel methodology is proposed to improve the accuracy and precision of the wave spectrum estimates, by combining sea state estimation methods and techniques for tuning the wave-to-motion transfer functions. The framework of the wave buoy analogy is used to derive estimates for each ship through the use of measured ship motion data and available initial estimates of transfer functions. Simultaneously, the wave-to-motion transfer functions of the individual ship are tuned by utilizing a weighted version of the wave data inferred on board the other ships in the network. The overall architecture of the procedure is modular, in the sense that various approaches may be implemented for obtaining sea state estimates and tuned transfer functions. The methodology is demonstrated through two case studies, one based on simulated vessel responses, and the other using model test data of ship motions in a wave tank. Both case studies consider a network of three ships in long-crested waves equipped with a dynamic positioning system. It is shown that the procedure provides good wave spectrum estimates, and leads to reduced uncertainty in the estimates via tuning of the vessel transfer functions.

In the present study, a Panamax container vessel sailing in the Northern Atlantic is taken as a case study. Deep neural networks are trained on results of cross-spectral analysis of 6-DOF accelerations to predict significant wave height, peak period, and relative wave direction, which were initially derived from measurements of an X-band wave radar. Both an Inception network and a residual network are compared. Overall, it is found that the vector decomposition of the relative wave heading shows superior prediction accuracy in contrast to other methods from state-of-the-art literature. Several possible extensions of the presented methodology are pointed out.

This paper presents a method that can be used to obtain the motion transfer functions of a ship when the detailed hull geometry is not available. In the method, any parameterised transfer function can be optimised from a data-driven approach that compares a measured response spectrum and the corresponding theoretical spectrum based on the combination of transfer function and wave spectrum. The established method facilitates association of 95% confidence bands which are useful to indicate a level of trust. The proposed method is evaluated on a conceptual level using both simulated and in-service measurements, and promising results are obtained.

This paper is concerned with a machine learning-based approach for sea state estimation using the wave buoy analogy. In-situ sensor data of an advancing medium-size container vessel has been utilized for the prediction of integral sea state parameters. The main novelty of this contribution is the rigorous comparison of time and frequency domain models in terms of accuracy, robustness and computational cost. The frequency domain model is trained on sequences of spectral ordinates derived from cross response spectra, while the time domain model is applied to 5-minute time series of ship responses. Multiple deep neural networks were trained and the sensitivity of individual sensor recordings, sample length, and frequency discretization on estimation accuracy was analysed. An Inception Architecture adapted for sequential data yields the highest out of sample performance in both considered domains. Additionally, multi-task learning was employed, as it is known for increased generalization capability and diminished uncertainty. Overall, it was found that the frequency domain method provides both superior performance and significantly less computational effort for training.

The present paper examines a semi-empirical framework for the estimation of added resistance in arbitrary wave heading under consideration of uncertainty quantification. In this respect, the calibration of the formula’s parameter vector is conducted based on particle swarm optimization as well as a database of model test results comprising 25 different ships and around 1100 samples. In the first iteration, the minimization of reducible systematic uncertainty is of interest and the effect of four objective functions on prediction accuracy is evaluated. Moreover, two different parameter combinations were obtained for blunt (CB≥0.70) and slender-type ships. Conversely, the irreducible statistical uncertainty, i.e. the inherent noise of the experimental data, is taken into account by a quantile regression procedure. Applying this approach, a 90% prediction interval for the formula’s estimates is implemented using the skewed version of the superior loss function in the previous iteration. The practical relevance of an uncertainty estimate for the prediction of the added resistance is emphasized in the final part, in which the proposed approach is validated in regular waves against model test data and other well-established prediction methods. In general, the validation studies suggest satisfactory performance and reliability of the adapted semi-empirical formulation.

This paper presents a new approach to attain estimates of the sea state based on short-time sequences of wave-induced ship responses. The present sea state estimation method aims at reconstructing the incident wave profiles in time domain. In order to identify phase components of the incident waves, the Prolate Spheroidal Wave Functions (PSWF) are employed. The use of PSWF offers an explicit expression of phase components in the measured responses and incident waves, indicating that estimations can be efficiently attained. A method to estimate the relative wave heading angle based on the response measurements and pre-computed transfer functions of the responses is also proposed. The method is tested with numerical simulations and experimental measurements of ship motions, i.e. heave, pitch, and roll, together with vertical bending moment and local pressure in a post-panamax size containership. Validation is made by comparing the reconstructed wave profiles with the incident waves. The accuracy and efficiency of the present approach are promising. At the same time, it is shown that the use of responses, which are more broad-banded in their frequency characteristics, is an effective means to cope with high frequency noise in reconstructed waves.

This paper studies real-time deterministic prediction of wave-induced ship motions using the autocorrelation functions (ACFs) from short-time measurements, namely the instantaneous ACFs. The Prolate Spheroidal Wave Functions (PSWF) are introduced to correct the large lag time errors in the instantaneous sample ACF, together with a modification of the autocorrelation (AC) matrix for ensuring its positive definiteness. The validity of the PSWF-based ACFs is first examined by using the ship motion measurements from model experiment under stationary wave excitations. It is shown that the use of PSWFbased ACFs leads to better prediction accuracy than direct use of sample ACFs. The validation is then extended to ship motion prediction using in-service data from a container ship, and an improvement of the prediction accuracy by PSWF-based ACFs is again found. Finally, the effectiveness of use of the instantaneous ACFs for non-stationary wave-induced responses is highlighted by comparing with the prediction results based on the ACFs from long-time measurements.

This paper addresses the connection between added wave resistance and required propulsion power of ships, having focus on the early stage of new ship designs, notably tankers and bulk carriers. The paper investigates how mean added wave resistance affects the required torque of a fixed pitch propeller and thus also the operational conditions of a directly coupled main engine. The interest of the study has its background in the assessment of minimum propulsion power, and the study considers the prescriptive guidelines of the IMO as basis. Specifically, the study focuses on an assessment of the minimum forward speed attainable under consideration of the propeller light running margin and static load limits of engines in the early phase of new ship designs, where details of hull geometry are not available. The study considers three semi-empirical methods for predicting mean added wave resistance. All methods are known to be applied in the industry, emphasising that only methods relying solely on main particulars, together with information about sea state and advance speed, are of interest. The paper contains a case study used to illustrate the importance of the added wave resistance prediction with respect to the loading of the main engine. It is shown that, despite small absolute differences, the consequence in relation to the loading of the propeller and hereby the directly coupled main engine can be relatively large. Furthermore, the study illustrates that the propeller light running margin of a fixed pitch propeller directly coupled to the main engine has crucial influence on the attainable speed during adverse weather conditions.

This paper investigates a semi-empirical model used to estimate added-wave resistance on a ship sailing in waves. The model relies on measurements from a continuous monitoring system, and produces an estimate - the indirect measurement - of added-wave resistance, based on the difference between, on the one side, the measured power and, on the other side, a summation of theoretically calculated resistance contributions but neglecting the component because of seaway. Hence, the added-wave power is obtained as the surplus. The model has been applied to more than three months of full-scale data recorded on an in-service operating container ship. The results indicate reasonable levels of the added-wave power, emphasising that both positive and negative values are obtained, since the model is not restricted to relative wave directions forward of beam. It is found that following waves often lead to 'negative resistance', in fact creating a thrust, but, importantly, following waves can also increase the total resistance significantly.