This paper presents a method for data-driven identification of passive systems, with a focus on systems with nonlinear friction models. This is achieved by approximating the passive dynamics of a given system in a reproducing kernel Hilbert space (RKHS). The main novelty of this approach is the use of a quadratic kernel model to ensure that the estimated friction dynamics is passive. The regularized regression is done by optimizing a convex Fenchel dual of the optimization problem. The method was validated with a simulation study, where it was first applied to the estimation of a Karnopp friction model with Stribeck effects. It was also applied to the identification of systems dynamics of a mass-spring-damper (MSD) system with the same friction model. The identification was performed using two different forms of input data: sampled input and output signals and the Legendre coefficients corresponding to these signals. Results demonstrated accurate identification of training data, as well as examples of generalization to test data. Passivity and causality of the approximated continuous system do not guarantee passivity and causality of its corresponding discretized or spectrally approximated system. Although simulation results demonstrate this, increasing the sample size could increase representation accuracy and thereby the preservation of these properties.
While physics-informed neural networks (PINNs) are commonly employed for time-domain problems, this work extends their application to the frequency domain for parameter estimation. The framework integrates the governing equation into the neural network (NN) training loss function by using a composite complex domain formulation to enforce data consistency, minimize physics residuals, and respect parameter bounds. The methodology is presented as a transfer function-based formulation, whose parameters can be estimated from available data. The effectiveness of the methodology is validated through both simulation and experimental datasets on axial active magnetic bearing (AMB) systems, which are open-loop unstable and thus make frequency-domain parameter estimation particularly relevant. The results show excellent agreement between measured and predicted responses. Furthermore, the proposed approach is compared with an existing frequency-domain identification approach. PINN-based estimation of physical parameters is closer to reference values due to physics residual and parameter bounds. The findings validate the proposed methodology, which indicates that the PINN-based frequency-domain framework can accurately estimate system parameters, demonstrating its accuracy and robustness for physics-constrained learning in the frequency domain.
Determination of the pose of workpieces is important for robotic applications in manufacturing, including handling, assembly, machining and welding. Established methods based on 3D sensors may fail for workpieces with highly reflective materials. In this paper, we take advantage of recent development in machine learning to determine the pose of reflective workpieces without the use of depth data. Our proposed method is based on deep iterative matching of image data of the workpiece with a computer-aided design model. Starting with an initial estimate of the workpiece pose, the method iteratively aligns the computer-aided design model projections with an image of the actual workpiece, adjusting the pose until computer-aided design model matches the image of the workpiece. The deep learning-based approach optimizes this alignment by updating the pose estimate at each iteration, achieving high precision even for geometrically complex or reflective surfaces. This refinement process enhances accuracy in robotic applications where precise workpiece positioning is critical, such as in automated welding and assembly tasks. We use photorealistic rendering to create two datasets for pretraining the network, which reduces both training time and the need for real labeled data. After the network is trained on synthetic data, it is fine-tuned and tested on real images of reflective aluminium workpieces. We show that the proposed deep iterative matching method outperforms established methods based on iterative closest point with two 3D scanners due to large errors in the scans caused by reflections.
This paper presents a data-driven framework utilising neural networks to approximate the Koopman operator for a discrete-time representation of an electrically actuated two-link mechanical system, enabling the application of linear control techniques. A validated dynamic model is used to generate training data. The Physics Informed Fully Embedded Koopman (PIFEK) framework embeds the original state space directly into the Koopman space, leading to sensible tuning of the controllers. The tuning was done using bandwidth analysis of the Koopman dynamics, yielding estimates of the true dynamics. This resulted in satisfactory control performance, including accurate tracking of a trajectory. Comparisons with a conventional PI controller show that while performance is similar, the data-driven PIFEK approach requires significantly less system-specific knowledge, underscoring its potential for efficient control design.
Species that have persisted over millions of years have done so because they have been able to track peaks in an adaptive landscape well enough to survive and reproduce. Such optima are defined by the mean phenotypic values that maximize mean fitness, and they are predominantly functions of the environment, for example the sea temperature. The mean phenotypic values over time will thus predominantly be determined by the environment over time, and the trait history may be found in the fossil record. Here I use fossil data from four cases found in the literature, and show that adaptive peak tracking models give better results than alternative weighted least squares and directional random walk models. The model performances are compared by use of weighted mean squared errors and Akaike information criterion results.
The safety property of dynamical systems has typically been studied in Euclidean spaces. In this work, we extend the notion of safety to a non-Euclidean geometry. Motivated by the role of time as a fourth dimension in physical models, we construct the 4-dimensional Heisenberg Lie group H4 and investigate the safety problem of dynamical systems defined on this group. Unlike odd-dimensional Heisenberg Lie groups, which admit a unique structure, even-dimensional cases allow multiple forms; in particular, H4 possesses four distinct forms. Focusing on one such form, we provide a detailed analysis of dynamical systems on H4. Moreover, using a diffeomorphism between the (2n+1)-dimensional Heisenberg Lie group and the Euclidean space of the same dimension, we establish their equivalence, and we extend safety result for H4. Several examples are presented to illustrate the applicability of the theoretical results.
Determination of the pose of workpieces is important for robotic applications in manufacturing, including handling, assembly, machining and welding. Established methods based on 3D sensors may fail for workpieces with highly reflective materials. In this paper, we take advantage of recent development in machine learning to determine the pose of reflective workpieces without the use of depth data. Our proposed method is based on deep iterative matching of image data of the workpiece with a computer-aided design model. Starting with an initial estimate of the workpiece pose, the method iteratively aligns the computer-aided design model projections with an image of the actual workpiece, adjusting the pose until computer-aided design model matches the image of the workpiece. The deep learning-based approach optimizes this alignment by updating the pose estimate at each iteration, achieving high precision even for geometrically complex or reflective surfaces. This refinement process enhances accuracy in robotic applications where precise workpiece positioning is critical, such as in automated welding and assembly tasks. We use photorealistic rendering to create two datasets for pretraining the network, which reduces both training time and the need for real labeled data. After the network is trained on synthetic data, it is fine-tuned and tested on real images of reflective aluminium workpieces. We show that the proposed deep iterative matching method outperforms established methods based on iterative closest point with two 3D scanners due to large errors in the scans caused by reflections.
In wind turbines, the pitch system is a critical subsystem for both regulating the power output and for the safety of the turbine. However, it is also one of the leading contributors to downtime in turbines, as it is prone to faults like internal and external leakage. In this article, a novel method for estimating load and leakage in hydraulic pitch systems is presented. The proposed method is based on Unscented Kalman Filters, where the method integrates load estimation to handle the stochastic nature of wind loads, enhancing the accuracy of leakage detection. Based on the developed method, simulation and experimental results are presented that demonstrate the feasibility and robustness of the method under varying operating conditions and for different parameter variations. From the results, it is therefore found that the method shows good promise for being applied in condition monitoring and fault detection in pitch systems and may therefore be used to reduce downtime and maintenance costs in wind turbine operations.
The maintenance costs associated with offshore wind turbines, particularly those related to logistics and system downtimes, are significantly influenced by the reliability of hydraulic components, especially the pitch control system. Accumulator failures, which constitute a notable percentage of system faults, often result from gas leakage and pressure drops, highlighting the need for efficient fault detection and diagnosis (FDD) methods. This paper presents a novel approach utilizing Long Short-Term Memory (LSTM) neural networks for detecting faults in hydraulic accumulators. Two LSTM models were developed: a regression model that estimates the exact pre-charge pressure and a classification model that predicts pressure ranges. The models were trained and validated using both experimental and simulation data from a hydraulic test setup. Results demonstrated that the regression network achieved a root mean square error (RMSE) of approximately 4.2 bar, while the classification network reached 78.75% accuracy. The findings show that LSTM networks provide precision similar to prior art but for a larger variation of load cases. Thus, the proposed non-invasive method is promising for early fault detection in offshore wind turbine accumulators, potentially reducing operational costs and enhancing maintenance strategies.
This paper provides a comprehensive examination of controller design for hydropower systems equipped with Francis turbines operating in isolated conditions. By employing a mechanistic modelling approach using differential algebraic equations, the study captures the complex interplay of hydraulic, mechanical, and electrical subsystems, enabling an in-depth analysis of system dynamics under varying load conditions. A two-step approach is adopted, where a PID controller is initially designed for a linearized model and subsequently tested on a nonlinear model, allowing for a systematic evaluation of its performance, in accordance with the Norwegian Transmission System Operator specifications. The controller design process emphasizes achieving critical stability margins, meeting industry standards, and addressing the challenges posed by nonlinear system behaviour. The novelty of this work lies in the use of a recently developed Francis turbine model, its application to a real-world hydropower plant using realistic parameters, and the presentation of the controller design from a control engineering perspective. Directions for future work include exploring optimization-based controller designs, incorporating realistic load profiles, and refining system model to address complex real-world scenarios.
A notable share of greenhouse gas (GHG) emissions from road freight in Europe stems from heavy-duty vehicles (HDVs). Despite being a small fraction of the overall vehicle fleet in Finland, the contribution of HDVs towards GHG emissions is disproportionately large. European Union (EU) aims to reduce the new HDV fleets emissions to 30 % by 2030, with Finland targeting a 50 % reduction in transport sector emissions by 2030 and complete elimination by 2045. This study aims for the estimation of energy and power demand for electrification of HDVs in Finland, however the approach can be applied to other regions and countries as well. Utilizing traffic volume data from 376 traffic measurement system (TMS) points on Finland's 28 main roads, the study classifies HDVs and calculates their fuel and electrical energy consumption (EEC). The results indicate a need for 4.89 TWh of annual peak energy for 100% electrification of HDVs, reflecting a minimum 0.614 GW power demand and requiring 1,755 chargers (each with a capacity of 350 kW at 22 h/day utilization). The analysis includes spatial mapping of energy density, energy demand, power requirements, and charging stations placement based on alternative fuels infrastructure regulations (AFIR) by EU. The obtained results can be future utilized to study local grid strength and possibility to participate in the frequency markets.
To explain transpiration results from experiments on stomatal oscillations in oat plants, it is shown by simulations on a model including both hydro-passive and hydro-active feedback that the model must include hydro-active control of the osmotic pressure of the subsidiary cells. Hydro-active feedback was used between the turgor pressure of the mesophyll cells, acting as sensor cells, and the osmotic contents of the subsidiary- and guard cells. In the model, a reduction of the turgor of the sensor cells results in an increase of the osmotic content of the subsidiary cells and a reduction of the osmotic content of the guard cells. Simulations showed that the hydro-active feedback to the subsidiary cells was always needed. However, it was also shown that it is an advantage to combine the hydro-active feedback to the subsidiary cells with hydro-active feedback to the guard cells. The added hydro-active feedback to the guard cells will prevent too high turgor levels in the guard cells when the stomata have closed at water stress with high light levels. The model consists of established evaporation- and plant waterflow models from literature, experimentally verified models on stomatal mechanics and new models of hydro-active feedback. The model explained results reported in experiments where the water potential of the root medium was lowered and in experiments where the potential rate of evaporation was increased.
Environmental disturbances such as wind, currents, and waves introduce uncertainties in hydrodynamic parameter estimation, affecting the accuracy of ship maneuvering models and trajectory prediction. This study investigates how these disturbances influence the estimation of hydrodynamic derivatives in the Abkowitz maneuvering model and their impact on their accuracy. Therefore, models are estimated using least squares regression based on data from four maneuvers conducted under different wind and current conditions. A comparison of the resulting hydrodynamic derivatives identifies parameters that exhibit greater variance as disturbance intensity increases. To further assess their influence on the trajectory prediction error, Sobol sensitivity analysis is applied to determine which parameter variations most significantly affect trajectory accuracy. The results reveal that while some parameters remain stable across environmental conditions, others exhibit slight variations. Additionally, the parameters most affected by disturbances are not necessarily those with the greatest impact on trajectory prediction error. These findings highlight the importance of accounting for environmental effects in vessel model estimation to improve prediction reliability and provide deeper insight into how disturbances impact the model estimation process.
With the popularity and use of s-domain models for control, time-domain models have been less worked upon. Easy availability of methods to solve ordinary differential equations (ODEs) and differential algebraic equations (DAEs) makes it possible to work directly with the time-domain models nowadays. This would provide the flexibility and model structure for various non-linear analysis and implementation of modern control schemes. This paper presents a detailed time-domain model of variable speed hydro power plant suitable for control, including the hydraulic and electric parts up to the grid. The converter configuration used on the grid side is replaced by a virtual synchronous generator considering only active power control. The model is simulated for step and ramp load changes and together with the governing action. Results show that the upstream water oscillation and pressure imbalance dynamics depends on the size of load change. The ramp load change seems to have more smooth operation on the upstream side. Furthermore, the turbine speed is allowed to deviate away from its reference value for a while. This opens the possibility of utilizing the rotational energy in the turbine generator unit with properly controlled output power without causing the small signal instability in upstream side. Also, it is possible to provide the ancillary services to the grid using this model and controlling it as required.
Studies of phenotypic responses in wild populations are often based on reaction norm models where the environmental drivers in many cases are related to climate change. Such input signals will never be exactly known, and there will always be measurement errors also in the recorded responses. In parameter estimation these errors give rise to errors-in-variables problems, especially in the form of overfitting caused by errors in the input measurements. A second important feature of such phenotypic response problems is that the environmental inputs must be given appropriate but largely unknown reference values. A third problem is that it is difficult to find good validation methods for predicted responses. Essential aspects of these problems are here studied by use of a reaction norm model in its simplest univariate form, characterized by a mean intercept value and a mean plasticity slope value, and the overall conclusion is that validated disentanglement of plasticity and genetic adaptation based on realistically short data for wild populations is a difficult task. In a proposed validation method, the available input-output data is split into one part for modeling and one part for validation, and the feasibility of this approach is studied in simulations with use of a prediction error method, which is essentially a maximum likelihood method. It is also argued that validation of a chosen or estimated reference environment in practice is impossible when the data comes from the (unintended) anthropogenic global warming experiment, where no independent experimental data exists. When the evolution is slow because of small genetic variances, overlapping generations and long lifetimes, or because of near optimal adaptive plasticity, the best quantitative genetics option may be to assume a constant plasticity slope value, equal to the initial value. It turns out to be easy to estimate this value, but that should be done without setting other unknown parameter values to zero. This option is appealing also because it removes the dependence of a guessed or estimated reference environment.
The energy transition of the Norwegian ocean-going fishing fleet is challenging since there are few widely available fuel alternatives. Many large fishing vessels stay at sea for long periods between each fuelling. To complete its tasks and have space for the caught fish, an energy system which efficiently utilizes its fuel is required. This work proposes a flexible optimization-based mixed-integer linear programming tool for sizing, scheduling and analysing maritime energy systems consisting of diesel gensets and batteries, which considers both part-load engine efficiency and battery degradation. The tool is applied on a 24-hour data selection from an existing trawler, and the results point to the main fuel savings being enabled by splitting the installed power capacity into smaller gensets that can run independently from each other, and from utilizing a battery system to both reduce the total installed genset capacity to increase their relative loading and by delaying the startup of new gensets. Smaller fuel savings can be achieved from peak shaving of already running gensets. The current application achieves a 7.3% fuel reduction in a cost-efficient manner over the period. It also suggests that a fuel-reducing system might not necessarily be cost-efficient, particularly for energy systems with small batteries.
A solution to the Perspective-n-Lines (PnL) problem is proposed where a large fraction of outliers can be handled. The approach estimates camera pose from 2D-3D line correspondences where outliers are in the form of line mismatches. The solution is based on graduated non-convexity (GNC) with truncated least squares with Dynamical Pose Estimation (DAMP) as a solver. The solution is simple to implement and does not require specialized optimization software. The method is compared to 11 state-of-the-art PnL methods using synthetic and real data and evaluated in terms of accuracy, running time, and sensitivity to noise and outliers. The results show that our proposed method scores among the top for accuracy and robustness.
This simulator is designed to support research on centralised game-theoretical algorithms for maritime traffic management. It supports an arbitrary number of vessels and land masses import. Vessels are modelled as agents whose motion is governed by the kinematic equations and the land masses are polygon shape files. In the simulator, each vessel has access to the reward oracle, which evaluates the agents' strategies by taking into account the risk of collision and grounding, the level of compliance with the traffic rules and the operational efficiency. A game-theoretical model predictive control then generates optimal trajectories for every traffic participant simultaneously. Vessels are engaged in repeated competitive polymatrix games, whose equilibria solutions are a series of waypoints, meant to be broadcast as navigational decision support by the Vessel Traffic Services. We convey the agents' and functions' modelling principle implemented in NetLogo and present the overall simulator structure and scope.
In offshore environments, safe management of heavy payloads requires precise crane operations to avoid collisions with obstacles and adjacent equipment. Uncontrolled residual swinging of suspended payloads can quickly evolve into high-risk situations, which, if left unchecked, might lead to significant equipment failures and associated costs. This paper explores a control methodology designed specifically to eliminate payload swing in offshore cranes. We present a trajectory tracking technique explicitly crafted for swing suppression under control, rooted in the principles of the iterative learning algorithm and based on physics. The proposed antiswing control strategy guarantees asymptotic convergence of the payload's swing, angular velocity, and angular acceleration to desired values. The method was tested on a Comau robot mounted on a Stewart platform at the Norwegian Motion Laboratory. Simulation and experimental results comparing payload transfers with and without applying the anti-swing control method validates it's effectiveness.
In many industrial applications, one of the primary advantages of using PID-based controllers is their simplicity, tunability, and ease of implementation. However, in the case of high-speed machines with magnetically suspended rotor systems, the stabilizing control solution often involves combining PID controllers with supporting filter structures. Depending on the case, this can lead to controllers with a significant number of tunable parameters, ranging from 10 to 35, which can be a challenging task when done manually. Therefore, a multiobjective genetic algorithm optimization is proposed in this paper to seek an optimal configuration for the controller parameters. This paper concentrates on optimizing PID-based controllers for AMB-suspended rotor systems, aiming to enable the analysis of outcomes within a standardized framework. Thus, the closed-loop performance is evaluated by the obtained damping properties and robustness. Moreover, an experimental AMB-rotor system is used to assess the performance of the controllers.
