ABSTRACT. Physics-based realistic simulation environments are vital for advancing research and education in AI, providing an accurate and controlled platform for testing algorithms and models. These environments simulate real-world physics, including dynamics, collisions, and sensor interactions, allowing AI systems to learn and adapt in complex, lifelike scenarios. In research, they enable experimentation with AI-driven robotics, autonomous systems, and reinforcement learning without the constraints of physical setups. For education, they offer hands-on experiences for students to explore AI concepts and algorithms in dynamic environments, bridging the gap between theoretical learning and practical application, fostering innovation and understanding. During this talk, we will present our efforts in the development of relevent simulation environments within the Robotics and AI group, at Lulea University of Technology, Sweden. These environments are used in the ongoing coruces as well as the advancement of autonomy algorithms towards their field implementation.

ABSTRACT. The European Union (EU) has set an ambitious target to reach carbon neutrality by 2050, prompting industries to develop roadmaps to achieve this goal. In this context, hydrogen and hydrogen-based fuels play a crucial role in achieving net-zero emissions. Instead of relying on hydrogen production from steam reforming natural gas systems, electrolysers offer a sustainable alternative to address climate and energy challenges. The integration of solar energy systems with electrolysers can further diminish carbon emissions and enhance sustainability. Typically, these processes are simulated using process simulation software platforms that employ first-principle models based on the mass and energy balances within the system. The adoption of Model Predictive Control (MPC) algorithms not only benefit from improves advance control methods and optimization but also facilitates the automation and efficient operation of these processes. This study aims to mathematically model and simulate an integrated photovoltaic (PV) and Proton Exchange Membrane (PEM) water electrolyser system for hydrogen production. Additionally, it assesses the impact of MPC algorithms on the system's efficiency. The study undertakes modeling of PV systems incorporating a maximum power point tracking algorithm to capitalize on optimal power generation and ensure a consistent direct current supply to the electrolysers. Mathematical modeling of PEM water electrolysers is performed to establish the current-voltage curve in steady-state mode and to assess water and gas permeability through the membrane in dynamic mode. Finally, the study identifies input variables, such as electrolyser temperature, and evaluates their effects on key indicators like system efficiency through performance analysis.

ABSTRACT. Valorisation hemicellulose into furans chemicals is of great interest to create sustainable furan alternatives to fossil-derived products. A route of particular interest is acid-catalysed dehydration of the hemicellulose pentoses in aqueous medium, with simultaneous extraction of furfural using organic solvent. Agitated Cell Reactor (ACR) could be effectively used to intensify this process and decouple mixing from the long reaction time. This study presents a mathematical model for dehydration of C5 sugars to produce furfural in an ACR. The model can be used to study the effect of feed concentration to the product properties, the concentration profiles along the reactor length, and the dynamic behaviour of the system under feed disturbances or flow rate adjustments. The model was successfully fitted to the experimental data of a laboratory scale ACR for the target product. A simulation study was conducted to analyse the controllability of the process. Operability analysis with the nominal input space and the design space was used for mapping the most feasible region for the process design to meet the flexibility or controllability already at the design phase of the reactor system.

ABSTRACT. Autonomously driving vehicles and robots that drive in public environments need to be safe and reliable under all weather conditions, including arctic winter conditions. Digital twins provide an opportunity to test autonomous vehicles in a safer, faster, and less expensive environment than carrying out tests in real-life conditions.

We developed the data connection via ROS (Robot Operating System) between a mobile robot and its digital twin. This allows for almost real-time exchange of commands, information, and sensor data between the twins.

The digital twins of the robot and the testing ground are constructed in a CARLA-based autonomous driving simulator, which simulates realistic arctic winter weather conditions.

The digital twin design was informed by the intended future use cases: Testing, optimizing, controlling, and monitoring autonomous driving and snow cleaning functions first with the digital twin, then in hybrid approaches.

In our test setup we tested the hybrid case, where both robot twins were moving in the simulation and the real-world test area at the same time. We verified our digital twin, assessed delays, and the applicability in the intended use cases.

Our results show that the digital testing ground would profit from inbuilt reference points to examine the alignment with its real-world counterpart. The communication via ROS was occurring in almost real-time, therefore, the digital twin setup was found to be applicable in hybrid digital twin testing.

In the future, we will introduce an autonomous car into this digital twin setup and equip the testing ground with a 5G network.

ABSTRACT. In this paper we apply Augmented Extended Kalman filters (AEKFs) to perform parameter estimation in two different biological controller motifs under both noise-free and noisy conditions. Based on measurements of the two states of the controller motifs, we show that under both noise conditions it is possible to estimate all 5 and 6 parameters, respectively, which is in accordance with previously published results that investigated the theoretical concept of structural identifiability. We further investigate how the level of process/measurement noise and the initial estimates of both the parameters and states in the AEKFs affect the estimation performance, and the results indicate that the degree of non-linearity affects filter performance.

ABSTRACT. This research explores the feasibility of using excess heat from data centers for biomass drying, enhancing the biomass energy value. A predictive model was developed to estimate exhaust air humidity from the dryer, indirectly indicating biomass moisture. Machine learning techniques, including linear regression model (LM), gradient boosting machines (GBM), eXtreme gradient boosting (XGBoost), random forest (RF), and multilayer perceptron (MLP), were used. Tree-based models GBM, RF, and XGBoost achieved a coefficient of determination (R²) of 0.88–0.89. Methods were enhanced with transparency through explainable artificial intelligence (XAI) techniques, which facilitated the analysis and visualization of humidity fluctuations. Key factors affecting drying efficiency include weather conditions, supply air humidity, and fan speed. The study provides actionable insights for optimizing the drying process, improving system air tightness, and advancing sustainable energy utilization through AI-driven solutions. The developed model enables future dynamic control of drying processes.

ABSTRACT. Gas turbines are vital in power generation and propulsion systems. However, these engines are exposed to complex and variable operating conditions, which makes early and accurate fault detection essential for predictive maintenance and minimizing unplanned downtime. This paper proposes a novel approach that combines convolutional neural networks (CNNs) with transformer architectures to address these challenges. The proposed Convolutional transformer model aims to enhance the accuracy and robustness of turbofan fault classification by integrating the feature extraction capabilities of CNNs with the contextual learning strengths of transformers. Through rigorous experiments, we seek to demonstrate our approach's performance in classification accuracy and generalization across different operating conditions. We utilize a comprehensive dataset from multiple aircraft engine units as the benchmark for this study. The results for the proposed model show an accuracy of 99.6% on the test dataset. The outcome has the potential to be extended and fine-tuned for different types of gas turbines for diverse applications.

ABSTRACT. Reconfigurable intelligent surfaces (RISs) have emerged as a promising solution that can provide dynamic control over the propagation of electromagnetic waves. The RIS technology is envisioned as a key enabler of sixth-generation networks by offering the ability to adaptively manipulate signal propagation through the smart configuration of its phase shift coefficients, thereby optimizing signal strength, coverage, and capacity. However, the realization of this technology's full potential hinges on the accurate acquisition of channel state information (CSI). In this paper, we propose an efficient CSI prediction framework for a RIS-assisted communication system based on the machine learning (ML) transformer architecture. Architectural modifications are introduced to the vanilla transformer for multivariate time series forecasting to achieve high prediction accuracy. The predicted channel coefficients are then used to optimize the RIS phase shifts. Simulation results present a comprehensive analysis of key performance metrics, including data rate and outage probability. Our results confirm the effectiveness of the proposed ML approach and demonstrate its superiority over other baseline ML-based CSI prediction schemes such as conventional deep neural networks and long short-term memory architectures, albeit at the cost of slightly increased complexity.

ABSTRACT. Energy efficiency, scalability, and reliability are increasingly important for sustainable data centers. In this paper, we focus on forecasting real-world resource usage using neural network time series models, specifically utilizing convolutional recurrent Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) architectures. In our analysis, we compare LSTM and GRU in terms of forecasting accuracy and computational complexity during model training. We demonstrate that recurrent neural networks are more accurate and robust compared to the traditional Autoregressive Integrated Moving Average (ARIMA) time series model in this complex forecasting problem. GRU achieved a 9% reduction and LSTM a 5% reduction in forecasting Mean Squared Error (MSE) compared to ARIMA. Furthermore, the GRU architecture with a 1D convolution layer outperforms LSTM architecture in both forecast accuracy and training time. The proposed model can be effectively applied to load forecasting as part of a data center computing cluster. In this application, the proposed GRU architecture has 25% fewer trainable parameters in the recurrent layer than the commonly used LSTM.

ABSTRACT. This study presents a kinetic reaction modeling method for direct air capture (DAC) process of CO2 adsorption using computational fluid dynamics (CFD). Here, CO2 is adsorbed by amine coated airsurface contact area. The Langmuir model is employed to represent the kinetics of CO2 adsorption. Despite neglecting the diffusive phase of the adsorption, which is dominant only in the later stages of adsorption, the surface reaction model gives a satisfactory representation of the adsorption for a major part of the process. Honeycomb reactors with coated adsorbent may yield a better control of reaction rate and pressure drop compared to commonly used packed bed adsorption columns. Their enhanced performance in distributing the flow homogeneously between and within channels creates unique features for the reactor. In this study, we have analyzed mechanical and electrical energy demand for adsorbing CO2 per unit mass of adsorbed CO2 as a function of air flow rate. Adsorption performance of honeycomb structure is anticipated to significantly improve in comparison to the packed beds.

ABSTRACT. The objective of this paper is to investigate the fluid flow and conjugate heat transfer in a 2D channel using lattice Boltzmann method (LBM). In this work, fluid flow and heat transfer are studied for the Reynolds numbers varying between 250 and 1000. The working fluid in the simulations is air with the Prandtl number of 0.72. At the Reynolds number of 600, the effect of different conductivity ratio (1, 10, 100, 400) between solid and fluid are investigated. Furthermore, at this Reynolds number, the distance between obstacles for the conductivity ratio of 10 is evaluated. The results show that any increase in Reynolds number leads to a heat transfer improvement. Moreover, increase in the conductivity ratio leads to an isothermal surface and enhanced heat transfer. The more the distance between the obstacles, the better the heat transfer rate. The results obtained from LBM are in good agreement with experimental and conventional computational fluid dynamics methods.

ABSTRACT. Circulating fluidized beds is one of the emerging technologies to convert waste to energy and also an attractive method on a large scale. Key components such as the loop seal, gas distributor and cyclone separator play pivotal roles in facilitating solid recirculation, heat transfer within the system. This study focuses on the design and optimization of a CFB reactor using data derived from Barracuda Virtual Reactor software. Initially, data from a small-scale CFB reactor with main dimensions of 84 mm diameter and a loop seal diameter of 34 mm is utilized for simulation validation. By comparing simulation results with experimental data, the accuracy and reliability of the computational model are ensured. Subsequently, different reactor models are constructed and analyzed to explore various configurations and operating conditions. The results obtained from simulation-based design and optimization provide valuable insights into achieving the optimal performance of the CFB system. By refining geometry, the efficiency and reliability of the reactor can be significantly enhanced. Overall, this study contributes to advancing the understanding and application of CFB technology in waste-to-energy conversion and large-scale industrial processes.

ABSTRACT. Fluidized bed technology known for its efficient heat and mass transfer and controlled material handling, is widely used across industries. However, CFD simulation of fluidized beds presents challenges that require extensive validation. This study leverages the Multiphase Particle-In-Cell (MP-PIC) method, a recent Lagrangian modeling technique to improve computational efficiency and accuracy. The CAD model was developed using SolidWorks 2020 and simulation was carried out in the commercial CFD package Barracuda VR 21.1.0. The sensitivity of grid size, drag models and the impact of recirculating pipe height after loop seal was examined. Sand particles 63-200 μm and air were used as bed material and fluidization gas respectively achieving full flow circulation at 650 SL/min and 12 SL/min aeration in the riser and loop seal. A total of 19 different simulations were conducted, varying grid size and drag models each for a duration of 45 seconds with a time step of 0.0005 seconds. Pressure transducers along the CFB walls provided validation data. The Wen-Yu Ergun drag model showed a minimal error margin of 0.60%, followed by the Wen-Yu 80000 model at 0.62%, demonstrating high predictive accuracy.

ABSTRACT. Improving the efficiency of oil recovery is a crucial necessity in the current energy landscape. The widespread adoption of advanced wells, equipped with Autonomous Inflow Control Devices (AICDs), represents a leading strategy for this purpose. However, the absence of a predefined and straightforward option for modeling advanced wells in dynamic multiphase flow simulators like OLGA® poses a significant challenge. To address the issue, this paper proposes a novel approach based on developing a mathematical model derived from experimental data characterizing the AICD behavior. The Algebraic Controller option in OLGA is then leveraged to integrate the AICD effects into the simulation seamlessly. The proposed methodology undergoes rigorous testing on the PUNQ-S3 reservoir model as a benchmark case study with Water Alternating Gas (WAG) injection. Results demonstrate that AICD has a better water reduction rate of 36.3% and 3.7% compared to OPENHOLE and ICD. This result also indicates the accurate modeling and simulation of AICD performance in the software, showcasing the effectiveness of the developed mathematical model. Comparative analyses of advanced wells with different Flow Control Devices (FCDs) underscore the conclusion that AICDs significantly enhance oil recovery efficiency, thereby maximizing profit and minimizing the carbon footprint.

ABSTRACT. The utilization of advanced multilateral wells to enhance well-reservoir contact, coupled with water injection, stands out as a common approach to boost oil extraction efficiency. It is imperative to develop precise, fully integrated, dynamic, well-reservoir models tailored for this type of oil recovery to enhance the design of advanced multilateral well completions. This study addresses the challenge by constructing a well model using OLGA®, which is, a dynamic multiphase flow simulator, and a reservoir model using EclipseTM, a reservoir simulator. Subsequently, these models are seamlessly integrated to perform comprehensive simulations. The proposed approach is tested on a case study involving oil recovery through an advanced multilateral well completed with various Flow Control Devices (FCDs) supported by water injection. Results from the simulations demonstrate the success of the integration approach, offering a reliable method for accurately modelling oil recovery from advanced multilateral wells to improve oil recovery. Notably, according to this study, wells completed with Autonomous Inflow Control Valves (AICVs) exhibit superior performance, optimizing oil recovery with a reduced carbon footprint.

ABSTRACT. The manufacturing industry is in a strong transition towards digital, intelligent, and sustainable manufacturing. However, SMEs in the manufacturing industry often lack the resources and know-how to utilize digital tools as part of their R&D activities. Thus, there is a need for concrete examples to show the benefits of these tools. This paper discusses a demonstration of warehouse optimization where a genetic algorithm is applied to optimize pallet transfers. The simulation model of an FMS cell includes a warehouse with nine Euro pallets and a stacker crane. Visual Components was used for the simulation and an external Python application for the algorithm. As a result of the optimization, the total duration of the transfers was reduced by approximately 20 seconds. The demonstration has been used to showcase the integration of optimization methods into simulation technology and has ignited longer-term collaboration with the local industry on the same theme.

ABSTRACT. This paper will describe the Autonomous Mill of the future as a mill that benefits from the use of Digital Twins utilizing a Process Model coupled with a Control Model of the real-time Control System to allow the Autonomous Mill to “run itself” with little or no human intervention. This paper will then give an overview of the unit operations and equipment common to pulp and paper mills and conclude with several examples of specific opportunities where control systems optimization through Advanced Process Control (APC) and Model-Based Predictive Control (MPC) can increase production; reduce costs, and autonomously operate the mill of the future. The pulp and paper mill is often divided into six main “islands” of automation; raw material receiving and preparation (the woodyard), the pulp mill, the powerhouse, the paper mill, converting and finishing, and effluent treatment. Each of these islands presents their own, unique set of unit operations; but, perhaps not surprisingly, you can see similar unit operations in various industries besides pulp and paper. For example, the powerhouse equipment, besides the main difference being that the fuel is “black liquor”, the equipment can be found in any other industrial power plant. In the paper machine “island”, the use of cascaded variable-speed drives to control the paper sheet tension is also seen in the draw line of a steel, textile, or fiber mill. And, as a final example, the effluent treatment facility of the paper mill has many of the same equipment you will find in a municipal water/wastewater plant.Several examples of specific control systems optimization included for each of these “islands” include chemical savings in the lime kiln and causticizing, pulping, screening and refining, washing, and bleaching processes of the pulp mill; energy savings in recovery boiler sootblowing and the lime kiln, pulp stock preparation including cleaning and refining and the paper pressing and drying sections of the paper mill; and the environmental savings involved in effluent treatment and recycling water.

ABSTRACT. This study conducts a sensitivity analysis to evaluate the influence of varying data volumes on model performance within multi-product batch processes in the iron and steel industry. Nine machine learning models, encompassing both ensemble and parametric methods, were rigorously tested using a data withholding approach. The results demonstrate that ensemble models, particularly Random Forest and Gradient Boosting, consistently outperformed parametric models across different data volumes, showcasing superior generalisation and robustness to outliers. These findings underscore the importance of careful model selection and comprehensive data preprocessing in enhancing model performance and suggest that ensemble methods are particularly well-suited for complex industrial applications where data quality and volume are critical.

ABSTRACT. Most engine models are developed for control purposes and, in some cases, hard coded with a single engine type usage in mind. This is a problem since a new model is also needed when new engines are developed, as it usually takes less time than changing or modifying the old one. To facilitate a more rapid development process, there is a desire to have control-oriented models that can be adapted to new types of hardware with ease while at the same time providing fundamental insights into the physics of the engine that limit the control performance. These objectives are fulfilled by creating an open-source mean value MATLAB/Simulink model, a generic engine model with parametrization and compatibility with both VVT/VVA and SI/CI combustion. The main idea is to build on a component-based structure where the components are designed to be reused for similar processes. The engine model models the air filter, intercooler, and exhaust system components as incompressible flow restrictions. Bypass, throttle, intake/exhaust valves, and wastegate are modeled as compressible flow restrictions. Adiabatic control volumes are placed between each component to keep track of masses, pressures, and temperatures. The few remaining components are modeled separately, with unique functions for each model. As a concept demonstration of the generality of the approach, two engines, a 6-cylinder 12.7-liter Scania diesel engine and a 4-cylinder 2.0-liter Volvo petrol engine, are used as case studies where the generic simulation platform is parameterized and validated against experimental data for both engines.

ABSTRACT. In this work, an analytical tire deformation model is created, which can be parameterized using simple measurements. The model consists of three equations which are solved to provide a shape function for the tire.

This model can be used to provide excitation input for energy harvesters embedded inside the tire for example in FEM simulations. Additionally the model can be used in differential equation based simulations for quick parameterized simulations. With this model it is possible to study the effect of tyre inflation state to the energy harvesting performance of the system.

Two different simulation cases are presented in this work. First is a vibration energy harvester simulation using the model with an inertial energy harvester. The second case illustrates an energy harvester using the deformation of the tire as the excitation for the energy harvester as opposed to inertial type harvester.

ABSTRACT. Research and innovation in Vehicle-in-the-loop (ViL) testing is garnering more attention than ever. Integrating cyber-physical systems (CPS) into the ViL setups further enhances their functionality and hybridity. Setting up any ViL infrastructure involves substantial investments and thus requires critical analysis of the resources to achieve the intended results. This study focuses on such a ViL testing infrastructure development at NUVE-LAB, aiming to provide state-of-the-art facilities for hybrid automotive testing. The facility includes physical components such as a heavy tractor (Valtra), dynamometers, an APGS system, and battery emulators (BE), complemented by digital twins (DTs) of each physical machine, process, and environment to automate the testing facilities. This research examines various interoperability challenges within the current ViL framework. Three distinct testing scenarios are created to assess the overall functionalities of the hybrid setup: dynamometer-in-the-loop, APGS-in-the-loop, and BE-in-the-loop. Analyzing individual cases highlighted the need for different modeling and simulation (M&S) tools to develop digital twins. Among the tools, SIMULINK is used to build and refine the models of DTs, whereas MATLAB is used to develop control algorithms. The study also explores the adoption of Functional Mock-up Interface (FMI) standards to facilitate seamless interoperability among modeling and simulation tools. Additionally, the potential integration of the Eclipse Arrowhead framework (EAF), an IoT middleware, is discussed to enhance efficient data management, service interoperability, and the integration of various cyber-physical system components. In conclusion, this paper outlines the interconnection of the digital and physical platforms to evolve a hybrid ViL test laboratory, envisioning the future trajectory of the NUVE-LAB.

ABSTRACT. Reactivity Controlled Compression Ignition (RCCI) stands out as a promising combustion method for the next wave of internal combustion engines, offering cleaner and more efficient operation, particularly in heavy-duty engines. A key approach within this strategy involves pairing diesel as the high reactivity fuel with natural gas (NG) as the low reactivity counterpart. Further optimization can be achieved by introducing hydrogen to replace portions of NG, thereby enhancing combustion quality while reducing greenhouse gas emissions. For accurate numerical simulation of engines employing this strategy, a specialized chemical kinetics reaction mechanism tailored for internal combustion engines becomes essential. To facilitate computationally efficient 3-D Computational Fluid Dynamics (CFD) simulations, the mechanism undergoes reduction, yielding a manageable number of species and reactions. in the present work, n-heptane serves as the surrogate for diesel fuel and methane and ethane represent NG. The mechanismh is meticulously tuned using optimization methods to align with available experimental data on ignition delay time (IDT) and laminar burning velocity (LBV) of the fuel blends. Validation of the predicted results from the mechanism is conducted against experimental IDT and LBV data. Subsequently, 3-D CFD simulations are employed to verify engine operating parameters in dual fuel and RCCI engines, ensuring alignment with available experimental data.

ABSTRACT. The electrification of heavy vehicles and work machinery is developing rapidly. The main motivators are green transition and requirements from the customers. In Finland, there are many high-tech market-leading companies in this segment. Mass-produced equipment and machines are suitable for general applications and thus tailoring design for specific conditions and/or needs results in better productivity and efficiency. In heavy electric vehicle applications, the challenge is to make new products economically viable and configure them to meet customer needs. In these applications, the number of solutions is an order of magnitude higher than in traditional mechanical solutions. However, electronic solutions enable new features and energy efficiency improvements to have measurable benefits in the application. The research investigates the effects of electric axle solutions for hybrid heavy duty vehicles. Modelling and simulations consider both the effects of engine and usage of battery charge and surroundings of vehicle, for example road profile, traffic, outdoor temperature, and friction. A system level model of a vehicle has been utilized to simulate its longitudinal dynamics interacting with estimated surroundings by utilizing trajectory optimization followed by model-based control. The planned route based on trajectory optimization can be made further favourable by utilizing real-time model predictive control (MPC) receiving online data from changing conditions. MPC gives new suggestions for optimal battery usage based on deviations from the best matching model from a database. Control strategy is important when considering economic benefits for a hybrid heavy duty vehicle with a high degree of freedom in system design.

ABSTRACT. Increasing demand for critical raw materials and energy transition metals sets new targets for the mineral processing, also resulting as higher requirements for the simulation tools during process design and optimization. This study presents a framework for global uncertainty evaluation of modelled plant-wide processes, where the propagation of uncertainty sources is addressed. The uncertainties exist, for example in operational and design parameters and in material properties. The approach was demonstrated with a typical mineral processing flowsheet simulated with commercial software. First, domain knowledge was adopted to screen the parameter space and then Monte Carlo simulation was performed. After this, the generated data set was used to identify surrogate models between the uncertain inputs and process performance indicators. Finally, a global sensitivity analysis was conducted to identify the effects of uncertainties to the decision-making in process design. The results were particularly used to locate the process points where accurate information is needed for the robust process design, or where on-line measurements would be preferred to establish on-line optimization.

ABSTRACT. The use of mathematical methods in simulation and optimization models is widely spread to solve the current and future problems of an efficient and sustainable energy supply. Especially MILP is commonly used for industrial and municipal energy systems, where hourly resolved demand profiles are addressed in the time frame of one year in a quasi-stationary optimization. Certain technical or regulatory circumstances make necessitate considering all time steps in one coupled optimization problem. This results in a level of model complexity where today's solvers often struggle to find a solution within a reasonable timeframe.

Application examples are annual maximum runtime restrictions or finding the optimum loading strategy of a seasonal storage. Regulatory examples in Germany include the full-load hour restricted CHP-surcharge, the high-efficiency-criterion or the maximum emission of a CO2-Budget which lead to an annual integral limitation.

In this work, we present a two-stage approach with a simplified year-round-coupled first stage and a fully resolved second stage with a rolling horizon. To compress the input data in the simplified first stage while maintaining the order of the time sequence, we use different resolutions of downsampling and LP-relaxation. For the second stage, we derive corresponding additional boundary conditions and evaluate these in this study. Various use-cases involving both MILP and MIQCP models are evaluated using different compression parameters. The aim is to achieve high accuracy while saving computation time and furthermore enabling the solution of problems that would otherwise be computationally unsolvable without this method.

ABSTRACT. This paper provides a comparative overview of four numerical methods widely employed in computational fluid dynamics and related fields: Finite Volume (FV), Lattice Boltzmann Method (LBM), Smoothed Particle Hydrodynamics (SPH), and Spectral Methods. FV discretizes the domain into control volumes, emphasizing conservation laws and flux integrals across cell faces. It's renowned for its robustness, particularly in complex geometries. LBM is a mesoscopic approach simulating fluid dynamics through particle interactions on a lattice grid. Its intrinsic parallelism and ability to handle complex boundary conditions make it suitable for multiphase flows and porous media simulations. SPH represents fluids as a set of particles, where properties are smoothed over neighboring particles using a kernel function. SPH excels in free surface flows, astrophysical simulations, and fluid-structure interaction due to its Lagrangian nature and adaptive resolution. Spectral Methods discretize functions using orthogonal basis functions, such as Fourier or Chebyshev polynomials, enabling high-order accuracy and spectral convergence. They are preferred for problems with smooth solutions and periodic boundary conditions, like turbulence simulations and wave propagation.

ABSTRACT. Nonlinearities become essential in various systems when the operating area widens. The linear models are special cases for narrow areas. The behaviour can become gradually steeper or flatter depending on the case. These differences can be seen in data distributions which can be used in analysing the nonlinearities. The analysis is valid for the chosen operating area. Further widening requires a new analysis. The widely used Gaussian distribution is seldom valid for a wide area. The behaviour is often asymmetrical. The variable specific scaling can be presented with two second order polynomial defined by five parameters which can be obtained with data analysis. The only requirement is that the functions need to be monotonously increasing. The parameters can be interpreted as the operating point and four corner points of the feasible range. This provides a good solution for combining expert knowledge with the data-based analysis. If the nonlinearities are analysed correctly, only linear interaction are needed in the models. As the analysis is based the same methodology in all these applications, monitoring of the machines can be combined with process data. Smooth operation and high quality of products is the main goal of all these applications, and this can be achieved by combining these indicators with process control in the same way as it has been one for smaller indicators used in lime kiln control and water treatment.

ABSTRACT. Previously presented method of calculating local average gradients for solving partial differential equations (PDEs) is enhanced by accelerating it with GPUs and combining a previous technique of interpolating between grid points in the calculation of the gradients instead of using interpolation to create a denser grid.

For accelerating the calculation with GPUs, we have ported the original naive Matlab implementation to C++ and CUDA, and after optimizing the code we observe a speedup factor of two thousand, which is largely due to the original code not being optimized.

Combining the interpolation with the calculations of the gradients allows one to use the original coarser grid in the calculation which results in over a factor of 4 speedup. Additonally, the new algorithm has the added benefit that second order partial derivatives can be calculated from the same local information that first order partial derivatives are calculated, which effectively halves the needed memory traffic for calculating the gradients.

ABSTRACT. Offshore steels are engineered to have an outstanding combination of high strength and toughness to withstand extreme conditions needed in offshore and marine applications. Low carbon and alloying content ensure the weldability of these steels. Ferrite and bainite generally form the base metal microstructure of 500 MPa offshore steels, whereas acicular ferrite is commonly produced in the weld metals used with these steels. Additionally, multiple welding passes may be needed when welding thick steel sections. In this case, microstructures of pre-existing passes are affected by the thermo-cycles caused by the subsequent passes. These heat-affected zones (HAZ) in the weld metal are less studied than the HAZs in the base metal next to the weld. HAZs after real welding process are relatively narrow and that way challenging to study and test reliably. Physical simulation provides an opportunity to produce different kind of HAZs on sufficiently large area for various types of microstructural and mechanical properties characterization. Moreover, the effect of different welding methods and parameters can be easily studied by adjusting the simulation settings. Therefore, the aim of this study was to produce the coarse-grained (CGHAZ-W), intercritical (ICHAZ-W) and intercritically reheated heat-affected zones of the weld metal (ICCGHAZ-W) using physical simulation. Submerged arc welding (SAW) method was used to produce the original weld. HAZs with 2 different cooling times from 800 °C to 500 °C (t8/5 = 5 and 30 s) were simulated utilizing Gleeble 3500 thermomechanical simulator. Microstructures were characterized using a Zeiss Sigma field emission scanning electron microscope. The results indicated that the original weld metal contained acicular ferrite nucleated on oxide inclusions, and thermal cycles induced microstructural changes in the weld metal, with each simulation variant resulted in distinctive features. Microstructures obtained by the physical simulation were supported by the numerical simulation results carried out by JMatPro software.