Linköping Electronic Conference Proceedings
Not a member yet
1113 research outputs found
Sort by
Prototypical Control for the Digital Twin of Aircraft Environmental Control System
Full text linkA digital twin of the overall aircraft EnvironmentalControl System is being developed as part of TheMa4HERA, alarge European Research Initiative. It shall supportverification and validation by virtually demonstrating thebehavior of the complete system in various conditions. Tothis end, also a prototypical control scheme needs to bedeveloped so that a dynamic simulation through completeflight missions is enabled. The prototypical control schemeis tuned using a simplified version of the Digital Twinwhich focuses on robustness and fast computation time,while making it robust enough to be stable when used withthe high-fidelity Digital Twin. An already stable andworking controller for simulating the detailed DigitalTwin, provides a significant gain of time, allowing forimmediate preliminary results on both steady-state andtransient system behaviors. This paper describes the basemodel of the Digital Twin and the methodology used todesign the prototypical control architecture
Aspects and Ideas for the FMI-based Modeling of Railway Digital Twins
Full text linkThis papers reports on activities in the European projectMOTIONAL that aims at the development of a digital twinenvironment which facilities the modularity,interoperability and composability of complex digital twinassemblies of railway systems. The approach that refers tothe Functional Mock-up Interface is justified by adiscussion of the comparable activities in industry and inthe automotive field compared to particularities in therailway system. The work was initiated by the selection andanalysis of nine use cases. An introductory digital twinexample illustrates the current implementation status andrelated aspects, while an outlook presents the integrationinto the Federated Rail Data Space as the business case andas a vision of the activity
Benchmarking the Modular Structural Analysis Algorithm
Full text linkIn a 2023 Modelica Conference paper, we proposed a novelmethod for the modular structural analysis of DAE systems,in which the structural analysis is not performed onflattened models, but rather at the class level. A newnotion of structural interface was proposed, in whichclasses are enriched with context information. That paperdeveloped our approach based on a few illustrative examples.In this paper, we provide the details of our algorithm. Itsperformance depends on the system architecture: theanalysis of models having a small number of classes (possi-bly instantiated many times), with a low treewidth systemarchitecture, scales up very efficiently with thisapproach. We then present additional benchmarks, amongwhich a urban heating network, a representative real-lifeexample on which a near-logarithmic scaling up is shown
A Dynamic Analysis of Refrigerant Mass in Vapor Compression Cycles
Full text linkNumerical simulation of a thermofluid vapor compressioncycle (VCC) model in Modelica, for example,can exhibit a variation in the total fluid (refrigerant)mass. This paper provides a dynamicanalysis of a commonly used VCC model, identifies andanalyzes the root causeof this variation, and proposes a number of remedies. Thecause lies within the dynamicequations that result from application of the principle ofmass conservation.In many common formulations, these equations express theconservation of mass asone or more differential equations that equate the timederivative of mass to zero.The resulting set of n ordinary differential equations (anda number of auxiliaryalgebraic equations) include the time derivative of a massconstraint function,but not the actual mass constraint function itself. As aresult, this modelingformulation has the following properties: (1) equilibriumsolutions of the systemare neither isolated, nor exponentially stable; (2) alinearization about any equilibriumsolution has at least one eigenvalue equal to zero, makingan equilibrium solutionstable, but not exponentially stable; (3) for a VCC modelformulated using two fluidstates per control volume, a one-dimensional equilibriummanifold exists containingall of the equilibrium solutions, and is parameterized bythe total fluid mass;(4) an (n-1) dimensional, stable, invariant manifold existstransverse to theequilibrium manifold, defined by the mass constraintfunction, and on which analyticsolutions to the model evolve and the total fluid massremains constant; and(5) numerical solutions may drift off of this manifold,resulting in an observeddrift of fluid mass. These properties have consequences forsimulation,control design, numerical model reduction, and stateestimation.A number of methods to stabilize the mass constraint areproposed and anumber of examples that illustrate the behavior, analysisand remedies are provided
Hybrid Simulation Models for Embedded Applications: A Modelica and eFMI approach
Full text linkHybrid simulation models combine physics equations withtrainable components to improve simulation results andperformance. Physics-enhanced neural ordinary differentialequations (PeN-ODE) are a promising type of hybrid modelsthat combine artificial neural networks (NN) with thedifferential equations of a dynamic system. Dynamicalsimulation models are often part of embedded controlalgorithms of cyber-physical systems (CPS); compliance withthe safety and real-time requirements of such embeddedenvironments is, however, challenging.In this work, we propose a workflow to incorporate trainedNNs in Modelica models to form hybrid simulation modelsthat are PeN-ODEs. We thereby focus on the transformationsteps from equation-based trained PeN-ODEs in Modelicatowards causal solutions suited for the embedded domain --up to and including MISRA C:2023 compliance checks andfinal software-in-the-loop (SiL) tests of generatedproduction code in the modeling environment -- for which weleverage eFMI standard compliant tools (Dymola and SoftwareProduction Engineering). It is of particular interest, howthe trained NNs of the hybrid model are implemented. Wepresent two approaches: (1) generation of C code usingexisting Open Neural Network Exchange (ONNX) tooling and(2) pure Modelica code with the tensor-flow represented asmulti-dimensional equations. Both approaches are discussed,highlighting why (2) is, in the long run, a better optiongiven the eFMI technology space
ShipSIM: A Modelica Library for Ship Maneuverability Modeling and Simulation
Full text linkThis paper introduces the ShipSIM, a novel free (standardconforming) Modelica library for modeling and simulation ofship maneuverability.This first Modelica library in the field of ship propulsionand maneuverability provides the components and flexibilityto model the ship propulsion, hydrostatics andhydrodynamics to develop ship maneuvering simulations fullycompatible with the Modelica Standard Library components.In this paper is presented the library’s key features andstructure and introduce the underlying physical andmathematical foundations and modeling approaches. Inaddition, current implementation status, applicabilitylimits and future development is discussed.The library development is coordinated by the authors andit is used in several MSc thesis. ShipSIM library isavailable on http://modelica-spain.org:3000/Basilio/ShipSIM
How Artificial Intelligence (AI) is Transforming the Aviation Industry
Full text linkAI is increasingly applied in aviation industry such as traffic management, predictive maintenance, flight operations and safety systems. The fast progress in AI increases the potential for improvements at all levels. Wise integration into aviation will enable improved performance, e.g. speed up the transition to reduction of environmental impact, reduce operational costs, and improve overall operation and safety.
We are conducting research in AI systems and developing “proof of concept” in a vide area of applications such as predictive maintenance, autonomous flying, hybrid human-AI systems, flight path planning, decision support, monitoring mental state (tiredness, stress, distraction, etc.). At the same time, it is a risk to become overconfident in AI systems and it is critical to develop safe and secure hybrid AI systems; there are many examples of naive deployment of AI where lack of understanding of the application domain together with lack of understanding of the different AI methods, techniques, and algorithms have led to serious events and even fatalities. Many AI systems we see today are by nature not fully trustable, since they heavily rely on statistical learning and lack reasoning capabilities and deeper understanding. We need to take this into account already when designing an AI system. Securing proper safeguards and validation already in the initial design phase is essential; otherwise, the system may become a dead end, unreliable, unscalable, or unsafe for deployment.
In conclusion, artificial intelligence is rapidly transforming the aviation industry in all areas, including safety, efficiency, and sustainability. We will over the next 5–10 years see more autonomous aircrafts and drones occupying our airspace, see an increase in safety, increased performance and reduced environmental impact as a consequence of increased deployment of AI. At the same time, we need to be in control and understand when, where and how to use AI, how much we can trust AI and how much responsibility we can delegate to AI
Validating the DLR Cables Library with Experiments and Parameter Optimization
Full text linkThe advantages of modelling and simulation are widely known: Optimizing systems before production, generating alternatives in a few clicks, reducing costs, monitoring, digital twin, etc. The quality of the simulation depends heavily on the quality of the modeling, making it an essential task. The DLR Cables library, which we presented in another work, allows the simulation of steel cables, focusing on use cases where their dynamic behavior is of interest, such as cranes and elevators, but also special motion systems using cables and amusement rides. There, the numerical approach based on finite elements is explained in detail and it is also shown that some simplifications are accepted in order to improve the computational effort. This paper presents the crucial tasks of validation and parameterization of the model, specifically focusing on the material properties of bending stiffness and bending damping. To achieve this, a series of experiments were carried out on four different cables. Optical systems are used to record the cables and to compare them with the simulation. For some of the experiments, we were able to show a good match between reality and simulation, but it also became clear that a linear approach may not be sufficient depending on the application
An Integrated Optimization and Orchestration Toolchain for Adaptive Optimal Control in Modelica Simulations
Full text linkThis paper introduces a novel Python-based toolchain, "OptiOrch", designed to enhance optimal control in Modelica-based simulations by integrating an optimization framework and an orchestration workflow. OptiOrch leverages the "MOO4Modelica" optimization framework, which supports both single- and multi-objective parameter optimization, and incorporates the "ModelicaOrch" orchestration workflow to dynamically adapt models based on real-time input data and goals. The toolchain features a user-friendly interface, feature model transformation, parallel computing, and automated workflow coordination, making it a powerful and generalized solution for various applications. Practical examples and a case study demonstrate how this toolchain can be effectively applied to Modelica systems for optimal control
Vehicle Health Monitoring for Driving Safety using Co-simulation between Dymola and Simulink
Full text linkA vehicle dynamics model-based health monitoring process is presented to enhance driving safety. The vehicle model can simulate driving by reflecting degradation performance of suspension and tires. The model was developed using Dymola, and driving simulation was performed by integrating the lane keeping assistant system with the vehicle model using Simulink. The degradation behavior was monitored with k-nearest neighbor and Gaussian mixture model. The remaining useful life for vehicle components was predicted using Gaussian process regression. The proposed method predicts remaining useful life with a 95% confidence level for vehicle components to improve safety for driving