Modular Co-Simulation Using Predicted Interface Variables
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Co-simulation in mechatronic systems faces accuracy and efficiency issues due to incompatible subunits, leading to numerical artefacts and instabilities, particularly when subunits are simulated separately and communicate at discrete times, rather than continuously.
Innovation Solution
The method involves disaggregating the system into subunits, predicting numerical data using a data-prediction-model, delaying data transmission, and adjusting the simulation step size based on error and tolerance, ensuring smooth communication and accurate simulation by using polynomial functions and Hermite interpolation to maintain continuity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If subunits are simulated separately with discrete time communications, then co-simulation can handle incompatible platforms and modular structures, but numerical artefacts and instabilities occur due to discrete sampling
Solution Approach 1:
The patent implements continuous interpolation of interface variables between discrete communication time steps using polynomial functions (e.g., Hermite interpolation). This creates a continuous signal representation from discrete samples, eliminating numerical artifacts while preserving the ability to simulate incompatible subunits separately. The interface variables are continuously updated based on historical data and polynomial extrapolation, ensuring smooth transitions without discontinuities that cause solver restarts.
Solution Approach 2:
The patent performs preliminary prediction of interface variables using polynomial extrapolation based on historical communication data before actual communication occurs. This prediction allows receiving subunits to have accurate estimates of interface variables during the simulation step, maintaining simulation accuracy without requiring continuous real-time communication between subunits.
2Measurement precision
If communication frequency is increased to improve accuracy, then numerical artefacts are reduced, but processing time increases
Solution Approach 1:
The patent performs preliminary prediction of interface variables using polynomial extrapolation based on historical communication data before actual communication occurs. This prediction allows receiving subunits to have accurate estimates of interface variables during the simulation step, maintaining simulation accuracy without requiring continuous real-time communication between subunits.
Solution Approach 2:
The patent introduces an intermediary polynomial interpolation function that mediates between discrete communication events. This intermediary continuously generates accurate interface variable values without requiring frequent direct communication between subunits, thus maintaining precision while reducing communication overhead and processing time.
3Productivity
If higher time steps are used for data exchanges to improve efficiency, then processing time is reduced, but accuracy of the global solution deteriorates
Solution Approach 1:
The patent implements continuous interpolation of interface variables between discrete communication time steps using polynomial functions (e.g., Hermite interpolation). This creates a continuous signal representation from discrete samples, eliminating numerical artifacts while preserving the ability to simulate incompatible subunits separately. The interface variables are continuously updated based on historical data and polynomial extrapolation, ensuring smooth transitions without discontinuities that cause solver restarts.
4Adaptability or versatility
If discrete time communications are used between subunits, then modular simulation is enabled, but discontinuities cause untimely restarting of internal solvers
Solution Approach 1:
The patent implements continuous interpolation of interface variables between discrete communication time steps using polynomial functions (e.g., Hermite interpolation). This creates a continuous signal representation from discrete samples, eliminating numerical artifacts while preserving the ability to simulate incompatible subunits separately. The interface variables are continuously updated based on historical data and polynomial extrapolation, ensuring smooth transitions without discontinuities that cause solver restarts.
Data Source
Figure 1A
Figure 1B
Figure 2
AI summary
A computer-system (CPS) and a computer-implemented method for numerical modular simulation of a system (SYS), comprising: (a) disaggregating said system (SYS) into at least two subunit (SSY) simulation subsystems (SMN), (b) simulating the respective subunits (SSY) stepwise repeatedly generating subsystem-step-output (MSO) from subsystem-step-input (MSI) during a respective subsystem-time-step (SMP). To improve accuracy and performance said method comprises the additional steps: (c) transmitting subsystem-step-inputs (MSI) to a receiving subsystem (SMR) and simulating this subsystem (SMN) over a delay-time (DLT) before its subsystem-step-outputs (MSO) are generated, (d) receiving connection interface variables (TRD) from a sending subsystem (SMS) comprising at least one of: - numerical data (DTA), - at least parameters of a data-prediction-model (DEM) of said numerical data (DTA), - a data-prediction-model (DEM) assigned to said numerical data (DTA), (e) predicting said numerical data (DTA) by a data-prediction-model (DEM) over said delay-time (DLT) to obtain predicted numerical data (EDT) of said interface variables (TRD) provided by said sending subsystem (SMS), (f) starting the next simulation step of said receiving subsystem (SMR) generating the next subsystem-step-output (MSO) from subsystem-step-input (MSI), wherein said subsystem-step-input (MSI) comprises said predicted numerical data (EDT).