Hybrid Simulation Model Tuning for Accurate Machine Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing computer simulations for machine control face inaccuracies due to lack of measurement data for training black box models, leading to uncertainties in model-based control solutions.
Innovation Solution
A system and method for generating accurate computer-aided simulation models using preconfigured model components, sensor data, and adaptive parameter tuning to create a hybrid simulation model that accurately represents machine behavior.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If black box models are used for machine elements with unknown physical properties, then the model can be created without detailed physical knowledge, but the model accuracy deteriorates due to lack of measurement data for training
Solution Approach 1:
The machine is divided into multiple machine elements, each modeled separately using appropriate modeling approaches (physical models for well-understood elements, black box models for elements with unknown properties). This segmentation allows targeted application of different modeling strategies to optimize overall model accuracy while maintaining ease of creation for complex elements.
Solution Approach 2:
The system adapts model parameters based on available sensor data from the actual machine operation. By continuously adjusting parameters of black box models using real operational data, the model accuracy improves over time without requiring complete physical understanding of all machine elements.
2Measurement precision
If physical models are used for machine elements with well-known physical relationships, then the model accuracy is improved, but the device complexity increases due to need for detailed physical knowledge and equations
Solution Approach 1:
The system segments the machine into elements based on the availability of physical knowledge. Only elements with well-understood physics receive detailed physical models, while other elements use simpler black box models. This selective segmentation reduces overall model complexity while maintaining accuracy where physically justified.
Solution Approach 2:
Different modeling approaches are applied locally to different machine elements based on their specific characteristics and the availability of physical knowledge. This local differentiation ensures that complex physical models are only used where necessary, reducing overall system complexity while maintaining local accuracy where needed.
3Measurement precision
If hybrid models combining physical and black box models are used, then the overall model accuracy is improved, but the difficulty of detecting and measuring increases due to integration of different modeling approaches
Solution Approach 1:
The system employs a universal graphical modeling interface that can represent both physical models and black box models using the same component types and connection methods. This universal interface simplifies the integration process by allowing different modeling approaches to be combined without requiring different representation methods, reducing the difficulty of detecting and measuring model behavior.
Solution Approach 2:
The graphical modeling tool acts as an intermediary that standardizes the integration of diverse model components. It provides unified methods for connecting physical and black box models, handling data flow consistency, and managing parameter propagation across different model types, thereby reducing the complexity of integrating heterogeneous modeling approaches.
Data Source
Figure 1
Figure 2
AI summary
The invention relates to a system for controlling a machine, wherein the machine comprises machine elements and wherein for some of the machine elements are computer-aided modelled using preconfigured model components, the system comprising: • a graphical generator configured to generate a computer-aided simulation model for modelling the physical behaviour of the machine using the preconfigured model components for the machine elements, wherein for at least one selected machine element a respective model component is to be configured, • an analysis unit configured to determine interfaces of the preconfigured model components of the machine elements adjacent to the selected machine element, • a database configured to provide a plurality of model component templates for computer-aided modelling of at least parts of the machine elements, • an input unit configured to read in a predetermined specification determining the mapping of the physical behaviour of the selected machine element, • a selection unit configured to select a model component template from the database for the selected machine element based on the interfaces of the preconfigured model components and based on the specification, and to provide the selected model component template to the graphical generator to insert it into the simulation model, • at least one sensor configured to receive sensor data of the machine, • a configuration unit of the graphical generator configured to adapt parameters of the selected model component template using the sensor data such that when running the simulation model it reproduces the sensor data, and • an output unit configured to provide the simulation model for controlling the machine.