Plant Model Parameterization for Accurate Simulation Setup
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current plant modeling methods, such as physical and Blackbox models, are complex and require extensive parameterization, often necessitating expert knowledge, which limits their accessibility and accuracy, especially in replicating behavior outside learned data sets.
Innovation Solution
A method and system for parameterizing plant models that automatically ascertain input and parameter values based on specified output variable values, using mathematical equations to map physical sub-processes, allowing for scalable and accurate simulations by non-domain users, even in rare operating points.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If physical models with detailed equations are used to accurately represent plant processes, then manufacturing precision and reliability are improved, but device complexity and difficulty of operation increase due to the large number of parameters requiring expert knowledge for setup
Solution Approach 1:
The patent creates simplified copies of physical models that replicate input-output behavior without requiring the full complexity of underlying physical equations. These surrogate models are trained on data from detailed physical models or experiments, enabling accurate simulations with reduced parameter requirements and simplified structure.
Solution Approach 2:
The patent transforms the model representation by changing from explicit physical parameters requiring expert knowledge to data-driven parameters learned from training data. This parameter transformation enables non-experts to use accurate models without needing to understand or configure complex physical parameters.
2Manufacturing precision
If physical models with many parameters are used to achieve accurate simulations, then manufacturing precision is improved, but ease of operation deteriorates because only domain experts can properly configure and use these models
Solution Approach 1:
The patent creates simplified copies of physical models that replicate input-output behavior without requiring the full complexity of underlying physical equations. These surrogate models are trained on data from detailed physical models or experiments, enabling accurate simulations with reduced parameter requirements and simplified structure.
Solution Approach 2:
The system performs automatic model configuration and parameter identification through data-driven methods. The surrogate models self-adjust their parameters during training, eliminating the need for expert intervention in model setup and making the tools accessible to non-expert users.
3Ease of operation
If Blackbox models are used to simplify model configuration and improve ease of operation, then device complexity is reduced, but manufacturing precision deteriorates because they cannot accurately replicate behavior outside their training data
Solution Approach 1:
The patent segments the modeling task into two parts: using simple Blackbox structures for ease of configuration and data-driven parameter identification for accuracy. This segmentation allows the model to maintain simple architecture while achieving high fidelity through learned parameters that capture complex physical behaviors.
Solution Approach 2:
The patent creates composite modeling approaches by combining the simplicity of Blackbox models with the accuracy of data-driven parameter identification. The resulting hybrid approach integrates the advantages of both simple configurability and high simulation accuracy across different operating conditions.
4Reliability
If detailed physical models are used to accurately map plant processes, then reliability is improved, but loss of time increases due to the laborious manual parameterization required
Solution Approach 1:
The patent performs preliminary data collection and model training during the model development phase. Surrogate models are pre-trained on comprehensive datasets that cover various operating conditions, so that during actual use, the models can be quickly configured without time-consuming manual parameterization for each new scenario.
Solution Approach 2:
The system performs automatic model configuration and parameter identification through data-driven methods. The surrogate models self-adjust their parameters during training, eliminating the need for expert intervention in model setup and making the tools accessible to non-expert users.
Data Source
AI summary
A method and system for parameterization of a plant model, wherein a model of a plant consisting of a plurality of model components, which represent physical sub-processes of the plant process and which correspond to physical models, is provided and at least one output value is specified in an operating point of the plant, where input values, parameter values and/or additional output values of the sub-processes are determined based on the model components and the at least one specified output value.