Compensating First Principle Simulation Models via Error Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
First principle-based process simulation models are prone to modeling errors due to their inability to account for uncertainties in process equipment characteristics, and empirical data-based models require time-consuming and costly updates when equipment changes occur.
Innovation Solution
Compensating first principle-based simulation models by generating an error model using test input and output data, which adjusts the model outputs to match the actual process system outputs, thereby improving accuracy and flexibility.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If first principle-based models are used to model process equipment, then the models can be easily updated when equipment changes occur, but the models suffer from modeling errors due to inability to account for uncertainty in equipment characteristics
Solution Approach 1:
The patent combines first principle-based models with empirical data-based models into a hybrid modeling approach. The first principle model provides the structural framework that can be easily updated when equipment changes, while the empirical model captures uncertainties and deviations from ideal behavior. By merging these two modeling paradigms, the system achieves both adaptability to equipment changes and reliability in predicting actual process behavior.
Solution Approach 2:
The patent applies local quality by using the first principle model for the overall process structure where general physical laws apply, while introducing empirical corrections at specific locations where uncertainties exist. The empirical model focuses on capturing local deviations and uncertainties in equipment characteristics, allowing the majority of the system to rely on the adaptable first principle framework while correcting specific areas of inaccuracy.
2Reliability
If empirical data-based models are used to capture dynamic transient phenomena, then the models can better represent actual process behavior, but new models must be developed when equipment changes occur which is time consuming and costly
Solution Approach 1:
The patent segments the modeling approach into two distinct components: a first principle model that handles the overall process structure and can be easily reconfigured when equipment changes, and an empirical model that specifically captures dynamic transient phenomena. This segmentation allows each component to fulfill its specialized function while minimizing the overall effort required when equipment changes occur.
Solution Approach 2:
The patent introduces dynamics by allowing the empirical model to adapt to changing equipment conditions through parameter adjustment rather than complete model redevelopment. The empirical component can be tuned to reflect new equipment characteristics while maintaining the overall model structure, reducing the time and cost associated with model updates when equipment changes.
3Measurement precision
If trial and error method is used to tune process system models, then the models can be adjusted to reflect drift in process data, but the tuning process requires multiple repetitions and is time consuming
Solution Approach 1:
The patent implements feedback mechanisms where the empirical model continuously learns from the difference between first principle model predictions and actual process measurements. This feedback loop automatically adjusts model parameters to reflect drift in process data due to equipment aging or fatigue, eliminating the need for repeated manual trial and error tuning while maintaining high measurement precision.
Data Source
AI summary
Methods and apparatus to compensate first principle-based simulation models are disclosed. An example method to compensate a first-principle based simulation model includes applying one or more first test inputs to a process system to generate first output data, applying one or more second test inputs to a first principle model to generate second output data, generating an error model based on the first and second output data, applying input data to the first principle model to generate simulation model output data, and compensating the model data via the error model to generate compensated model output data.