Mean Value Model Identification Using Segmented Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing engine modeling approaches face challenges in accurately identifying mean value models for real-time optimal control, particularly in emissions control, due to issues with model validity, stability, and the need for high-dimensional optimization, which can lead to inefficient parameter optimization and divergence.
Innovation Solution
A system and method for identifying mean value models using logical matrices and quadratic programming to constrain model signals within valid ranges, ensuring model validity and optimizing control parameters through iterative processes that separate inner and outer optimizations, thereby addressing the challenges of model stability and dimensionality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional high-dimensional optimization approaches are used for mean value model identification, then model accuracy can be improved, but computational complexity and optimization time increase significantly
Solution Approach 1:
The optimization problem is segmented into two distinct levels: inner optimization for state variables at each operating point and outer optimization for model parameters. This segmentation reduces the complexity of the overall optimization by handling high-dimensional state variable optimization separately from the lower-dimensional parameter optimization, enabling accurate model identification without excessive computational burden.
Solution Approach 2:
The approach transforms the high-dimensional optimization problem into a structured two-level hierarchy where the inner optimization handles the high-dimensional state space at each operating point, while the outer optimization handles the lower-dimensional parameter space. This dimensional restructuring enables manageable optimization of mean value model parameters while maintaining accuracy.
2Adaptability or versatility
If model constraints are not enforced during identification, then optimization flexibility is maintained, but model validity is compromised
Solution Approach 1:
Physical and mathematical constraints on model states and parameters are incorporated into the optimization formulation to prevent violation of model validity conditions. By embedding these constraints directly in the optimization problem, the approach proactively prevents invalid model states from being selected, ensuring reliable model identification while maintaining optimization flexibility within valid ranges.
3Manufacturing precision
If iterative optimization is performed without convergence criteria, then parameter refinement continues, but computational time increases indefinitely
Solution Approach 1:
The optimization algorithm incorporates convergence criteria that monitor the reduction of the objective function and changes in parameters across iterations. When convergence thresholds are met, the algorithm automatically terminates, providing feedback that balances continued parameter refinement with computational time limits. This ensures sufficient parameter accuracy is achieved without unnecessary computational expenditure.
Data Source
AI summary
A system or approach for identifying mean value models with a set of equations and appropriate constraints which define the model validity. A model may be used to design an algorithm for an engine system, collecting sensed data, optimizing control parameters based on the models and data, and providing control of the engine system. These processed may be reiterated for updating control of the engine system.