Diesel Engine Reference Governor for Constraint Handling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing diesel engine control devices face challenges in achieving high versatility for prediction models and reducing calculation load when imposing constraints on state quantities, leading to increased calculation requirements for target value correction.
Innovation Solution
A diesel engine control device incorporating a reference governor that uses an nth-order function model to predict future state quantities and correct target values by minimizing an evaluation function, which includes terms for distance and constraint violation, allowing for non-iterative calculation of target values that satisfy upper or lower limit constraints.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If an nth-order function model is used to predict future state quantities, then the versatility of the prediction model is improved, but the calculation complexity increases
Solution Approach 1:
The patent changes the order parameter of the function model from fixed (second-order) to variable (nth-order), allowing the system to adapt the model complexity to match the actual dynamic characteristics of the plant. This enables versatile modeling while managing calculation complexity through selective use of higher-order terms only when necessary.
Solution Approach 2:
The patent introduces dynamic selection of model order based on the specific application requirements and plant characteristics. The nth-order function model can dynamically adjust its complexity level, using higher orders for complex systems requiring high versatility and lower orders for simpler systems where calculation efficiency is prioritized.
2Manufacturing precision
If iterative calculation is used to correct target values with constraint imposition, then the accuracy of constraint satisfaction is improved, but the calculation load increases
Solution Approach 1:
The patent extracts the constraint satisfaction requirement from the iterative correction process and incorporates it directly into the evaluation function. By formulating the target value correction as an optimization problem where the evaluation function inherently penalizes constraint violations, the system achieves accurate constraint satisfaction without requiring separate iterative calculation steps.
Solution Approach 2:
The patent performs preliminary formulation of the evaluation function to include constraint satisfaction criteria before the actual target value correction process. The evaluation function is designed in advance to guide the optimization toward feasible solutions that satisfy constraints, preventing the need for iterative adjustments after the fact.
3Productivity
If a fixed second-order model is used for prediction, then the calculation load is reduced, but the adaptability to different plant dynamics deteriorates
Solution Approach 1:
The patent changes the fixed model order parameter from second-order to a variable nth-order parameter. This allows the prediction model to adapt its complexity to match the actual dynamic characteristics of different plants or operating conditions, improving versatility without permanently increasing the calculation load for all applications.
Data Source
Figure 1
Figure 2~3
Figure 4~5
AI summary
A plant control device includes a feedback controller (54) that determines a control input of a plant (56), and a reference governor (52) that corrects an initial target value and outputs the corrected initial target value to the feedback controller (54). The reference governor (52) determines a target value satisfying an upper limit constraint to be a target value candidate that minimizes a value of an evaluation function. The reference governor (52) determines the target value satisfying the upper limit constraint to be a value of the upper limit constraint when a value acquired by substituting, with the value of the upper limit constraint, a variable of a differential function acquired by differentiating the evaluation function with respect to the target value candidate is greater than or equal to zero.