Prediction Method for Control System Parameter Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional fictitious reference iterative tuning (FRIT) techniques require actual operation of a control system to understand controller behavior, leading to a high workload for designers due to residual errors between control output and target response.
Innovation Solution
A prediction method that calculates parameter values ρ* and θ* to minimize an evaluation function, allowing for the prediction of control input and output based on transfer functions, enabling accurate analysis of controller behavior without physical experimentation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If the calculated parameter value from evaluation function is simply set as the controller parameter, then the calculation process is simple, but an error remains between control output and target response
Solution Approach 1:
The patent applies preliminary action by predicting controller behavior before actual implementation. The prediction unit calculates predicted control inputs and outputs using the optimized parameters ρ* and θ* obtained from the evaluation function, allowing designers to verify accuracy before deploying the controller in real systems.
2Loss of information
If actual operation of control system is performed to understand controller behavior, then accurate behavior knowledge is obtained, but workload for designer increases
Solution Approach 1:
The patent uses copying by creating a virtual model of the controller's behavior through prediction. Instead of physically operating the control system, the prediction unit generates copies of control inputs and outputs based on the transfer functions and optimized parameters, providing sufficient behavioral knowledge without physical experimentation.
3Manufacturing precision
If multiple iterations of actual system operation are performed to reduce error, then control accuracy improves, but time and resources are consumed
Solution Approach 1:
The patent performs preliminary optimization by calculating the parameter value ρ* that minimizes the evaluation function before actual system operation. This pre-optimization step, combined with behavior prediction, allows designers to achieve high accuracy without multiple iterative trials of the actual control system.
Solution Approach 2:
The patent implements feedback through the evaluation function J(ρ, θ, um, ym) that quantifies the error between target response and control output. This feedback mechanism guides the optimization of parameters ρ and θ, enabling systematic improvement of control accuracy without random trial-and-error experimentation.
Data Source
AI summary
A prediction method includes: acquiring time-series data of sets of a control input um and a control output ym as a sample of a control input u and a control output y; calculating, based on the time-series data, a value ρ* that minimizes a value of an evaluation function J(ρ,θ,um,ym) in a state where a parameter θ is set to a fixed value θ0, calculating a value θ* that minimizes the evaluation function J(ρ,θ,um,ym) in a state where a parameter ρ is set to the value ρ*; and calculating a prediction value up of the control input u and a prediction value yp of the control output y corresponding to a desired value r, based on a transfer function C(ρ*) in which the parameter ρ is set to the value ρ* and on a target response transfer function Td(θ*) in which the parameter θ is set to the value θ*.