Adaptive Feedforward Control Model for Disturbance Compensation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing feedforward control devices face challenges in updating their configurations quickly in response to disturbances and changes in control process characteristics, leading to difficulties in maintaining accurate control performance.
Innovation Solution
A control device that determines a feedforward compensation value using a prediction model and performs machine learning on supervised data to adapt to disturbances and control object characteristics in real time, improving the accuracy of feedforward control by selecting and refining data based on error values and disturbance levels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If the configuration of the feedforward compensator is predefined and designed as an inverse model of the control process, then the control structure is simple and easy to implement, but it is difficult to update the configuration quickly in response to fluctuations of disturbance and changes in the characteristics of the control process
Solution Approach 1:
The patent transforms the static, predefined feedforward compensator configuration into a dynamic system that can adapt in real-time. The learning unit continuously updates the prediction model parameters based on supervised data from the control process, enabling the system to respond to changing disturbance characteristics and control process variations without requiring complete redesign.
Solution Approach 2:
The system performs self-learning and self-updating through the learning unit that automatically acquires supervised data from the control process and updates the prediction model parameters. This eliminates the need for manual reconfiguration or external intervention when disturbance characteristics change, allowing the system to adapt autonomously.
2Adaptability or versatility
If the feedforward compensator configuration is updated frequently to adapt to changes, then the adaptability to disturbance improves, but the complexity of the control system increases
Solution Approach 1:
The patent implements a feedback mechanism where the learning unit continuously monitors the control process output and uses the actual control amounts and disturbance data as supervised data to update the prediction model. This closed-loop learning approach ensures the system adapts to changes while maintaining a relatively simple structure by leveraging existing process data.
Solution Approach 2:
The patent combines the feedforward control function with the learning/update function into an integrated system. The prediction model serves dual purposes: generating feedforward compensation values and being updated through machine learning. This merging reduces overall system complexity by eliminating separate adaptation mechanisms.
3Measurement precision
If machine learning is performed continuously on the prediction model using all available data, then the accuracy of feedforward control improves, but the computational load and processing time increase
Solution Approach 1:
The patent applies partial action by selectively using supervised data that meets specific criteria (error between target value and control amount, and disturbance characteristics) for updating the prediction model. Instead of processing all available data continuously, the system identifies and processes only the most relevant data points, reducing computational load while maintaining accuracy.
Solution Approach 2:
The patent changes the parameter of data selection criteria dynamically. The learning unit adjusts which data combinations are used for supervised learning based on the current control state, error magnitude, and disturbance characteristics. This selective parameter approach optimizes the balance between learning accuracy and processing efficiency.
Data Source
AI summary
A control device includes a feedback controller that determines a feedback operation amount based on an error between a target value and a control amount, a feedforward compensation device that determines a feedforward compensation value from a disturbance using a prediction model, and a learning device that performs machine learning on the prediction model using supervised data, obtains a first combination which includes the target value, the feedback operation amount corresponding to the target value, the disturbance, and the control amount corresponding to both the disturbance and the feedback operation amount, and adds, to the supervised data, a second combination which includes the disturbance and the feedback operation amount when the absolute value of the error is smaller than a first reference value.