Compensation Learning Model for Faster Press Control Adjustment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In press systems, adjusting optimal die height and control variables for processing objects is time-consuming due to the need for trial and error across various material parameters like hardness and temperature, affecting processing efficiency.
Innovation Solution
A learning device that generates a compensation amount for command values in a control system using data-driven methods like VRFT, FRIT, or ERIT, allowing for efficient learning without actual operations, using a learned model based on specific object parameters to optimize control variables without damaging the object.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If trial and error method is used to adjust control variables for different material parameters, then appropriate control values can be obtained, but the adjustment time increases significantly
Solution Approach 1:
The system performs preliminary learning by collecting operation data and generating teacher data in advance, building a learned model that stores the relationship between material parameters and optimal control variables. When processing new materials, the system queries the pre-built model to obtain control variables immediately, eliminating the need for time-consuming trial and error adjustments.
Solution Approach 2:
The system creates a virtual copy of the control object through a learned model that replicates the relationship between material parameters and control variables. This model serves as a digital twin that can predict optimal control settings without requiring physical trial and error on the actual processing system, thus saving time while maintaining accuracy.
2Adaptability or versatility
If multiple combinations of material parameters are tested to find optimal control variables, then comprehensive learning can be achieved, but the number of required objects to be processed increases
Solution Approach 1:
The system uses virtual copying through learned models to represent multiple material parameter combinations without requiring physical objects for each combination. The model generalizes from a limited set of actual learning objects to predict control variables for a wide range of material parameters, achieving comprehensive adaptability without proportionally increasing the number of physical objects needed.
Solution Approach 2:
The system achieves comprehensive learning coverage by varying material parameters within defined ranges during the learning phase and storing the relationships in the learned model. This allows the system to adapt to multiple parameter combinations without physically processing objects for every possible combination, as the model interpolates and extrapolates based on the learned parameter relationships.
3Reliability
If actual press machine operations are performed for learning, then real control data can be collected, but the objects to be processed may be damaged or consumed
Solution Approach 1:
The system creates virtual copies of the press machine operations through a learned model that simulates the control relationships. During the learning phase, the system uses a limited number of actual objects to collect authentic operation data, then stores this data in the learned model. Subsequent learning and optimization are performed on virtual copies, eliminating further object consumption while maintaining data authenticity from the initial real measurements.
Solution Approach 2:
The system performs preliminary data collection using actual press machine operations with a minimal set of objects to establish the learned model. Once the model is built from these initial authentic measurements, all subsequent learning, optimization, and adaptation are performed virtually through the model, preventing further object damage while maintaining reliability of the control data.
Data Source
AI summary
This learning device provides a learned model to an adjuster including the learned model learned to output a predetermined compensation amount to a controller based on parameters of an object to be processed, in a system including the controller outputting a command value obtained by compensating a target value based on a compensation amount; and a control object performing a predetermined process on the object and outputting a control variable as a response to the command value. The learning device includes: a learning part generating candidate compensation amounts based on operation data including a target value, command value and control variable, learning with the generated candidate compensation amounts and the parameters of the object as teacher data, and generating or updating the learned model; and a setting part providing, to the adjuster, the generated or updated learned model.