Hybrid Physical Model Identification for Stable Manufacturing Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automatic control systems for manufacturing lines face challenges in improving model precision, particularly for new lines with no data and non-linear models, and stability issues when input values exceed predetermined ranges, limiting their applicability and effectiveness.
Innovation Solution
A physical model identification system that uses actual data to enhance model precision by integrating a calculator, data sampling device, and physical model identification device, which includes data edition, statistical model learning, validity verification, relationship specification, and model coefficient identification units to update correction coefficients and improve prediction and control values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a statistical model is used for system identification, then model precision can be improved through data-driven learning, but the system cannot be applied to new manufacturing lines with no actual data
Solution Approach 1:
The patent introduces a physical model as an intermediary between the statistical model and the manufacturing line. The physical model provides a theoretical framework that can be applied to new manufacturing lines without requiring historical data, while the statistical model refines the physical model's parameters using available actual data. This intermediary physical model enables the system to function in data-scarce scenarios while still benefiting from data-driven precision improvements.
2Measurement precision
If a statistical model is relearned after changes in the controlled object, then model precision is maintained, but time and resources are consumed for data collection and relearning
Solution Approach 1:
The patent establishes a physical model in advance that incorporates fundamental physical relationships and principles. This preliminary physical model provides a robust baseline that remains valid even when manufacturing conditions change. When changes occur in the controlled object, only the statistical parameters need adjustment rather than complete relearning, significantly reducing the time and resources required for model updates.
3Adaptability or versatility
If a physical model is used, then the system can be applied to new manufacturing lines and handle non-linear models, but model precision is difficult to improve without actual data
Solution Approach 1:
The patent merges the physical model and statistical model into a hybrid approach. The physical model provides the structural framework and theoretical foundation that ensures applicability to new manufacturing lines and handles non-linear relationships. The statistical model component uses actual data to refine parameters and improve precision. This combination allows the system to simultaneously achieve broad applicability and high precision by leveraging the strengths of both modeling approaches.
4Adaptability or versatility
If a statistical model receives input values outside predetermined ranges, then calculation precision of predicted values largely degrades
Solution Approach 1:
The patent employs parameter changes to adapt the model to different input ranges. The physical model component provides a theoretical framework that can handle a broader range of inputs based on fundamental physical principles. When input values fall outside the training range of the statistical model, the physical model's parameter relationships guide the predictions, preventing the sharp degradation in precision that would occur with a pure statistical approach. This allows the system to maintain reasonable precision across varied input ranges.
Data Source
AI summary
A physical model identification system according to the present invention first uses actual data sampled from a manufacturing line to learn a statistical model expressing a controlled object in the manufacturing line. Next, this system creates numerical quantity data quantifying an input-output relationship between a registered input variable and a response variable of a learned statistical model. Next, this system identifies a correction coefficient of a physical model such that a relationship, expressed by the numerical quantity data, between the registered input variable and the response variable is maintained. An identified correction coefficient is reflected in the physical model implemented in a calculator controlling the manufacturing line. Accordingly, prediction precision of preset values and control values of the controlled object by a practical model is improved, and a stable operation of the manufacturing line and high quality production become possible.