Asymmetric Process Parameter Control via Adaptive Biasing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Manufacturing processes face challenges in controlling asymmetrical quality parameters, where certain parameters are not directly measurable, leading to prediction errors due to imperfections in modeling algorithms and unmeasured disturbances, and off-process measurements have inconsistencies, necessitating a technique to adaptively correct these errors more quickly in one direction than the other.
Innovation Solution
An adaptive biasing technique is employed, where a filtering factor is selected based on the direction of the prediction error, with a higher filtering factor applied when the error is in the less tolerable direction, and a lower filtering factor when it is in the more tolerable direction, to compute a biasing factor that adjusts the predicted values, ensuring quicker correction in the weaker direction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If off-process measurements are used to correct prediction errors, then measurement accuracy is improved, but measurement consistency deteriorates due to human/machine errors and sample variability
Solution Approach 1:
The system implements feedback by comparing off-process measurements with inferential model predictions, calculating prediction errors, and using these errors to update future predictions through a biasing factor. This closed-loop feedback mechanism continuously improves prediction accuracy while accounting for measurement inconsistencies through adaptive filtering.
Solution Approach 2:
The system dynamically adjusts the biasing factor based on the direction and magnitude of prediction errors, applying asymmetric filtering that changes parameters adaptively. When errors indicate a trend toward unacceptable quality, the system increases the biasing factor more aggressively, thereby changing the response parameters to maintain product quality despite measurement variability.
2Manufacturing precision
If asymmetric control with higher biasing factor is applied in less tolerable direction, then quality specification compliance is improved, but production cost increases due to aggressive control
Solution Approach 1:
The system applies asymmetric control by using different biasing factors for different directions of prediction error. A higher biasing factor is applied when predictions indicate potential non-compliance with minimum specifications, while a lower factor is used when predictions are comfortably within specifications. This asymmetric approach ensures quality compliance while minimizing unnecessary production adjustments.
Solution Approach 2:
The system applies partial action by adjusting the biasing factor proportionally to the predicted deviation from specifications. Rather than always applying maximum control, the system applies just enough biasing to bring predictions back toward target values, avoiding excessive control actions that would increase production costs unnecessarily.
3Speed
If asymmetric filtering is applied to prediction errors, then correction speed is improved in critical direction, but control stability deteriorates due to differential response
Solution Approach 1:
The system implements dynamic control by adjusting the biasing factor based on real-time prediction errors and their direction. The filtering is not static but adapts continuously to process conditions, increasing responsiveness when quality issues are predicted while maintaining stability when process conditions are favorable. This dynamic approach balances correction speed with control stability.
Data Source
AI summary
A technique is disclosed for asymmetrically controlling a process parameter based upon the direction of a prediction error between a predicted value determined using an inferential model and a laboratory measurement of the parameter. The present technique provides for the adaptive biasing of the predicted value based upon the direction of the prediction error. In one embodiment, a biasing factor may be determined by filtering the prediction error, such that the prediction error is emphasized more heavily in the biasing factor if the prediction error is in a less tolerable direction and emphasized less heavily if the prediction error is in the opposite direction. The biasing factor may further be determined as a function of a previous biasing factor computed during the process. Asymmetric control of the process parameter may be performed by controlling the parameter using model predictive control techniques based on the biased predicted values of the parameter.