Process Quality Prediction Using State-Parameter Conversion Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In product production processes, a large number of process condition parameters can lead to overtraining of neural networks, making accurate prediction of quality parameters difficult, especially when using neural networks or functions/table-based models, as it complicates the configuration and results in inaccurate predictions for conditions outside the teacher data.
Innovation Solution
The implementation of an optimization support device with conversion units that convert operating condition parameters into state parameters and then into quality parameters, and vice versa, using trained learning models to facilitate accurate prediction by reducing the complexity of input parameters and preventing overtraining, allowing for accurate prediction of various conditions in product production processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a large number of process condition parameters are input to the prediction model, then the model can capture more comprehensive process information, but the neural network becomes overtrained and prediction accuracy decreases
Solution Approach 1:
The patent segments the process condition parameters into two distinct groups: operating condition parameters (independent variables that can be controlled) and state parameters (dependent variables that reflect process state). This segmentation allows the prediction model to use only the essential operating condition parameters as inputs, rather than all process condition parameters, thereby preventing overtraining while maintaining comprehensive process information capture.
2Ease of manufacture
If a function or table is used instead of a neural network to obtain quality parameters, then the configuration becomes simpler, but a large number of process condition parameters still complicates the function or table configuration
Solution Approach 1:
The patent extracts only the essential operating condition parameters from the complete set of process condition parameters. By taking out only the independent variables that directly influence product quality and removing redundant state parameters, the function or table configuration becomes manageable and simple, while still capturing the critical relationships needed for accurate quality prediction.
3Adaptability or versatility
If all process condition parameters are used as input, then the prediction model can cover more process variations, but it becomes difficult to create an appropriate function or table configuration
Solution Approach 1:
The patent segments parameters into operating condition parameters (independent variables) and state parameters (dependent variables). This segmentation enables the model to cover process variations through the relationship between these two parameter groups rather than by including all parameters as inputs, thus maintaining adaptability while simplifying configuration.
Solution Approach 2:
The patent introduces state parameters as intermediary variables that connect operating condition parameters to quality parameters. Instead of directly mapping all process condition parameters to quality parameters, the state parameters serve as mediators that capture process state information, enabling comprehensive variation coverage with a manageable configuration structure.
Data Source
AI summary
The optimization support device includes a first conversion unit that converts an operating condition parameter indicating an operating condition of a process for producing a product into a state parameter indicating a state of the process, and a second conversion unit that converts the state parameter into a quality parameter indicating a quality of the product.