Neural Network Control Apparatus for Injection Molding Quality Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing neural network-based control systems for injection molding machines face challenges in maintaining prediction accuracy during mass-production molding, as monitor values deviate over time, leading to unsatisfactory product quality and increased costs due to the need for frequent revisions of the quality prediction function.
Innovation Solution
A control apparatus with a neural network that determines a quality prediction function through repeated estimation of weight factors and thresholds during test molding, and includes an upper/lower control limit determination unit and a function revision need determining unit to adjust the quality prediction function and generate alarm signals when monitor values deviate, ensuring continuous production and reducing measurement costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a neural network quality prediction function is established during test molding and used for mass-production molding, then productivity increases by eliminating individual measurement, but measurement precision deteriorates as monitor values deviate over time
Solution Approach 1:
The system continuously monitors actual quality results during mass-production molding and feeds this information back to the neural network. When deviations exceed predetermined thresholds, the system automatically relearns by incorporating new data, thereby maintaining prediction accuracy without interrupting production flow.
Solution Approach 2:
The quality prediction function transitions from a static model established during test molding to a dynamic system that can be continuously updated. The neural network's weighting factors and thresholds are adjusted through repeated learning cycles based on actual production data, allowing the system to adapt to changing conditions while maintaining high productivity.
2Measurement precision
If the quality prediction function is revised frequently to maintain accuracy, then measurement precision is maintained, but productivity decreases due to increased revision time
Solution Approach 1:
Instead of continuously revising the quality prediction function, the system performs partial updates only when necessary. Revision triggers are based on predetermined conditions such as deviation thresholds or elapsed time, allowing the system to maintain adequate accuracy without the overhead of frequent complete relearnings.
Solution Approach 2:
The system performs preliminary quality predictions using the established neural network during mass-production molding. Only when predictions fall outside acceptable ranges does the system initiate a revision cycle, allowing most production to proceed with the original model while maintaining quality control.
3Reliability
If monitor values are continuously monitored during mass-production molding, then reliability of quality control is improved, but device complexity increases
Solution Approach 1:
The control apparatus integrates multiple functions into a unified system. The same neural network used for quality prediction also serves as the basis for monitoring and revision triggering. The control unit performs both prediction and deviation detection functions, reducing the need for separate monitoring systems and minimizing overall complexity.
Solution Approach 2:
The system monitors its own performance and automatically determines when revision is needed. The control unit compares actual monitor values against the neural network predictions and autonomously triggers relearning when deviations exceed thresholds, eliminating the need for external quality control personnel or additional monitoring equipment.
Data Source
AI summary
Test molding and mass-production molding are performed by an injection molding machine that includes a control apparatus in which neural networks are used. A quality prediction function determined based on the test molding is revised as necessary during mass-production molding.