Reinforcement Learning Injection Molding Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing injection molding machine systems face challenges in adjusting molding conditions quickly and efficiently, particularly in the absence of an operator, due to excessive computer resource consumption and time required for learning, especially when dealing with variations in external environments and mechanical wear.
Innovation Solution
An injection molding machine system that incorporates a machine learner performing reinforcement learning using physical data from the machine and defect types, along with a defect judging device and classifier for supervised learning, to adjust molding conditions based on defect states, reducing the need for extensive computer resources and learning time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If reinforcement learning is used to adjust molding conditions, then automation and productivity are improved, but computer resource consumption and learning time increase excessively
Solution Approach 1:
The patent segments the molding condition adjustment process into two distinct modules: a defect judging device that detects molding defects using vision recognition, and a machine learner that adjusts molding conditions based on defect information. This segmentation allows the system to avoid excessive computer resource consumption by dividing the complex reinforcement learning task into smaller, more efficient sub-tasks, where the defect judging device handles detection and the machine learner handles adjustment decisions.
Solution Approach 2:
The patent introduces defect information as an intermediary between the defect judging device and the machine learner. Instead of directly processing complex visual data for molding condition adjustment, the system uses defect information (defect type, location, and severity) as a simplified intermediary representation. This intermediary reduces the computational burden on the machine learner while still enabling effective molding condition adjustments.
2Manufacturing precision
If comprehensive physical data is used for reinforcement learning, then molding condition accuracy is improved, but learning time and computer resources increase
Solution Approach 1:
The patent extracts only the essential information needed for molding condition adjustment: defect type, defect location, and defect severity. Instead of using comprehensive physical data from the injection molding machine, the system extracts specifically the defect-related information that is most relevant for adjusting molding conditions. This extraction principle reduces learning time and computer resource requirements while maintaining sufficient accuracy for effective molding condition control.
Solution Approach 2:
The patent applies local quality by focusing on specific defect characteristics (type, location, severity) rather than analyzing all physical parameters of the molding process. The defect judging device specifically identifies and reports only the relevant defect properties, allowing the machine learner to concentrate computational resources on adjusting molding conditions based on these localized defect characteristics rather than processing comprehensive global data.
3Measurement precision
If operator monitoring is used to detect molding defects, then defect detection accuracy is improved, but productivity decreases due to continuous operator presence requirement
Solution Approach 1:
The patent implements self-service by equipping the injection molding machine with an autonomous defect detection and adjustment system. The defect judging device automatically detects molding defects, and the machine learner automatically adjusts molding conditions based on detected defects, eliminating the need for continuous operator monitoring. This self-service capability maintains high defect detection accuracy while enabling continuous operation and significantly improving productivity.
Solution Approach 2:
The patent establishes a feedback loop where the defect judging device continuously monitors molded products, reports defects to the machine learner, and the machine learner adjusts molding conditions in response. This automated feedback mechanism replaces operator monitoring, maintaining high defect detection accuracy while enabling continuous operation without human intervention, thus improving productivity.
Data Source
AI summary
Provided is an injection molding machine system (1) that performs control of molding conditions in an injection molding machine (2) by an agent (6) including a machine learning device which performs reinforcement learning. In the present learning, physical data obtained from the injection molding machine (2) and a defect type indicating the type of a molding defect in a molded article are used as states, molding conditions are used as actions, and a defect state indicating the defect level of the molding defect is used as a reward.