Injection Molding Dataset Creation for AI Parameter Tuning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing methods for creating datasets for machine learning models that adjust molding condition parameters in injection molding machines are time-consuming and labor-intensive, requiring manual intervention to induce and correct molding defects.
Innovation Solution
A method that automatically creates datasets by systematically changing molding condition parameters to degrade and then improve product quality, associating physical quantity data with these changes to generate a dataset for machine learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a large volume of unstructured data is collected from multiple sources for AI training, then the applicability and accuracy of the trained AI model are improved, but the cost and complexity of data collection, management, and processing increase significantly
Solution Approach 1:
The patent introduces a structured data collection system with predefined collection rules, data formats, and validation mechanisms that act as intermediaries between multiple data sources and the training dataset. This intermediary layer standardizes data from diverse sources (social media, databases, APIs, files) into a unified structured format, reducing the complexity of managing unstructured multi-source data while maintaining high data quality and model accuracy
Solution Approach 2:
The system dynamically adjusts data collection parameters such as collection frequency, data volume thresholds, and source priorities based on predefined rules and training requirements. By changing these parameters adaptively, the system optimizes the balance between data diversity (for model accuracy) and collection complexity, ensuring efficient data acquisition without overwhelming system resources
2Reliability
If data collection frequency and volume are increased to improve training data comprehensiveness, then the accuracy of the trained model is improved, but the time and resources required for data processing increase
Solution Approach 1:
The patent implements preliminary data validation, filtering, and structuring during the data collection phase itself. Collection rules are predefined to validate data format, completeness, and relevance before data is fully acquired. This preliminary action prevents the accumulation of invalid or redundant data, reducing the time needed for later processing while ensuring comprehensive and high-quality training data
Solution Approach 2:
The system establishes continuous data collection and processing pipelines that operate systematically according to predefined schedules and triggers. By maintaining continuous structured data flow from multiple sources with automated processing, the system achieves comprehensive data coverage over time without requiring intensive batch processing, thus reducing total processing time while improving data comprehensiveness
3Reliability
If strict data collection rules and validation are implemented to ensure data quality, then the reliability of the training dataset is improved, but the complexity of the data collection system increases
Solution Approach 1:
The patent segments the data collection and validation system into modular components: data source modules, collection rule engines, validation modules, and processing modules. Each module handles specific tasks independently with well-defined interfaces. This segmentation allows strict validation rules to be implemented systematically without creating monolithic complexity, as each segment can be developed, tested, and maintained independently while ensuring overall data quality
Data Source
Figure 1
Figure 2
Figure 3
AI summary
According to the present invention: physical quantity data indicating the state of a molded article produced by modifying a first molding condition parameter, among first and second molding condition parameters set in a molding machine, so as to degrade the quality of the molded article, or indicating the state of the molding machine, is acquired; physical quantity data indicating the state of a molded article produced by modifying the second molding condition parameter set in the molding machine, or indicating the state of the molding machine, is acquired; the second molding condition parameter before modification, the physical quantity data obtained at that time, the second molding condition parameter after modification, and the physical quantity data obtained when the second molding condition parameter after modification was set are associated with one another and stored; and a data set for machine learning is created by repeatedly modifying the first and second molding condition parameters and acquiring the physical quantity data.