Injection Molding State Determination Using Process-Specific Learning Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current state determination methods for injection molding machines lack accuracy in diagnosing abnormalities, particularly due to the failure to consider combinations of drive parts and molding processes, leading to reduced productivity and increased costs, as they often require suspending production for measurements and necessitate multiple state determination devices.
Innovation Solution
A state determination device and method that uses machine learning to classify time-series physical quantities from injection molding machines, employing multiple learning models specific to different processes and component consumption periods, allowing for efficient and accurate abnormality estimation and displaying alerts for safety.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple state determination devices are used to accurately diagnose abnormalities of different drive parts, then diagnostic accuracy is improved, but device complexity and cost increase
Solution Approach 1:
The patent implements a single state determination device that can determine abnormalities of multiple different drive parts (injection cylinder, mold clamping cylinder, molded-product ejecting device) by selecting and applying different learning models corresponding to each drive part and its associated molding processes, eliminating the need for multiple separate devices
Solution Approach 2:
The patent segments the diagnostic function by creating multiple specialized learning models, each trained on time-series data from specific drive parts and molding processes. The system divides the overall diagnostic task into process-specific segments (injection process, dwelling process, metering process, mold closing process, mold opening process) and selects the appropriate segment based on current operation context
2Measurement precision
If production is suspended for direct measurement of check valve dimensions, then measurement accuracy is improved, but productivity decreases
Solution Approach 1:
The patent replaces direct mechanical measurement methods (removing screws to measure check valve dimensions) with an information-based diagnostic system that uses machine learning models to analyze time-series operational data, enabling non-intrusive abnormality detection without suspending production
Solution Approach 2:
The patent introduces time-series data from sensors as an intermediary that indirectly reflects the state of drive parts. Instead of directly measuring physical dimensions, the system uses intermediate data (current, speed, pressure) that correlates with component condition, allowing diagnosis during normal operation
3Device complexity
If a single learning model is used for all molding processes, then device complexity is reduced, but determination accuracy decreases due to mixed drive part combinations
Solution Approach 1:
The patent makes the learning model selection dynamic by determining the current molding process and selecting the corresponding learning model based on which drive part is being operated. The system adapts its diagnostic approach in real-time according to the active process (injection, dwelling, metering, mold closing, mold opening) rather than using a static single model
Data Source
AI summary
A state determination device acquires data on an injection molding machine and stores conditions for classifying the acquired data on the injection molding machine and a plurality of learning models. The state determination device further classifies the acquired data based on the stored classification conditions and settles a learning model to which the classified data are applied, among the plurality of stored learning models. Subsequently, the state determination device performs machine learning for the learning model settled as an application destination, based on the classified data.