Injection Molding Failure Diagnosis via Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional failure analysis methods for injection molding machines rely heavily on the knowledge and experience of analysts, making it difficult to accurately diagnose failure causes, and existing systems cannot reliably predict conditions to prevent failures based on molding conditions and alarming histories.
Innovation Solution
A failure cause diagnostic device utilizing machine learning to analyze internal and external state data from injection molding machines, predicting and identifying the cause of failures, calculating correlations, and adjusting state data to prevent future occurrences, regardless of analyst knowledge or experience.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual failure analysis methods are used by users and manufacturers' engineers, then the analysis process can be performed with existing resources, but the reliability of failure cause identification is limited due to restricted knowledge and experience
Solution Approach 1:
The patent introduces a management device as an intermediary between multiple injection molding machines and users/manufacturers. This management device collects molding conditions, alarming histories, and quality information from multiple machines, performs centralized failure cause diagnosis using accumulated data, and provides diagnostic results back to users. This intermediary approach enables reliable failure cause identification without requiring individual users or manufacturers to possess extensive knowledge and experience.
2Extent of automation
If conventional analysis systems are used, then the system structure remains simple, but the system cannot identify failure causes or calculate optimal molding conditions based on acquired data
Solution Approach 1:
The management device performs preliminary actions by continuously collecting and accumulating molding conditions, alarming histories, and quality information from multiple injection molding machines before failures occur. This accumulated data serves as a knowledge base that enables the system to automatically diagnose failure causes and calculate optimal molding conditions when failures do occur, without requiring complex real-time analysis structures.
3Reliability
If data from multiple injection molding machines are processed individually, then each machine can be analyzed separately, but the overall failure diagnosis reliability and predictive capability are reduced
Solution Approach 1:
The patent merges data from multiple injection molding machines by collecting molding conditions, alarming histories, and quality information from various machines into a centralized management device. This combining of data from multiple sources creates a larger, more diverse dataset that improves the reliability of failure cause identification and enables the system to recognize patterns and correlations that would not be apparent from analyzing single-machine data in isolation.
Data Source
AI summary
A failure cause diagnostic device of the present invention receives input of internal and external state data on injection molding machines and diagnoses failure cause of the injection molding machines by means of a machine learning device. An internal parameter of the machine learning device is obtained by performing machine learning using the state data obtained from the injection molding machines subject to failure cause and the state data obtained from the injection molding machines free of failure cause.