Manufacturing Condition Ranking for Imbalanced Change Histories
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing manufacturing condition setting technologies face challenges in maintaining data diversity and precision during model updates and analyses, particularly when condition changes result in failures due to external factors, leading to imbalanced training data and reduced effectiveness in future changes.
Innovation Solution
A manufacturing condition setting automating apparatus that includes a facility data acquiring unit, quality judging unit, manufacturing condition candidate creating unit, and imbalance-preventing unit, which collects and analyzes facility data, computes process capability, searches past condition change history for similar cases, and adjusts candidate rankings to prioritize successful changes while accounting for external factors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If only the most implemented condition change is executed under the same conditions, then the decision-making process is simplified, but the diversity of training data deteriorates and model update precision decreases
Solution Approach 1:
The system performs preliminary classification of condition change history data into successful and failed cases before model updates. This pre-processing step ensures that diverse training data is prepared in advance, allowing the model to learn from both successful and failed experiences without complicating the actual decision-making process during operation
Solution Approach 2:
The condition change history data is segmented into distinct categories (successful cases and failed cases) based on outcome analysis. This segmentation allows the system to maintain data diversity by preserving multiple types of experiences while still enabling simplified decision-making through structured presentation to the model
2Manufacturing precision
If failed condition change cases are excluded from training data, then model precision is improved, but future adaptability deteriorates due to loss of valuable failure information
Solution Approach 1:
The system converts failed condition change cases, which initially appear harmful or useless, into valuable training data by analyzing failure causes and outcomes. Failed cases are reclassified and retained in the training dataset, transforming what would be discarded information into beneficial learning material that enhances both model precision and future adaptability
Solution Approach 2:
The system implements feedback mechanisms that analyze the outcomes of condition changes (both successful and failed) and use this information to refine future decisions. By incorporating feedback from failed cases about what not to do, the model learns to avoid repeating mistakes while maintaining precision through structured outcome analysis
3Manufacturing precision
If condition change history data with failures is retained, then data diversity is maintained, but data quality for training deteriorates due to inclusion of unsuccessful cases
Solution Approach 1:
The system applies different quality standards to different portions of the training data based on their value and reliability. Successful cases and failed cases are processed and weighted differently in the training process, allowing the system to maintain data diversity while ensuring that high-quality, reliable data has greater influence on model learning than lower-quality data
Data Source
AI summary
A manufacturing condition setting automating apparatus includes: a quality judging unit that computes a present process quality from facility data at predetermined time intervals, and judges whether or not it is in a quality tolerance range; a manufacturing condition candidate creating unit that computes a feature quantity, searches a database for condition change cases having similar feature quantities, tabulates condition change cases basis on whether the condition change cases are successes or failures, and outputs manufacturing condition candidates in descending order of rates of successes; an imbalance-preventing manufacturing condition candidate creating unit that changes scores that decide ranks of manufacturing condition candidates, and creates a ranking of manufacturing condition candidates; and a manufacturing condition output unit that outputs a set value of a condition change of a top manufacturing condition candidate to the manufacturing facility, and registers a new condition change in the condition change history.