Clustering-Based Defect Prediction Adaptation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning systems for predicting manufacturing defects struggle to adapt to changing manufacturing conditions, leading to inaccurate defect predictions due to shifting relationships between trace data and defects.
Innovation Solution
A system and method utilizing a cluster classifier to cluster manufacturing data into multiple modalities, reassigning input vectors based on cluster labels, and retraining machine learning models to improve prediction accuracy, incorporating a tree-structured multimodal regressor design for addressing varying manufacturing conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a single machine learning model is used for defect prediction, then the system is simple to implement, but it cannot adapt to changing manufacturing conditions leading to inaccurate predictions
Solution Approach 1:
The patent segments the manufacturing data into multiple clusters based on different manufacturing conditions, and trains a separate machine learning model for each cluster. This allows the system to adapt to changing conditions by selecting the appropriate model for the current data cluster, resolving the contradiction between adaptability and complexity through structured division of the prediction task.
Solution Approach 2:
The patent changes the parameter of model selection based on the cluster label assigned to input data. By dynamically selecting different models based on the identified manufacturing condition cluster, the system achieves adaptability to changing conditions while maintaining manageable complexity through parameter-based model switching.
2Measurement precision
If cluster labels are assigned based on initial training, then the training process is straightforward, but prediction accuracy decreases when manufacturing conditions change
Solution Approach 1:
The patent implements periodic retraining of the cluster classifier at specified intervals or when triggered by significant condition changes. This periodic action refreshes the cluster assignments and model selections to reflect current manufacturing conditions, maintaining high prediction accuracy while controlling retraining frequency to manage time loss.
Solution Approach 2:
The system incorporates feedback mechanisms that monitor prediction performance and manufacturing condition changes, triggering retraining when accuracy degradation is detected. This feedback-driven approach ensures high prediction accuracy by ret training only when necessary, balancing accuracy maintenance with time efficiency.
Data Source
AI summary
A method of training a system for making predictions relating to products manufactured via a manufacturing process includes receiving a plurality of input vectors and a plurality of defect values corresponding to the plurality of input vectors, identifying a plurality of first cluster labels corresponding to the plurality of input vectors based on the defect values, training a cluster classifier based on the input vectors and the corresponding first cluster labels, reassigning the input vectors to a plurality of second cluster labels based on outputs of the cluster classifier, retraining the cluster classifier based on the input vectors and the second cluster labels, and training a plurality of machine learning models corresponding to the second cluster labels.