Robot Control With Dimensionally Aware Weld Quality Rules
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for confirming weld quality in ultrasonic welding of sheet metals are non-destructive and sensitive to environmental variables, requiring tedious feature identification and manual re-building of black-box classifiers, which lack physical understanding and are not adaptable to changes in environmental conditions.
Innovation Solution
A dimensionally aware rule mining approach using genetic programming and automated rule discovery to generate meaningful rules from welding machine sensor data, enabling classification of good and bad welds and informing controller adjustments for improved quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If black-box classifiers are used to confirm weld quality, then classification accuracy can be achieved, but the physical understanding of the welding process is lost and the system becomes difficult to adapt to environmental changes
Solution Approach 1:
The patent introduces dimensionally aware rule extraction as an intermediary between raw sensor data and classification decisions. This intermediary layer extracts interpretable rules with explicit dimensional relationships (e.g., force × distance = energy), providing both classification accuracy and physical understanding. The rules serve as a mediator that maintains adaptability while achieving precise weld quality assessment.
Solution Approach 2:
The system dynamically adjusts classification parameters based on environmental conditions by extracting rules that explicitly model relationships between welding parameters and quality outcomes. When environmental variables change, the rule extraction process adapts by re-evaluating dimensional relationships in the new conditions, allowing the system to maintain accuracy without complete retraining.
2Measurement precision
If manual feature identification and classifier rebuilding is performed for every environmental change, then accurate classification can be maintained, but the process becomes extremely time-consuming and complex
Solution Approach 1:
The system performs preliminary dimensionally aware rule extraction from training data to establish a rule template that captures the fundamental dimensional relationships in welding processes. This preliminary action creates a reusable framework that can be quickly adapted to new environmental conditions without requiring complete manual feature identification and classifier rebuilding from scratch.
Solution Approach 2:
The extracted rules serve multiple functions: they provide classification decisions, explain physical relationships, and serve as an adaptable template for different environmental conditions. This universal rule structure eliminates the need for separate manual feature identification processes for each condition, significantly reducing time and complexity.
3Loss of information
If dimensionally aware rule extraction is used instead of black-box classifiers, then physical understanding and interpretability are improved, but the system complexity increases due to rule extraction and validation requirements
Solution Approach 1:
The system segments the rule extraction process into distinct dimensional components (force, distance, time, energy) that can be independently analyzed and validated. This segmentation makes the complex rule extraction process more manageable by breaking it down into smaller, verifiable dimensional relationships rather than treating it as a monolithic complex system.
4Adaptability or versatility
If automated rule discovery methods are employed, then the system can adapt to environmental changes, but the rules may lack dimensional correctness and physical meaning
Solution Approach 1:
The system incorporates dimensional validation feedback into the automated rule discovery process. Extracted rules are evaluated against dimensional consistency criteria (ensuring that relationships respect physical dimensions like energy = force × distance). This feedback mechanism guides the rule extraction algorithm to prefer rules that are both adaptive and dimensionally correct, resolving the conflict between automation and physical meaning.
Data Source
AI summary
A system includes a memory storing a dimensionally aware model generated based on a training set and guided by feature dimensions and instructions for execution a processor. The instructions include, in response to receiving a set of data from a user device, identifying a set of features from the set of data and applying the dimensionally aware model to the set of features by implementing a boundary representation. The instructions include classifying the set of features as acceptable in response to the implementation of the boundary representation indicating the set of features are outside the boundary representation, classifying the set of features as unacceptable in response to the implementation of the boundary representation indicating the set of features are inside the boundary representation, generating, for display on the user device, an alert based on the classification and controlling a user device to obtain product features within the boundary representation