Welding Data Clustering for Accurate Weld Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current welding and cutting monitoring systems face challenges in accurately identifying and grouping individual welds or cuts without relying on weld or cutting profile identification numbers, leading to incorrect data clustering and defect detection issues due to reused or undefined identification numbers.
Innovation Solution
A system comprising a server computer with an analytics component and data store that receives and analyzes welding or cutting data, including core and non-core data, to identify and group same individual welds or cuts without relying on profile identification numbers, using cluster analysis and additional parameters like pre-idle times and non-welding movements to correctly label and group data for machine learning algorithms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If weld profile identification numbers are used to identify and group welds, then data organization is simplified, but incorrect grouping occurs due to reused or undefined identification numbers
Solution Approach 1:
The patent replaces the mechanical identification system (weld profile identification numbers) with a pattern recognition system based on cluster analysis. Instead of relying on explicit numerical identifiers that can be reused or undefined, the system uses mathematical clustering algorithms to group welds based on their inherent parameter patterns, thereby eliminating identification errors while maintaining data organization.
Solution Approach 2:
The patent changes the identification approach from discrete parameter (identification numbers) to continuous parameter analysis (cluster analysis of welding parameters). By transforming the identification method from categorical labeling to continuous pattern recognition, the system achieves more reliable weld grouping that is resistant to identifier reuse or omission.
2Measurement precision
If cluster analysis is performed without identification numbers, then weld grouping accuracy improves, but data processing complexity increases
Solution Approach 1:
The patent implements a self-service identification system where welds automatically group themselves into clusters based on their parameter patterns without requiring external identification numbers. The cluster analysis algorithm autonomously identifies and groups welds by their inherent characteristics, eliminating the need for manual labeling while improving grouping accuracy.
Solution Approach 2:
The patent introduces cluster analysis as an intermediary computational layer between raw welding data and final weld grouping. This intermediary process transforms the complex task of weld identification into a series of manageable mathematical operations, making the system more robust while maintaining processing feasibility through standardized algorithms.
3Reliability
If detailed welding parameters are collected for analysis, then defect detection capability improves, but data storage requirements increase
Solution Approach 1:
The patent extracts only the essential welding parameters needed for cluster analysis and defect detection, separating them from unnecessary data. By selectively collecting core parameters (current, voltage, speed, arc length) and excluding redundant information, the system maintains high defect detection capability while minimizing data storage requirements.
Solution Approach 2:
The patent segments welding data into distinct parameter categories (core welding parameters, process parameters, quality parameters) that can be analyzed independently. This segmentation allows the system to focus storage and processing resources on the most critical parameters for defect detection, reducing overall data volume while maintaining detection accuracy.
Data Source
AI summary
Embodiments of systems and methods providing pattern recognition and data analysis in welding and cutting are disclosed. In one embodiment, a system includes a server computer and a data store connected to the server computer. The server computer receives welding data, including core welding data and non-core welding data, over a computer network from welding systems used to generate multiple welds to produce multiple instances of a same type of part. The server computer performs an analysis on the welding data to identify and group same individual welds of the multiple welds without relying on weld profile identification numbers as part of the analysis. A group of the same individual welds corresponds to a same weld location on the multiple instances of the same type of part. The data store receives the welding data from the server computer and digitally stores the welding data as identified and grouped.