Machine Learning Welding Procedure Specification for Faster WPS Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current welding procedures require significant time and resources, and the reliability of welding operations can be compromised if the welding zone is not satisfactory, leading to potential product failures.
Innovation Solution
A method and apparatus using a machine learning algorithm to automatically generate a welding procedure specification (WPS) by collecting welding-related information and using a pre-trained machine learning model to generate a suitable WPS, improving reliability and ease of use through a web service.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional welding procedure specification methods are used, then welding operations can be performed with established procedures, but significant time and resources are required for manual creation and verification of WPS
Solution Approach 1:
The patent replaces the manual mechanical process of WPS creation with an automated machine learning system. The ML model automatically generates WPS by processing welding-related information (material type, thickness, joint configuration) and predicting appropriate welding parameters, eliminating the need for manual expert intervention in routine WPS generation tasks.
Solution Approach 2:
The system enables self-service WPS generation where the machine learning model autonomously creates welding procedure specifications without requiring manual input from welding experts. The model learns from historical WPS data and automatically generates new WPS based on input parameters, making the system self-sufficient for routine WPS creation.
2Reliability
If manual welding procedure specification creation is used, then expert knowledge can be applied, but the reliability of WPS may be compromised due to human error and inconsistency
Solution Approach 1:
The patent replaces the human expert system with an automated machine learning system to eliminate human errors and inconsistencies. The ML model provides consistent, reproducible WPS generation based on learned patterns from historical data, improving reliability by removing variability associated with manual expert judgment.
Solution Approach 2:
The system incorporates feedback mechanisms where the machine learning model is trained on historical WPS data and welding outcomes. The model continuously learns from past performance data, refining its predictions to improve WPS reliability over time while maintaining consistent quality standards.
3Productivity
If automated machine learning WPS generation is implemented, then productivity and consistency are improved, but the system complexity and training data requirements increase
Solution Approach 1:
The patent applies preliminary action by pre-training the machine learning model on extensive historical WPS data before deployment. This offline training phase prepares the model to rapidly generate WPS during operation, separating the complex training process from the simple inference process used in actual WPS generation.
Solution Approach 2:
The machine learning model is designed to handle multiple welding scenarios and material types through a unified framework. By learning general patterns from diverse training data, the model can generate WPS for various welding applications without requiring separate specialized systems for each welding type.
Data Source
AI summary
Provided are a method and apparatus for automatically generating a welding procedure specification (WPS) by using a machine learning algorithm. The method, performed by a processor of a WPS generating apparatus, of automatically generating a WPS by using a machine learning algorithm, includes collecting a WPS transmission request signal together with welding-related information including a welding material for a welding target and a thickness of the welding material, generating a WPS corresponding to the welding-related information by using a machine learning model pre-trained to generate the WPS by using the welding-related information, and transmitting a WPS response signal together with the generated WPS, in response to the WPS transmission request signal, wherein the machine learning model is a model trained through supervised learning using training data in which the welding-related information is an input and the WPS corresponding to the welding-related information is a label.