Additive Weld Bead Shape Modeling for Machine-Agnostic Condition Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The complexity of adjusting welding conditions in additive manufacturing is exacerbated by numerous condition combinations and the challenge of accounting for machine differences between power supplies and robots, leading to difficulties in specifying appropriate welding conditions accurately.
Innovation Solution
An additive manufacturing system that includes a shape sensor, a power supply, and an information processing device using machine learning to derive adjustments for welding conditions based on bead shape data, independent of specific machine configurations, thereby improving accuracy and reducing the need for a machine-specific database.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If a database is created to store welding condition combinations and bead shape data, then the accuracy of welding condition adjustment is improved, but the device complexity and workload increase significantly
Solution Approach 1:
The system performs self-learning by automatically collecting bead shape measurement data and welding condition data, storing them in a database, and generating learning models without requiring manual database creation. This self-service mechanism reduces the workload and complexity of database creation while maintaining high welding condition adjustment accuracy.
Solution Approach 2:
The system performs preliminary learning by collecting and storing welding condition data and bead shape data before actual additive manufacturing operations. This preliminary data collection and model generation enables the system to provide accurate welding condition adjustments from the start, avoiding the need for complex manual database creation during operation.
2Manufacturing precision
If machine-specific databases are created for different power supplies and robots, then the accuracy for specific machines is improved, but the device complexity and number of databases required increase
Solution Approach 1:
The system creates a single general-purpose database that can be used across different power supplies and robots. The learning model is designed to be machine-agnostic, allowing the same database to provide accurate welding condition adjustments for multiple different machines, thereby eliminating the need for separate machine-specific databases.
Solution Approach 2:
The system uses measurement data from actual beads produced on specific machines to create learning models that can be copied and applied to the same machine type. This copying approach allows the system to maintain accuracy across machines of the same type without creating separate databases for each machine, reducing overall system complexity.
3Adaptability or versatility
If manual extraction of welding conditions is performed, then adaptability to specific cases is improved, but the time and labor required increase significantly
Solution Approach 1:
The system implements automatic feedback by measuring actual bead shapes, comparing them with target shapes, and using the learning model to determine appropriate welding condition adjustments. This closed-loop feedback mechanism automatically adapts welding conditions to specific cases without requiring manual extraction, significantly reducing the time and labor required while maintaining high adaptability.
Solution Approach 2:
The system replaces manual mechanical extraction of welding conditions with an automated information processing system. The learning model automatically processes measurement data and welding condition data to determine optimal adjustments, substituting manual operations with automated computational processes that are both faster and more consistent.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enhances the accuracy of welding condition adjustments during additive manufacturing, allowing for precise control of bead shape without requiring separate databases for each machine configuration, and can be applied to various additive manufacturing systems using a general-purpose database.
Implementation Method 1
an arc is generated from the filler wire M to manufacture an additively-manufactured object W
Implementation Method 2
a shape sensor 101 that detects a shape of the additively-manufactured object W
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
Provided is a machine-learning device which performs machine learning for welding conditions when a laminated molding article is molded by welding a filler material and laminating weld beads, wherein the machine-learning device comprises a training process means which performs a training process for generating a trained model in which two pieces of shape data of the welding beads or the difference between the two pieces of shape data is taken as input data, and the difference between welding conditions, which corresponds to the difference between the two pieces of shape data, is taken as output data.