Pulse Arc Welding State Determination via Normal Pattern Deviation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining the welding state in pulse arc welding require extensive data on abnormal patterns under various conditions, making it difficult to achieve accurate and efficient evaluation.
Innovation Solution
A welding state determination device and method that acquires and preprocesses pulse waveforms or probability densities of pulse currents/voltages, shaping them to match normal patterns, allowing for state determination based on differences with past data, thereby simplifying the identification of normal or abnormal welding states without needing extensive abnormal data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If neural network learning method is used to determine welding state, then welding state determination accuracy can be improved, but extensive abnormal pattern data preparation is required which increases complexity and time consumption
Solution Approach 1:
Instead of learning to identify abnormal patterns from extensive abnormal data, the invention inverts the approach by learning to identify normal patterns from normal data only. The neural network is trained exclusively on normal welding state data, and any deviation from the learned normal pattern is identified as abnormal. This eliminates the need to collect and prepare extensive abnormal pattern data while maintaining determination accuracy.
Solution Approach 2:
The invention extracts and focuses only on the essential characteristics of normal welding patterns, separating them from the complexity of abnormal pattern classification. By extracting features that define normal welding states and training the neural network solely on these, the system simplifies the data preparation process while preserving the ability to accurately determine welding states.
2Adaptability or versatility
If extensive abnormal pattern data is collected through experiments, then welding state determination coverage can be improved, but experimental time and resource consumption increase significantly
Solution Approach 1:
The invention reverses the traditional approach of collecting extensive abnormal data by instead collecting only normal welding data. The neural network learns the characteristics of normal welding states, and any deviation from this learned normal pattern is automatically identified as abnormal. This inversion eliminates the need for time-consuming experiments to reproduce various abnormal patterns while maintaining comprehensive welding state determination coverage.
Solution Approach 2:
The system enables the welding state determination to serve itself by automatically identifying abnormalities through deviation detection from normal patterns, without requiring manual preparation of abnormal test cases. The neural network self-adapts to recognize abnormal states by comparing against the learned normal pattern, reducing the need for extensive experimental data collection.
3Reliability
If traditional welding state determination method is used, then comprehensive abnormal pattern recognition can be achieved, but the method becomes difficult to implement due to data preparation requirements
Solution Approach 1:
The invention inverts the traditional approach by training the neural network to recognize only normal welding patterns instead of both normal and abnormal patterns. This is achieved by preparing only normal welding data for training, and the system reliably identifies abnormal states by detecting deviations from the learned normal pattern. This inversion significantly improves implementation ease by eliminating the need to collect, prepare, and manage extensive abnormal pattern data while maintaining reliable abnormal pattern recognition.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
Provided is a welding state determination device that can easily determine the welding state. The welding state determination device includes: an acquisition unit that acquires a pulse waveform of a pulse current or a pulse voltage supplied to an electrode for pulse arc welding, the pulse waveform including a falling portion, a rising portion, and a flat portion therebetween; a preprocessing unit that shapes the pulse waveform such that the flat portion has a predetermined width; and a determination unit that determines a state of the pulse arc welding based on a difference between the shaped pulse waveform and a normal pattern created based on a plurality of past shaped pulse waveforms.