FSW Joint Tensile Strength Prediction Using Temperature and Axial Force
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for predicting the tensile strength of friction stir welding (FSW) joints fail to accurately consider the dynamic influence of welding process parameters, particularly axial force and temperature, leading to incomplete understanding of their impact on joint quality.
Innovation Solution
A one-dimensional convolutional neural network (1D CNN) model is employed to predict tensile strength by using time series data of temperature and axial force from the advancing and retreating sides of the weldment surface, processed through infrared thermal imaging and force sensing, with data normalization and model training using a Huber robust loss function.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional welding parameter prediction methods are used, then the prediction process is simple, but the prediction accuracy is insufficient and cannot accurately reflect the dynamic influence of welding parameters on tensile strength
Solution Approach 1:
The patent replaces traditional mechanical/statistical prediction methods with an intelligent 1D CNN deep learning model. This substitution enables the system to automatically learn complex nonlinear relationships between welding parameters (temperature, axial force) and tensile strength, achieving high prediction accuracy (average relative error of 2%) without requiring manual feature engineering or simplified assumptions.
Solution Approach 2:
The patent introduces dynamic time-series data collection during the FSW process, capturing temperature and axial force variations at different welding stages. The 1D CNN model processes these temporal sequences to reflect the dynamic influence of welding parameters on tensile strength, moving from static parameter analysis to dynamic process characterization.
2Loss of information
If welding process parameters are used directly for prediction, then the measurement process is simple, but the influence of axial force and temperature time series data on tensile strength is not comprehensively considered
Solution Approach 1:
The patent implements a feedback mechanism by collecting real-time temperature and axial force data during the FSW process using infrared thermal imagers and force sensors. This feedback loop captures the actual welding process state, enabling the model to learn the true relationship between process parameters and joint quality, thereby preventing information loss about the dynamic welding behavior.
Solution Approach 2:
The patent introduces infrared thermal imagers and force sensors as intermediary devices to indirectly measure temperature and axial force that are difficult to access directly during welding. These intermediaries convert physical quantities into measurable signals, enabling comprehensive data collection without interfering with the welding process.
3Productivity
If destructive tensile tests are performed to evaluate joint quality, then the measurement is direct and accurate, but the assessment process is time-consuming and not suitable for real-time quality control
Solution Approach 1:
The patent creates a virtual copy of the tensile strength property through the 1D CNN prediction model. Instead of physically testing each joint, the model generates a predicted tensile strength value based on processed welding parameter data, providing a non-destructive alternative that maintains accuracy while dramatically improving assessment efficiency and enabling real-time quality control.
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
The model achieves accurate prediction of tensile strength with an average relative error of approximately 2%, providing a non-destructive assessment of FSW joint quality.
Implementation Method 1
the temperature of the AS and RS feature points of the weldment surface... are obtained by infrared thermal imager
Implementation Method 2
axial force... obtained by infrared thermal imager and force sensor
Implementation Method 3
The shoulder provides heat through friction during FSW process, which softens the welded metal
Data Source
AI summary
The invention belongs to the field of friction stir welding (FSW) quality prediction and relates to an intelligent prediction method for the tensile strength of FSW joints considering welding temperature and axial force. The invention uses a combination of experiment and theory. FSW experiment is carried out, the infrared thermal imager and force sensor are used to obtain the temperature of the feature points on the advancing side and retreating side of the outside of the shoulder of the weldment surface and the axial force during FSW process. The obtained data is used to train and test the one-dimensional convolutional neural network. The tensile strength prediction of friction stir welding is realized, which provided a reference for welding process control.