BiW Spot Welding Deformation Prediction Using Graph Convolution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The deformation mechanism of body-in-white (BiW) structures during spot welding is complex and lacks effective prediction and control methods, hindering precision control and production quality improvements in automobile manufacturing.
Innovation Solution
A method is developed using a graph convolutional network (GCN) to construct a BiW spot welding deformation prediction model by acquiring 3D coordinate measurements, extracting welding features, encoding them into hidden space vectors, constructing a graph topology, and decomposing Laplacian eigenvectors to predict deformations through multi-layer GCN layers with adaptive frequency domain filters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If traditional welding control methods are used, then the welding process can be completed, but the deformation prediction precision is insufficient and cannot achieve precision control
Solution Approach 1:
The patent replaces traditional mechanical deformation control methods with an information processing system. A graph convolutional network (GCN) based on artificial intelligence is used to process welding data, extract features, and predict deformation. This substitution of mechanical control with intelligent information processing enables precise deformation prediction and control that was not achievable with conventional methods.
Solution Approach 2:
The patent transforms physical welding parameters (current, voltage, time, temperature) into data representations suitable for AI processing. By converting physical quantities into numerical parameters and using graph convolutional networks to process these parameters, the system achieves accurate deformation prediction. The transformation of physical parameters into processable data forms enables the application of advanced algorithms for precision control.
2Strength
If more spot welds are added to improve BiW structure, then the structural strength increases, but the deformation control difficulty increases due to complex combined action
Solution Approach 1:
The patent segments the complex welding process into discrete, analyzable components. Each spot weld is treated as an independent unit with its own features and parameters. The graph convolutional network processes these segmented weld units individually while capturing their relationships, enabling manageable analysis of complex multi-weld structures without being overwhelmed by the overall system complexity.
Solution Approach 2:
The patent introduces an AI-based graph convolutional network as an intermediary between the complex welding process and the control system. This intermediary processes the complex data from multiple spot welds, extracts meaningful patterns, and provides deformation predictions. The GCN acts as a mediator that simplifies the complexity by transforming raw welding data into actionable predictions, making the control of complex multi-weld structures feasible.
3Measurement precision
If traditional data processing methods are used, then the processing speed is acceptable, but the prediction accuracy is insufficient for quality improvement
Solution Approach 1:
The patent performs preliminary processing of welding data by constructing graph structures and extracting features before the actual prediction process. By pre-processing the data into meaningful representations and organizing it in graph format, the system prepares the information in advance for efficient processing by the GCN. This preliminary action enables both high accuracy and fast prediction speed during actual production.
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 enables rapid and accurate prediction of BiW deformations, learning thermodynamic and kinetic representations, and allowing for quick overall deformation prediction in automobile manufacturing with GPU acceleration.
Implementation Method 1
During spot welding, a large current of several hundred to tens of thousands of amperes is applied between two electrodes and the workpiece, and the interface of the workpiece is melted through the combined action of pressure and Joule heating
Data Source
AI summary
A method for constructing a body-in-white (BiW) spot welding deformation prediction model based on a graph convolutional network (GCN) includes: 1) acquiring a welding feature and 3D coordinates of a spot weld to form an eigenvector and extracting designed 3D coordinates at each 3D coordinate measurement point; 2) encoding, by an encoder, eigenvectors and designed 3D coordinate vectors into hidden space vectors of spot welds and hidden space vectors of the coordinate measurement points, respectively, and constructing a graph topology G through a k-nearest neighbors algorithm; 3) decomposing a Laplacian eigenvector of the constructed graph topology G to acquire frequency domain components, and linearly transforming eigenvalues corresponding to the frequency domain components to construct a multi-layer GCN; 4) inputting the thermodynamic and kinetic information of each coordinate measurement point into a deep neural network and decoding a final deformation at each coordinate measurement point; and 5) optimizing the model.