Infrared Powder-Bed Defect Detection With CNN Defect Masks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for detecting defects during laser additive manufacturing are expensive, error-prone, and lack reliability, necessitating a more efficient approach to ensure quality control in the manufacturing process.
Innovation Solution
A method utilizing a convolutional neural network to process infrared spectrum images of the powder layer surface during laser scanning, generating a defect mask to identify defects such as fusion lacks, burned areas, or contamination, which can be represented in images or tensors, enhancing defect detection accuracy and reducing human error.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional photo-based defect detection is used, then defect detection is performed, but the method is expensive and error-prone with low reliability
Solution Approach 1:
The patent replaces the conventional mechanical/optical photo-based detection system with a thermal imaging system that detects temperature distributions. This substitution uses thermal radiation detection instead of visible light photography, enabling automated defect identification through temperature anomalies caused by defects like lack of fusion or contamination, thereby improving reliability while reducing operational complexity
Solution Approach 2:
The patent changes the detection parameter from visual appearance (photo-based) to temperature distribution (thermal imaging). By monitoring temperature fields during laser additive manufacturing, the system can identify defects through thermal anomalies, transforming the detection approach from qualitative visual inspection to quantitative thermal measurement, which enhances reliability and enables automated analysis
2Productivity
If manual photo interpretation is used, then defect detection is performed, but it requires specialized operators and is time-consuming
Solution Approach 1:
The patent implements an automated thermal imaging system that performs defect detection without requiring specialized human operators. The system captures thermal images, processes temperature distributions, and identifies defects automatically through algorithmic analysis, enabling the detection process to serve itself without external expert intervention, thereby dramatically improving productivity and eliminating time losses associated with manual inspection
Solution Approach 2:
The patent replaces the manual photo interpretation process with automated thermal image processing systems. Instead of relying on human operators to visually analyze photos, the system uses computational algorithms to automatically detect temperature anomalies and identify defects, substituting human cognitive processing with automated computational analysis, which accelerates detection speed and eliminates delays
3Reliability
If conventional detection methods are used, then quality control is performed, but it lacks real-time capability for immediate corrective actions
Solution Approach 1:
The patent implements continuous real-time thermal monitoring during the laser additive manufacturing process. The thermal imaging system continuously captures temperature distributions throughout the manufacturing cycle, enabling uninterrupted defect detection and immediate identification of quality issues, which maintains reliable quality control while eliminating delays and enabling instantaneous corrective actions
Solution Approach 2:
The patent establishes a real-time feedback loop where thermal imaging data is continuously processed and analyzed during manufacturing. The system provides immediate feedback on temperature anomalies and potential defects, enabling operators to take corrective actions promptly. This closed-loop feedback mechanism ensures reliable quality control while minimizing response time, as defects are identified and addressed during the manufacturing process rather than afterward
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 method effectively detects defects during laser additive manufacturing, improving the reliability and cost-effectiveness of quality control by using neural networks to analyze images in real-time, enabling immediate corrective actions and enhancing the overall manufacturing process.
Implementation Method 1
a first image is captured, said first image being the image, captured in the infrared spectrum, of an upper surface of a layer of powder exposed to laser scan
Data Source
AI summary
A method for detecting defects during laser additive manufacturing, comprising the following steps: B21) a first image is captured, the first image being the image, captured in the infrared spectrum, of an upper surface of a layer of powder exposed to laser scan; B23) the first image is processed using a first convolutional neural network of self-encoding type, in such a way as to produce a defect mask indicating the location of defects at the upper surface of the layer of powder. A method for manufacturing parts during which the presence of defects is detected using the preceding method. A data-processing device, computer program and storage medium for the implementation of this method.