Surface Modification Defect Detection Using Overlapping Image Sequences
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automated quality assurance methods for surface modification processes, such as laser beam brazing and welding, are labor-intensive and require large training datasets to achieve accurate defect detection, leading to inefficiencies and inaccuracies, especially in detecting small defects.
Innovation Solution
A computer-implemented method using an image sequence with overlapping individual images, where a trained neural network assigns images to defect or non-defect classes, and outputs a defect signal if a predetermined number of consecutive images are classified as defective, allowing for real-time detection of both surface and device defects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual visual inspection is used for quality control, then labor intensity is high, but detection accuracy for small defects is limited
Solution Approach 1:
The patent replaces manual visual inspection with an automated optical inspection system that uses a camera to capture images of the surface modification process. This substitution eliminates human labor while providing more consistent and accurate defect detection, as the system can identify small defects that may be missed by human inspectors.
Solution Approach 2:
The patent introduces image processing algorithms as an intermediary between the camera and the quality assessment. These algorithms analyze captured images to automatically identify defects, serving as a mediator that transforms raw visual data into actionable quality information with higher precision than manual inspection.
2Productivity
If automated quality assurance using image analysis is implemented, then productivity increases, but measurement precision decreases due to difficulty in detecting small defects
Solution Approach 1:
The patent changes the parameters of the inspection system by using high-resolution imaging and advanced image processing algorithms. These parameter changes enable the automated system to detect small defects that would otherwise be missed, thereby improving measurement precision while maintaining high productivity through automation.
Solution Approach 2:
The patent transitions from manual 2D visual inspection to a multi-dimensional analysis approach using digital image processing. By capturing images in multiple dimensions (spatial resolution, grayscale values, and temporal sequences), the system achieves superior defect detection capability compared to traditional manual inspection methods.
3Measurement precision
If neural network-based image analysis is used, then defect detection accuracy improves, but device complexity and training data requirements increase
Solution Approach 1:
The patent applies preliminary action by pre-training the neural network model with extensive defect data before deployment. This preliminary training phase establishes the foundation for accurate defect classification, allowing the system to achieve high measurement precision while keeping the operational complexity manageable during actual inspection tasks.
Solution Approach 2:
The patent implements self-service through the neural network's ability to automatically learn and improve from captured images without requiring constant human intervention or recalibration. Once trained, the system autonomously performs defect classification, reducing the need for complex manual configuration and maintenance.
4Device complexity
If conventional image analysis methods are applied, then device complexity is reduced, but false positive and false negative rates increase
Solution Approach 1:
The patent incorporates feedback mechanisms where the neural network continuously refines its defect classification based on captured images and known defect patterns. This feedback loop reduces false positives and false negatives by learning from previous classifications, thereby improving reliability without significantly increasing device complexity.
Solution Approach 2:
The patent optimizes key parameters of the neural network model, such as learning rate, batch size, and network architecture depth, to achieve the best balance between reliability and complexity. By carefully tuning these parameters, the system minimizes false detections while maintaining manageable system complexity.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A computer-implemented method for determining defects 7 occurring during the execution of a surface modification process of a surface area 8 of a component is specified.The procedure comprises the following steps: 1. Providing a sequence of images 5 of a surface area 8 to be assessed, comprising several individual images 6, wherein each individual image 6 shows a section 9 of the surface area 8 and the sections 9 of the individual images 6 overlap at least partially (S2); 2. Assigning the individual images 6 to at least two image classes 10a, 10b, of which at least one image class 10b bears the attribute "defective", hereinafter referred to as defect image class 10b (S4); 3. Checking whether several individual images 6 of a predefined number of directly consecutive individual images 6 of the defect image class 10b in the sequence 5 have been assigned (S5); and if several individual images 6 of the predefined number of directly consecutive individual images 6 of the defect image class 10b have been assigned, 4. Outputting a defect signal (S6).In addition, a device and a computer program for determining defects 7 occurring during the execution of a surface modification process of a surface area 8 of a component are specified.