Convolutional Neural Network for Absorbent Article Defect Inspection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current human-generated inspection algorithms for absorbent article manufacturing are limited in speed and complexity, constraining the scope of in-process inspections and requiring manual coding for sophisticated defect detection.
Innovation Solution
A method utilizing a convolutional neural network (CNN) to generate inspection algorithms, trained with databases of images or signals of defective and non-defective absorbent articles, enabling automated detection of defects and characteristics during the manufacturing process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If human-generated inspection algorithms are used, then the system is easier to implement, but the inspection speed and sophistication are limited
Solution Approach 1:
The patent replaces traditional human-generated inspection algorithms with a machine learning-based convolutional neural network (CNN) system. This substitution enables automated defect detection that operates at production line speeds without requiring manual programming of inspection logic, thereby resolving the contradiction between ease of implementation and inspection speed.
Solution Approach 2:
The patent transforms the inspection algorithm from a static, manually-coded system to a dynamic, self-learning system. By training the CNN on labeled defect images, the system automatically adapts its detection parameters and patterns, enabling both high-speed operation and sophisticated defect recognition without manual intervention in algorithm development.
2Device complexity
If human-generated inspection algorithms are used, then the system structure is simpler, but the scope of sophisticated in-process inspection is constrained
Solution Approach 1:
The patent replaces complex manual algorithm design with an automated machine learning system. The CNN architecture handles the complexity of sophisticated defect pattern recognition internally, allowing the system to achieve high adaptability and inspection scope without increasing operational complexity for users.
Solution Approach 2:
The patent performs preliminary training of the inspection algorithm using labeled defect images before deployment. This pre-training phase enables the system to learn complex defect patterns and characteristics in advance, so that during actual production inspection, the system can handle sophisticated inspection tasks without requiring complex real-time processing or manual intervention.
3Measurement precision
If traditional inspection algorithms are used, then manual coding is required for sophisticated defect detection, but this limits the speed of algorithm development
Solution Approach 1:
The patent replaces manual algorithm coding with automated machine learning model training. Instead of programmers manually writing inspection logic, the system automatically learns defect detection patterns from labeled images, dramatically reducing algorithm development time while maintaining or improving detection precision.
Solution Approach 2:
The patent enables the inspection system to self-improve through automated training on labeled defect data. The CNN automatically adjusts its parameters and learns new defect patterns without human intervention in the coding process, allowing the system to adapt to new defect types and improve precision over time without requiring manual algorithm reprogramming.
Data Source
AI summary
A method for inspecting absorbent articles is provided. The inspection is performed using an inspection algorithm generated with a convolutional neural network having convolutional neural network parameters. The convolutional neural network parameters are generated by a training algorithm. Based on the inspection, characteristics of the absorbent articles, such as defects, can be identified. Absorbent articles having identified characteristics can be rejected, or other actions can be taken.