Component Mounting System Machine Learning Inspection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing component mounting systems face challenges in ensuring the quality of boards with components mounted under inferior inspection conditions, where training data has not been sufficiently collected, leading to reduced inspection accuracy.
Innovation Solution
The system employs a learning device to capture post-component-mounting board images, acquire training data, and create learning data using machine learning. An inspection device performs component recognition image processing using this learning data, even under conditions with inferior inspection accuracy, while a memory device records images and inspection information for later review.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If inspection is performed under inferior inspection conditions (insufficient training data), then production efficiency is improved, but inspection accuracy deteriorates
Solution Approach 1:
The system performs preliminary actions by capturing and storing post-component-mounting board images and inspection results even when training data is insufficient. These images and results are stored in advance in the storage device, so that when sufficient training data becomes available, the learning device can retroactively improve inspection accuracy for previously inspected boards without requiring re-inspection, thus maintaining production efficiency while enabling future accuracy improvement.
2Productivity
If inspection is performed under inferior inspection conditions (insufficient training data), then production efficiency is improved, but board quality assurance deteriorates
Solution Approach 1:
The system implements feedback by continuously collecting inspection results and board images, storing them in the storage device. When sufficient training data accumulates, the learning device processes this feedback to improve inspection accuracy. This feedback mechanism ensures that board quality assurance improves over time while allowing production to continue at full speed even when training data is initially insufficient.
3Measurement precision
If learning data is created using machine learning based on training data, then inspection accuracy is improved, but the system complexity increases
Solution Approach 1:
The system introduces a storage device as an intermediary between the inspection device and the learning device. This intermediary stores board images and inspection results, allowing the learning device to process training data at its own pace without disrupting the inspection workflow. This intermediary approach improves inspection accuracy through machine learning while managing system complexity by decoupling the inspection and learning processes.
Data Source
AI summary
A component mounting system includes multiple component mounting devices lined up in a board conveyance direction, with each component mounting device having an imaging device to capture an image of a board. The component mounting system includes a learning device, an inspection device, and a memory device. The learning device captures a post-component-mounting board image after component mounting, acquires training data based on the captured post-component-mounting board image, and creates learning data using machine learning based on the acquired training data. The inspection device captures a post-component-mounting board image, and performs an inspection of the post-component-mounting board under a first inspection condition or under a second inspection condition having inspection accuracy inferior to that of the first inspection condition by performing component recognition image processing on the captured post-component-mounting board image using the learning data.