PCB Solder Paste Anomaly Detection via Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current printed circuit board inspection methods, such as SPI, are inadequate in accurately detecting anomalies in solder pastes printed on PCBs, as they rely on optical technologies that may not capture subtle shape variations effectively.
Innovation Solution
A machine-learning based model is employed to derive probability values for measurement shape information of solder pastes printed through specific apertures, using both measurement and aperture shape information to detect anomalies by comparing the distribution of shape data from multiple apertures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If optical technology is used for SPI inspection, then the inspection process can be automated, but the detection precision for subtle shape variations is insufficient
Solution Approach 1:
The patent changes the inspection parameter from direct optical shape measurement to statistical distribution analysis. By comparing whether solder paste shapes conform to expected statistical distributions derived from aperture patterns, the system achieves higher precision in detecting subtle anomalies while maintaining automated inspection.
2Productivity
If traditional optical inspection methods are used, then the inspection speed is maintained, but the accuracy in detecting solder paste anomalies is inadequate
Solution Approach 1:
The patent replaces the mechanical/optical measurement system with a statistical computing system. Instead of relying on optical precision to detect shape variations, the system uses statistical distribution comparisons to identify anomalies, thereby improving detection accuracy without compromising inspection speed.
3Measurement precision
If detailed shape measurement is performed on each solder paste, then detection accuracy improves, but the complexity of the inspection system increases
Solution Approach 1:
The patent transforms the inspection approach from measuring detailed geometric parameters of each solder paste to comparing statistical distribution characteristics. This parameter transformation simplifies the inspection system while maintaining high detection accuracy, as statistical comparisons are computationally simpler than detailed shape analysis.
Data Source
AI summary
A printed circuit board inspection apparatus obtains measurement shape information about each of a plurality of solder pastes printed on a first printed circuit board through a plurality of apertures and aperture shape information about each of the plurality of apertures, obtains probability values that a first solder paste printed through a first aperture of the plurality of apertures and each of a plurality of second solder pastes printed through second apertures other than the first aperture of the plurality of apertures have the measurement shape information when the first solder paste and the plurality of second solder pastes are printed on the first printed circuit board, by applying the measurement shape information and the aperture shape information to a machine-learning based model, and detects whether an anomaly in the first solder paste occurred based on the probability values.