PCBA Inspection Classifier for AOI False-Positive Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automated optical inspection (AOI) systems for printed circuit board assemblies (PCBAs) produce a high number of false positives, leading to increased manual inspection efforts, reduced productivity, and elevated production costs due to rework and manual handling, which complicates the manufacturing process.
Innovation Solution
A machine learning-based classifier using numerical measurement results from AOI systems to filter out pseudo errors, reducing manual inspection workload by classifying inspection results as either 'true errors' or 'pseudo errors', and controlling production flow accordingly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual inspection is performed to verify AOI results, then false positives can be identified and removed, but inspection time and labor costs increase
Solution Approach 1:
A machine learning classifier is introduced as an intermediary system between the AOI system and manual inspection. The classifier processes numerical measurement results from the AOI system and automatically identifies false positives, serving as a mediator that filters out erroneous detections before they reach manual inspectors. This reduces the burden on manual inspectors while maintaining high accuracy.
Solution Approach 2:
The patent replaces the mechanical manual inspection process with an automated machine learning-based classification system. Instead of relying solely on human inspectors to verify AOI results, the system uses trained classifiers that automatically analyze numerical measurement data and distinguish true defects from false positives, thereby reducing inspection time while maintaining or improving accuracy.
2Measurement precision
If multiple cameras and viewing angles are used in the AOI system, then inspection coverage is improved, but false positives increase
Solution Approach 1:
The system implements a feedback mechanism where numerical measurement results from multiple cameras and viewing angles are collected and processed by a machine learning classifier. The classifier learns from the combined data to distinguish between true defects and false positives, using the feedback from multiple perspectives to improve detection accuracy while reducing false alarms.
Solution Approach 2:
The patent transforms visual inspection data from multiple cameras into numerical measurement parameters that can be processed by machine learning classifiers. By changing the representation of inspection data from raw images to structured numerical features, the system enables automated classification that can handle complex multi-angle data while reducing false positives.
3Productivity
If AOI inspection is performed at high speed, then productivity is maintained, but false positive rate increases
Solution Approach 1:
The system performs preliminary automated classification of AOI results using machine learning models before manual inspection. By pre-processing and filtering AOI results through trained classifiers that analyze numerical measurement data, the system quickly identifies and removes false positives, maintaining high inspection speed while improving accuracy.
Solution Approach 2:
The machine learning classifier operates autonomously to self-correct false positives generated by the AOI system. The classifier independently analyzes numerical measurement results and automatically distinguishes true defects from false alarms without requiring immediate human intervention, thereby maintaining high productivity while reducing false positive rates.
Data Source
AI summary
A computer-implemented method of inspecting a printed circuit board assembly (1), PCBA, comprising the steps of: obtaining (S1) numerical measurement results (Feat 1, . . . Feat n) of one or more inspection types from a plurality of inspection types of an automated optical inspection, AOI, system (10), wherein the numerical measurement results are generated by the AOI system, entering (S2) the numerical measurement results into at least one section (a1, a2, an) of a feature vector (FV) associated with the one or more inspection types, the feature vector (FV) comprising a plurality of sections each of which is associated with a respective inspection type from the plurality of inspection types, selecting (S3) input features (IF) from the feature vector (FV) by dimensionality reduction of the feature vector (FV), inputting (S4) the input features (IF) into a classifier, wherein the classifier is capable of determining an error class (EC) of the PCBA (1) based on the input features (IF), outputting (S5) the error class (EC) as an inspection result of the PCBA (1).