Karteikarten: Automated PCB Fault Diagnosis — 56 Karten

What is a printed circuit board made of?

Glass-reinforced plastic with etched copper tracks.

Why are PCBs used in almost all commercial electronic devices?

Because they are rugged, inexpensive, and highly reliable.

What are the main steps in bare PCB fabrication?

Copper foil coverage, photoresist film formation, etching, and film stripping.

Name one type of defect identified in PCB manufacturing.

Missing components.

Name another type of PCB defect besides missing components.

Misalignment.

What PCB defect involves incorrect electrical connection polarity?

Wrong polarity.

What tools does the research use for image annotation and model training?

Roboflow is used for image annotation and model training.

Which metrics evaluate the fault-diagnosis system's performance?

Accuracy, precision, recall, F1-score, mean Average Precision, and inference time.

What inspection methods are compared in the study?

Automated computer-vision inspection and conventional manual inspection.

Which aspects are compared between automated and manual inspection?

Defect-detection accuracy, inspection time, and consistency.

Which activities are excluded from the study's scope?

PCB repair, real-time deployment, hardware integration, 3D analysis, thermal imaging, X-ray, electrical testing, and cost evaluation.

What is the first step in the Roboflow workflow for PCB images?

Uploading PCB images.

Which stages follow annotation in the Roboflow workflow?

Preprocessing and augmenting images.

How does the Roboflow workflow prepare datasets for model training?

By versioning datasets, splitting data, and exporting datasets.

What is the initial step in the complete data-processing cycle for PCB defect detection?

PCB image collection.

Which step comes after Roboflow annotation in the data-processing cycle?

Preprocessing and augmentation.

What are the final steps in the complete data-processing cycle?

Quantitative evaluation, statistical analysis, and interpretation of results.

What inputs does the conceptual framework for PCB defect detection include?

PCB image dataset, Roboflow annotation tool, defect categories, training parameters, and CNN or object-detection algorithm.

What assumptions does the study make about the images used?

Images are correctly labeled, clear, representative of defects, and independent between training and testing.

What is a convolutional neural network in visual inspection?

A deep-learning architecture that learns complex visual defect features automatically.

What trade-offs do YOLO, Faster R-CNN, and RetinaNet offer?

Different balances between object-detection inference speed and accuracy.

What mAP did YOLOv4 achieve on a custom PCB dataset?

92.3% mAP.

How many images and defect categories were in the YOLOv4 PCB dataset?

2,500 annotated images across eight defect categories.

What CNN accuracy is reported on standardized PCB defect datasets?

Exceeding 95%.

Name one research gap identified in PCB defect detection literature.

Limited Roboflow integration.

What is a limitation related to practical deployment in PCB defect detection research?

Insufficient practical deployment evaluation.

What issue involves dataset generalization in PCB defect detection studies?

Limited dataset generalization.

What datasets are created from PCB images in the methodology?

Training, validation, and testing datasets are created.

Which tool is used for preprocessing and annotating PCB images?

Roboflow is used for preprocessing and annotating.

What is the final step in the PCB defect detection methodology?

Evaluating the model's results quantitatively.

How many respondents were interviewed in the study?

Thirty respondents were interviewed.

Who were included among the study's respondents?

IT experts and laboratory personnel were included.

Name two software tools included in the proposed environment.

Python 3.10 or later and Roboflow are included.

What is the minimum recommended processor for the hardware?

An Intel Core i5 or Ryzen 5 processor is recommended.

What GPU is recommended for the hardware setup?

An NVIDIA RTX 3060 or higher GPU is recommended.

How is accuracy calculated using TP, TN, FP, and FN?

Accuracy = (TP + TN) / (TP + TN + FP + FN).

What does precision measure in defect prediction?

Precision measures how many predicted defects are actually defects.

How is recall calculated in defect detection?

Recall = TP / (TP + FN).

What does the F1-score balance in performance metrics?

F1-score balances precision and recall.

How is inference time calculated for image processing?

Inference time = total processing time divided by number of images.

Which test compares two models independently?

Independent tests compare two models.

What is the purpose of Tukey HSD in statistical analysis?

Tukey HSD is used after significant ANOVA results.

What accuracy did the reported model achieve?

The model achieved 96.8% accuracy.

What is the first step in the PCB inspection flow?

Capturing a PCB image.

What does the system display when no defect is detected?

It displays “No Defect”.

Which defect categories can the system classify?

Missing holes, short circuits, mouse bites, open circuits, spurs, and spurious copper.

What functions does Roboflow simplify in the inspection workflow?

Image annotation, dataset management, augmentation, model training, and deployment.

What image processing steps occur before detection in the inspection flow?

Resizing, enhancing, normalizing, and cropping the image.

What outputs are shown when a defect is detected?

Bounding boxes, class labels, and confidence scores.

What are the main steps after image preprocessing in the inspection flow?

Sending through Roboflow pipeline, loading model, performing detection, classifying faults, displaying and storing results.

What did the developed system detect and classify in PCB images?

Multiple PCB defects.

How does the developed system's image processing compare to human inspection?

It processes images more rapidly and consistently than human inspection.

What aids inspectors in locating defects and verifying classifications in real time?

Visualization of bounding boxes and class labels.

What factors affect detection performance in PCB defect detection?

High-quality annotated data, dataset diversity, image quality, lighting, camera position, PCB design complexity, and rare defect representation.

What should future improvements in PCB defect detection include?

Expanding datasets, comparing architectures, enhancing preprocessing, integrating equipment, retraining models, and evaluating classification.

Name some potential future integrations for the PCB defect detection system.

Industrial cameras, conveyor systems, robotic platforms, IoT devices, cloud computing, manufacturing execution systems, and predictive analytics.