Predicting a Quality of a Printed Circuit Board Assembly from Production Data

Simple SummaryContent extracted from patent full text and abstract with AI.

This patent describes a computer-implemented method and system for predicting the quality of printed circuit board (PCB) assemblies by analyzing production data instead of requiring direct, post-production X-ray imaging. The system uses machine learning models (regression algorithms and autoencoders) trained on real X-ray images and related production data to map the production information onto a latent feature space. This mapping can then be used to either predict quality indicators or generate artificial X-ray images for further inspection, reducing the need for time-consuming and costly physical X-ray inspections.

Use CasesContent extracted from patent full text and abstract with AI.

• In-line quality prediction of PCB assemblies during electronics manufacturing, allowing early detection of faulty boards without needing physical X-ray scans.
• Automated quality inspection in high-volume PCB production environments to reduce human intervention and dependence on expert operators.
• Supporting predictive maintenance and yield analysis by identifying faulty production steps from aggregate production data and algorithmic assessment.
• Generating artificial X-ray images for operator review, training purposes, or further inspection by automated vision systems, simulating actual X-ray outputs without exposure.
• Customization for different regions of interest (e.g., specific solder joints) on complex multi-component PCBs for targeted inspection.

BenefitsContent extracted from patent full text and abstract with AI.

• Reduces costs and time associated with physical X-ray inspection by predicting PCB quality from readily available production data.
• Minimizes operator workload and dependence on specialized inspection expertise.
• Improves inspection throughput and enables earlier detection of defects in the production process.
• Reduces safety concerns and regulatory burdens associated with physical X-ray imaging.
• Enables comprehensive and consistent quality prediction, including for specific regions or types of potential fault on complex PCB assemblies.
• Facilitates integration with automated manufacturing and data-driven quality control systems.

Technical Classifications (CPCs)

Inventors & Applicants

Patent Abstract

A computer-implemented method of predicting a quality of a printed circuit board, PCB, assembly (1) comprising the steps of:Obtaining (S0) production data (101) relating to the production (100) of the PCB assembly (1),Mapping (S1), preferably based on a trained regression algorithm (102), the production data onto a latent vector (103) of a latent space of a trained adaptive algorithm (104), wherein the trained adaptive algorithm (104) is trained on real X-ray images (20) of PCB assemblies and/or serves for generating X-ray images (105) of PCB assemblies, and Determining (S2) a subspace (106) of the latent space related to the latent vector (103), the subspace (106) indicating a quality (107) of the PCB assembly (1), in particular of a region of interest (R1) of the PCB assembly (1), and/or Generating (S3), by the trained adaptive algorithm (104), based on the latent vector (103), an X-ray image (105) of the PCB assembly (1), in particular of the region of interest (R1), in order to determine a quality (107) of the PCB assembly (1).