A neural network construction method for combined optimization of visual detection and parameter regulation of a screen printing machine

CN122242581APending Publication Date: 2026-06-19FUJIAN FUQIANG PRECISION PRINTED CIRCUIT BOARD CO LTD

Patent Information

Authority / Receiving Office: CN · China
Patent Type: Applications(China)
Current Assignee / Owner: FUJIAN FUQIANG PRECISION PRINTED CIRCUIT BOARD CO LTD
Filing Date: 2026-04-28
Publication Date: 2026-06-19

AI Technical Summary

Technical Problem

Existing technologies for visual inspection and parameter control in screen printing machines suffer from problems such as inaccurate cross-modal feature alignment, severe information loss, insufficient gradient collaborative optimization, high network inference latency, and difficulty in meeting real-time closed-loop control requirements.

Method used

A neural network for joint optimization of visual detection and parameter control is constructed. Printing defect features are extracted through multi-scale convolution and spatial attention processing, and process parameter features are extracted by combining temporal convolution. A cross-modal fusion module is used for feature splicing and dimensionality reduction. Adaptive task weight allocation and physical constraint regression are introduced to achieve end-to-end joint training.

Benefits of technology

It improves cross-modal feature alignment accuracy, reduces information loss, enhances model generalization stability and robustness, meets the needs of real-time closed-loop control, and has the advantage of plug-and-play engineering deployment.

✦ Generated by Eureka AI based on patent content.

Smart Images

Figure CN122242581A_ABST

Patent Text Reader

Abstract

This invention relates to the field of neural network architecture design technology, and in particular to a neural network construction method for joint optimization of visual inspection and parameter control in screen printing machines. The method acquires time-synchronized images of screen-printed products and real-time process parameters of the screen printing machine, and completes standardized preprocessing. It constructs a dual-branch network for visual feature extraction and parameter encoding, respectively extracting spatial distribution features of printing defects and dynamic evolution features of process parameters. A spatial parameter cross-attention mechanism is used to achieve deep fusion of cross-modal features, generating a unified joint feature representation. A dual-head parallel prediction structure for defect detection and parameter control is designed, simultaneously outputting defect detection results and process parameter control increments. A multi-task joint loss function is constructed, and an adaptive task weight and gradient collaborative optimization strategy are introduced to achieve end-to-end joint training of the network. This invention achieves integrated closed-loop control of visual inspection and parameter control in screen printing machines, effectively improving defect recognition accuracy and parameter control response speed.

Need to check novelty before this filing date? Find Prior Art