M.A. Romashchenko1, D.V. Vasilchenko2, D.A. Pukhov3, S.Yu. Beleckaya4

1-4 Voronezh State Technical University (Voronezh, Russia)

Statement of a problem. At the moment, a large number of both software and hardware are involved in the production process of any electronic device. In the manufacture of electronic products, the elements are automatically installed on the board using robotic devices. Along with the installation of components, the quality of the printed circuit board pattern is also of serious importance. Quality control at any stage of production is assigned to a person, which significantly reduces not only productivity, but also the quality of rejection.

Purpose. Improving the quality of rejection of defective products in the conditions of conveyor production and reducing the number of defective products in the batch.

Results. A method for detecting defects on printed modules using artificial intelligence and machine vision has been developed and tested. Modern architectures of neural networks are analyzed, the most suitable one is selected and its own architecture based on them is implemented. The developed algorithm makes it possible to evaluate the printing module for the presence of defects such as mechanical damage (chips, scratches), excessive solder, the presence of a short circuit between the printed conductors, inaccuracy of component installation and errors in silkscreen printing.

Practical importance. The proposed solution is primarily aimed at improving the quality and increasing the speed of modern production of printed modules, the main indicator that it can provide is a reduction in the percentage of defects. A special feature is the high accuracy of defect detection and simultaneous communication with the automated control system of the production line, which will allow in real time to find the sources of emerging defects, enter information about them into the database and eliminate them in the process of work.

Romashchenko M.A., Vasilchenko D.V., Pukhov D.A., Beleckaya S.Yu. Method of carrying out defects of printed modules using neural networks. Radiotekhnika. 2022. V. 86. № 7. P. 44−49. DOI: https://doi.org/10.18127/j00338486-202207-08 (In Russian)

1. Romashhenko M.A., Vasil'chenko D.V., Rozhnenko S.N. Metodika ocenki vlijanija jelektromagnitnyh pomeh na funkcioniro-vanie jelektronnyh sredstv v processe ih proektirovanija. Radiotehnika. 2021. T. 85. № 6. S. 57-61 (In Russian).
2. Vasil'chenko D.V., Nekljudov A.L., Romashhenko M.A. Programmno-apparatnyj kompleks testirovanija jelektronnyh sredstv na vozdejstvie jelektromagnitnyh pomeh. Sb. trudov XXVI Mezhdunar. nauch.-tehnich. konf. «Radiolokacija, navigacija, svjaz'». 2020. S. 386-391 (In Russian).
3. Romashhenko M.A., Vasil'chenko D.V., Nekljudov A.L., Glotov V.V., Glotova T.S. Skaner blizhnego jelektricheskogo polja dlja dvuhstoronnih i mnogoslojnyh pechatnyh plat. RU 189820 U1, 05.06.2019. Zajavka № 2019108722 ot 26.03.2019 (In Russian).