Розроблення Інтелектуального Метода Контроля Якості Pad Контактів Цифрового Двійника Fek Delvotec 6400 Для Моделювання Процесу Клинового Зварювання У Мікроелектроніці
Year: 2026
Type: article
Abstract: The object of the study is the process of automated quality control of bonding pads in the digital twin of the FEK Delvotec 6400 equipment used for modeling the wedge bonding process in microelectronics. The subject of the study comprises the methods, mathematical models, and algorithmic tools for intelligent quality control of bonding pads in the digital twin of the FEK Delvotec 6400 equipment for modeling the wedge bonding process in microelectronics. The aim of the study is to develop an intelligent method for the automated assessment of bonding pad conditions, providing improved positioning accuracy, enhanced reliability of digital modeling, and better decision-making quality during the technological process. To achieve this aim, mathematical models for machine vision image formation, coordinate space calibration, neural network-based recognition of bonding pads and fiducial markers, refinement of center coordinates, coordinate transformations, automatic correction of the wire bonding trajectory, positioning accuracy assessment, and integrated validation of computer vision results prior to their use in the digital twin were developed. Methodologically, the study is based on mathematical modeling, modern computer vision techniques, deep learning, convolutional neural networks, object detection methods, digital image processing, automated analysis of the geometric parameters of bonding pads, and numerical modeling. For experimental validation, an augmented dataset of bonding pad images was created, a neural network model was trained, and a series of experiments was conducted to determine bonding pad detection completeness, the accuracy of center localization, errors in measuring geometric parameters, the reliability of defect classification, the integral validation index of the digital twin, and the processing time per image. The obtained results confirmed the high efficiency of the proposed approach, providing bonding pad detection completeness of up to 100%, center localization errors within 1.35–3.05 μm, errors in determining geometric parameters of up to 4.7%, a high level of defect classification reliability, and an image processing time not exceeding 50 ms. The practical significance of the study lies in the possibility of integrating the proposed method into the FEK Delvotec 6400 digital twin to enable automated quality control of bonding pads, improve the accuracy of wedge bonding process modeling, ensure timely defect detection, and maintain the stability of the technological process in microelectronic manufacturing.
Language: Ukrainian
Subfield: Industrial and Manufacturing Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7214123712
