scholarly journals SMT Assembly Inspection Using Dual-Stream Convolutional Networks and Two Solder Regions

2020 ◽  
Vol 10 (13) ◽  
pp. 4598
Author(s):  
Young-Gyu Kim ◽  
Tae-Hyoung Park
Keyword(s):  
Neural Network ◽  
Convolutional Neural Network ◽  
Classification Accuracy ◽  
Printed Circuit Board ◽  
Circuit Board ◽  
Printed Circuit ◽  
Convolutional Networks ◽  
Single Stream ◽  
Automated Optical Inspection ◽  
Network Weight

The automated optical inspection of a surface mount technology line inspects a printed circuit board for quality assurance, and subsequently classifies the chip assembly defects. However, it is difficult to improve the accuracy of previous defect classification methods using full chip component images with single-stream convolutional neural networks due to interference elements such as silk lines included in a printed circuit board image. This paper proposes a late-merge dual-stream convolutional neural network to increase the classification accuracy. Two solder regions are extracted from a printed circuit board image and are input to a convolutional neural network with a merge stage. A new convolutional neural network structure is then proposed that is able to classify for defects. Since defect features are concentrated in solder regions, the classification accuracy is increased. In addition, the network weight is reduced due to a reduction of the input data. Experimental results for the proposed method show a 5.3% higher performance in F1-score than a single-stream convolutional neural network based on full chip component images.

A novel convolutional neural network for electronic component classification with diverse backgrounds

2021 ◽  
Author(s):  
Longfei Zhou ◽  
Lin Zhang
Keyword(s):  
Neural Network ◽  
Computer Vision ◽  
Convolutional Neural Network ◽  
Printed Circuit Board ◽  
Manufacturing Industries ◽  
Circuit Board ◽  
Electronics Manufacturing ◽  
Electronic Components ◽  
Printed Circuit ◽  
Single Category

The rapid development of computer vision techniques has brought new opportunities for manufacturing industries, accelerating the intelligence of manufacturing systems in terms of product quality assurance, automatic assembly, and industrial robot control. In the electronics manufacturing industry, intensive variability in component shapes and colors, background brightness, and visual contrast between components and background results in difficulties in printed circuit board image classification. In this paper, we apply computer vision techniques to detect diverse electronic components from their background images, which is a challenging problem in electronics manufacturing industries because there are multiple types of components mounted on the same printed circuit board. Specifically, a 13-layer convolutional neural network (ECON) is proposed to detect electronic components either of a single category or of diverse categories. The proposed network consists of five Convolution-MaxPooling blocks, followed by a flattened layer and two fully connected layers. An electronic component image dataset from a real manufacturing company is applied to compare the performance between ECON, Xception, VGG16, and VGG19. In this dataset, there are 11 categories of components as well as their background images. Results show that ECON has higher accuracy in both single-category and diverse component classification than the other networks.

Hierarchical pyramidal subsampling in deep convolutional networks for visual pattern recognition

2021 ◽  
Author(s):  
A. E. Averyanikhin ◽  
A. I. Vlasov ◽  
E. V. Evdokimova
Keyword(s):  
Printed Circuit Boards ◽  
Printed Circuit Board ◽  
Flaw Detection ◽  
Input Image ◽  
Circuit Board ◽  
Digital Data ◽  
Deep Convolutional Neural Networks ◽  
Size Scale ◽  
Printed Circuit ◽  
Convolutional Networks

The main problem of known deep convolutional neural networks (CNN) is that they require a fixed-size input image. This requirement is “artificial” and can reduce recognition accuracy for images or its parts of arbitrary size/scale. The paper proposes a strategy of combining “hierarchical pyramidal subselection” to eliminate the above restriction. The structure of the neural network using the proposed combining strategy allows the generation of prediction regardless of the size/scale of the original image, and also improves the accuracy of recognition. Features of application of CNN for identification and recognition of defects of conducting pattern of printed circuit board blanks have been considered. Features of defects of conductive pattern of printed circuit board blanks have been briefly discussed. The invention proposes the use of artificial CNN, which have advantages in speed and accuracy in solving problems of object recognition on images relative to existing methods. The focus is on the architecture of CNN using hierarchical pyramidal subselection. Capabilities of application of CNN for recognition of defects of conducting pattern of printed circuit board blanks have been shown. Proposed method of hierarchical pyramidal subselection in deep convolutional networks has been implemented in software complex, which allows processing digital data of photographs of conducting pattern of printed circuit boards, in particular during their flaw detection, and can be used for localization of existing defects of conducting pattern. The conclusion draws the possibilities of using methods and means of image processing in flaw detection of radio-electronic equipment and instruments

Inspection System of Soldering Joint on Printed Circuit Board by Using Neural Network

1995 ◽  
Vol 7 (3) ◽  
pp. 225-229
Author(s):  
Shunichiro Oe ◽  
◽  
Kennichi Kaida ◽  
Daisuke Nagai ◽  
Mituo Nakamura ◽  
...  
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Semiconductor Laser ◽  
Printed Circuit Board ◽  
The State ◽  
Circuit Board ◽  
Inspection System ◽  
Sensor Unit ◽  
Printed Circuit ◽  
The Neural Networks

This paper deals with a new inspection system of soldering joint on printed circuit board by using neural network. A sensor unit of this system consists of a semiconductor laser unit, four PSDs, and a pin photo-diode. We can obtain four types of images which are called height image, PSD brightness image, vertical image and vector image, by using four sensor units. We extract the features which show the state of soldering joint from these images and develop an inspection system using the neural networks constructed for the features and the state of soldering joint.

scholarly journals Estimation of the PCB Production Process Using a Neural Network

2020 ◽  
Vol 15 ◽  
pp. 01-07
Author(s):  
Kuo-Hsien Hsia ◽  
Jr-Hung Guo
Keyword(s):  
Neural Network ◽  
Production Process ◽  
Manufacturing Process ◽  
Printed Circuit Boards ◽  
Printed Circuit Board ◽  
Circuit Board ◽  
Circuit Boards ◽  
Electronic Products ◽  
Printed Circuit ◽  
Board Industry

Printed Circuit Boards (PCB) are an integral part of all electronic products, and the production process for printed circuit boards is quite complex. As the life cycle of electronic products becomes shorter and shorter, and the precision and signal bandwidth of electronic products become higher and higher, the manufacturing process of printed circuit boards is further complicated. Therefore, how to pre-evaluate the production difficulty before starting the production will effectively increase the efficiency and quality of printed circuit board production. Gerber file is the most commonly used data format for the printed circuit board industry. This file contains most of the parameters required for the manufacture of printed circuit boards. Therefore, this study uses a neural network to evaluate new PCB products before they are produced through the production parameters that are more influential in the PCB manufacturing process. This makes it possible to evaluate the difficulty and the required production process before the new PCB product is produced. This will be very beneficial for the PCB production schedule, quality control, and cost.

scholarly journals Comparative results of the classification accuracy of MSTAR dataset radar images by convolutional neural networks with different architectures.

2021 ◽  
Vol 2021 (11) ◽  
Author(s):  
I.F. Kupryashkin ◽  
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Convolutional Neural Network ◽  
Classification Accuracy ◽  
Training Data ◽  
Convolutional Networks ◽  
Radar Images ◽  
Neural Network Classification ◽  
Learning Technique ◽  
Comparative Results

The results of MSTAR objects ten-classes classification using a VGG-type deep convolutional neural network with eight convolutional layers are presented. The maximum accuracy achieved by the network was 97.91%. In addition, the results of the MobileNetV1, Xception, InceptionV3, ResNet50, InceptionResNetV2, DenseNet121 networks, prepared using the transfer learning technique, are presented. It is shown that in the problem under consideration, the use of the listed pretrained convolutional networks did not improve the classification accuracy, which ranged from 93.79% to 97.36%. It has been established that even visually unobservable local features of the terrain background near each type of object are capable of providing a classification accuracy of about 51% (and not the expected 10% for a ten-alternative classification) even in the absence of object and their shadows. The procedure for preparing training data is described, which ensures the elimination of the influence of the terrain background on the result of neural network classification.

Sign in / Sign up

Export Citation Format

Share Document