Printed Circuit Board Defect Detection Using Deep Learning via A Skip-Connected Convolutional Autoencoder

Abstract

As technology evolves, more components are integrated into printed circuit boards (PCBs) and the PCB layout increases. Because small defects on signal trace can cause significant damage to the system, PCB surface inspection is one of the most important quality control processes. Owing to the limitations of manual inspection, significant efforts have been made to automate the inspection by utilizing high resolution CCD or CMOS sensors. Despite the advanced sensor technology, setting the pass/fail criteria based on small failure samples has always been challenging in traditional machine vision approaches. To overcome these problems, we propose an advanced PCB inspection system based on a skip-connected convolutional autoencoder. The deep autoencoder model was trained to decode the original non-defect images from the defect images. The decoded images were then compared with the input image to identify the defect location. To overcome the small and imbalanced dataset in the early manufacturing stage, we applied appropriate image augmentation to improve the model training performance. The experimental results reveal that a simple unsupervised autoencoder model delivers promising performance, with a detection rate of up to 98% and a false pass rate below 1.7% for the test data, containing 3900 defect and non-defect images.

Do Autoencoders Need a Bottleneck for Anomaly Detection?

“…Several works have investigated the use of skip connections in AEs for tasks such as image denoising [22], [23] and audio separation [24]. Our work differs from current works on overcomplete and skip-AEs for anomaly detection [12]- [15], [25]: we have investigated a wider range of non-bottlenecked AEs, in which overcomplete or skip-AEs are only one type, and experimented with more datasets.…”

“…Skip connections allow a better flow of information in NNs with many layers, leading to a smoother loss landscape [41] and easier optimisation, without additional computational complexity [42]. Recent works [14], [15], [25] have reported that AEs with skip connections outperform those without on image anomaly detection. In preventing the skip-AEs from learning the identity function, Collin et al [14] and Baur et al [15] have implemented a denoising scheme and a dropout mechanism, respectively.…”

Do autoencoders need a bottleneck for anomaly detection?

“…Skip connections allow a better flow of information in NNs with many layers, leading to a smoother loss landscape [17] and easier optimisation, without additional computational complexity [18]. Recent works [19,20,21] have reported that AEs with skip connections outperform those without on image anomaly detection. In preventing the skip-AEs from learning the identity function, Collin et al [20] and Baur et al [19] have implemented a denoising scheme and a dropout mechanism, respectively.…”

A Novel Feature Extraction Algorithm and System for Flexible Integrated Circuit Packaging Substrate

“…Hu and Wang [ 15 ] used ResNet50 with a feature pyramid network as the backbone of feature extraction to better detect small defects on the PCB. Kim et al [ 16 ] proposed an advanced PCB inspection system based on skip connection convolutional autoencoders for the problem of setting pass/fail criteria for small fault samples. Ding et al [ 17 ] were inspired by the similarity measurement and proposed a multi-layer deep feature fusion method to calculate the similarity between the template and the defective circuit board.…”