Printed circuit boards (PCBs) are a nearly ubiquitous component of every kind of electronic device. With the rapid development of integrated circuit and semiconductor technology, the size of a PCB can shrink down to a very tiny dimension. Therefore, high-precision and rapid defect detection in PCBs needs to be achieved. This paper reviews various defect detection methods in PCBs by analysing more than 100 related articles from 1990 to 2022. The methodology of how to prepare this overview of the PCB defect detection methods is firstly introduced. Secondly, manual defect detection methods are reviewed briefly. Then, traditional image processing-based, machine learning-based and deep learning-based defect detection methods are discussed in detail. Their algorithms, procedures, performances, advantages and limitations are explained and compared. The additional reviews of this paper are believed to provide more insightful viewpoints, which would help researchers understand current research trends and perform future work related to defect detection.

Effective detection of dense printed circuit board (PCB) components contributes to the optimization of automatic flow of production. In addition, PCB component recognition is also the essential prerequisite for early defect detection. Current PCB component detection approaches are not adept in both rapid and precise detection. YOLOv8 models have exhibited effective performances for detecting common objects, such as person, car, chair, dog etc. However, it is still tricky for YOLOv8 models to inspect dense and disparate PCB components precisely. Thus, a novel convolution neural network (CNN) model is proposed for dense PCB component detection by introducing several modifications onto YOLOv8. First, creative C2Focal module is designed as the core element of the backbone, combining both fine-grained local and coarse-grained global features concurrently. Then, the lightweight Ghost convolutions are inserted to effectively reduce the computation cost, meanwhile maintaining the detection performance. Finally, a new bounding box regression loss that is Sig-IoU loss, is proposed to facilitate the prediction regression and promote the positioning accuracy. The experiments on our PCB component dataset demonstrate that our proposed model performs the highest mean average precisions of 87.7% (mAP@0.5) and 75.3% (mAP@0.5:0.95) respectively, exceeding other state-of-the-arts. Besides, the detection speed hits 110 frames per second using RTX A4000, which is potential to realize the real-time inspection.

With the continuous evolution of Moore's law, the power density brought by three-dimensional integrated packaging technology has surged. Although the microchannel heat dissipation technology can effectively improve the heat dissipation efficiency, the traditional structure still has problems such as large pressure drop loss and uneven temperature field distribution. This study innovatively proposed a radiation manifold microchannel structure and constructed a three-dimensional fluid-solid coupling model. The single-objective sequential optimization strategy based on parameter scanning is adopted. First, the optimal outlet layout is determined, and then the width of the collector ring and the width of the microchannel end are optimized in turn. Finally, the global optimal solution is obtained by parameter combination. The results show that under the condition of four boundary heat sources with 3 × 3 mm2 and heat flux q of 400 W/cm2, when the system configuration is four-side opening outlet mode, the width of the collector ring is 0.75 mm, the width of the microchannel end is 0.75 mm, the width of the manifold is 2 mm, and the width of the central shunt channel is 2 mm, the comprehensive index of heat dissipation performance and flow characteristics is optimal. Compared with the initial model, the heat transfer efficiency is increased by 2.02%, and the flow performance is increased by 86.33%. The comprehensive heat dissipation performance and flow characteristics of the optimized model are significantly improved compared with the model before optimization, which provides a new idea for solving the heat dissipation problem of multi-heat source chips.

The yield and quality of blueberry cultivation depend heavily on effective fruit harvesting. Accurate and real-time identification of blueberries at various maturity stages is a prerequisite for precise harvesting. Accordingly, a lightweight network named MG-YOLO is developed for blueberry fruit detection in this study, through multiple enhancements on YOLO11-nano. First, the GBC block integrating gated convolutions and Bottleneck convolutions is adopted to improve adaptive feature discrimination ability, especially for densely occluded blueberries. Second, a novel MG-C3k2 module incorporating multi-scale Ghost convolution is proposed to strengthen feature representation while maintaining model efficiency. Finally, IF-CIoU loss, focusing on inner overlapping areas of bounding boxes and the balance between hard and easy samples, is designed to mitigate the class imbalance issue and accelerate bounding box regression, improving localisation precision. Experimental results demonstrate that the proposed MG-YOLO attains 85.7% mAP@0.5 and 60.6% mAP@0.5:0.95, representing remarkable improvements of 4.0% and 5.1% over the baseline, respectively. Furthermore, the parameter count is reduced by 12%, and the model exhibits a detection speed of 51 FPS measured on an NVIDIA RTX 3060 GPU, indicating its potential for for real-time automated harvesting applications.

Electronic components in the printed circuit board assembly (PCBA) exhibit great diversities in categories, visual appearances and geometric dimensions. Efficient and precise detection for dense components is crucial and challenging for early quality inspection. Therefore, this paper proposes a novel dense component detection network (DCD-Net) established on a few improvements upon YOLO11-nano. First, a novel lighter GFConv block integrating Ghost convolution and a channel fusion operation is built to downsize the model, meanwhile enhancing the detection precision. Second, a creative C3DS module incorporating dynamic snake convolution is designed for excellent adaptability to diverse geometric forms and the large-scale variance of electronic components. Finally, FC-IoU loss involving the hard-easy sample distribution is proposed to facilitate the bounding box regression, boosting the positioning accuracy. The experiment results verify that the proposed DCD-Net achieves remarkable mean average precision of 91.6% (mAP@0.5) and 77.9% (mAP@0.5:0.95), exceeding 0.9% and 2.4% increments separately than YOLO11-nano. In addition, a 93 frames/s inference speed is achieved for inputs with 640 × 640 pixel resolution, demonstrating its notable potential to realize real-time industry applications. The code is available at https://github.com/lqin0818/DCD-Net/tree/master.