IBE enables the world’s best brands. Through our unrivalled customer-centric approach, we partner with leading companies in numerous markets to deliver solutions for their most complex challenges.

Gallery
Contacts

IBe Industry Building, ShenZhen, China

ibe@ibepcbaaa.com

+86-75581785031

News
AI and Its Application in PCB AOI Test

With the continuous increase of labor costs, most of the physical labor and even a small part of mental labor in today’s PCB manufacturing have been replaced by mechanization, electrification, automation and information technology.

PCB practitioners try to use AI technology to solve practical problems in PCB manufacturing, further improve production efficiency and product quality, and even replace some workers. So what can AI do in the PCB AOI test? Where can AI be applied to PCB manufacturing?

Table of Contents

What is AI (artificial intelligence)?

Artificial intelligence is a discipline that studies how to make computers simulate certain human thinking processes and intelligent behaviors (such as learning, reasoning, thinking, planning, etc.), and is a branch of computer science. Research in this field mainly includes robotics, language recognition, image recognition, natural language processing, simulation systems and expert systems, etc. When people talk about artificial intelligence, concepts such as machine learning (ML), deep learning (DL), deep neural networks (DNN), and convolutional neural networks (CNN) are often mentioned.

AI (artificial intelligence)
AI (artificial intelligence)

Application scenarios of AI
The artificial intelligence system is mainly composed of three parts: 

①Information input. Perceive the dynamically changing physical world through various sensing devices to obtain a large amount of data; 

②Decision processing. Apply a large amount of acquired data to the model obtained by machine learning for reasoning, prediction or decision-making; 

③Execute output. According to the results of inference or prediction, corresponding actions are performed. In short, it is to build a prediction model on a large amount of input data through regression, integration and other machine learning algorithms, and apply the built model to the actual data set to obtain the prediction result. 

AI has been widely used in finance, medical care, education, public security, transportation, communications, agriculture, meteorology, and service industries.

AOI automatic optical inspection

The front briefly introduced the basic concepts of AI artificial intelligence, and mentioned the computer vision algorithms commonly used in AI, and these vision algorithms are also widely used in AOI.

AOI automatic optical inspection is evolved from manual visual inspection. The working principle is as follows: firstly, the required image feature information is “learned” on the standard CAM data through the visual algorithm, and then used in the training on the scanned image of each PCB.

PCB AOI (automatic optical inspection)
PCB AOI (automatic optical inspection)

The model learned on the set is used for feature extraction, the obtained feature image is compared with the standard data, and the problem points that need to be detected are reported according to the given rules (detection standards). Since AOI is a typical application of computer vision, it has the same difficulties as computer vision.

Visual difficulties in AOI

  • Information loss during imaging
  • Image occlusion
  • Local windows and global views

AI applications in AOI

As a key process of quality control in PCB manufacturing, in the AOI process, the confirmation process on the maintenance machine requires more manual participation. Operators need to classify false defects, repair defects, and scrap defects, and perform corresponding repair or marking actions on the board, and record quality reports at the same time. In these links, how to improve production efficiency, reduce production costs (especially labor costs) and reduce quality anomalies caused by human factors are issues of concern to the industry.

AI applications in AOI
AI applications in AOI

Therefore, AI technology is gradually being applied to AOI, including:

Loading and unloading
Loading and unloading include loading and unloading of AOI and maintenance machine. More and more manufacturers are already implementing automation solutions such as conveyor belts and robotic arms to improve efficiency and reduce labor costs. AGV trolleys are also considered for material transfer between AOI scanning machines and maintenance machines. If intelligent robots are used for loading and unloading, the cost and efficiency of intelligent robots need to be evaluated.

AOI detection
It mainly includes two parts: AOI equipment operation and logical operation. The AOI application system has simplified various operations (material number retrieval, alignment, light correction and application setting), and basically realizes one-button operation. Some manufacturers use the method of scanning QR codes to retrieve the part numbers that need to be scanned to ensure fast and accurate. Therefore, as far as the current situation is concerned, the operation process of AOI does not need AI technology optimization for the time being.

If it is mandatory to use a vision system or voice recognition system to operate the AOI system, the whole process will become more complicated. The detection logic of AOI has been iterated and optimized for many years, and evaluated according to the training process of deep learning, it is already the optimal detection model.

As for the efficiency of logic operations, like AI, a lot of matrix multiplication and convolution operations are used in AOI’s visual algorithms, because GPUs can efficiently handle matrix multiplication and convolution operations. It is foreseeable that GPU will be increasingly used in AOI to improve the efficiency of logic operations.

Therefore, post-AOI processing is a key link in the application of AI technology, that is, AI technology is applied to the confirmation of maintenance stations to reduce equipment investment and labor costs.

CVR confirmation
This process is the most concerned part, mainly including two parts of false point filtering and true point classification processing. If the rate of false points can be reduced, board handling, investment in maintenance equipment, and labor costs for maintenance will be reduced accordingly . The false points here do not refer to logical false points (defects reported without any abnormalities), but defects that do not want to be reported, such as dust, non-sticky foreign matter, and oxidation. According to statistics, such “false points” account for as little as 30% of the total defects and as many as 80% .

In actual production, when encountering such defects, maintenance station operators will use auxiliary tools (compressed air, sticky dust roller, eraser or fiber wipe, etc.) to clean and deal with them, and then confirm the type of defect. If you want to add an AI vision system to reduce such false points in this link, you must consider how to overcome the interference of various types of foreign matter, oxidation, etc., which requires the system to use 3D samples as processing objects, just like the abnormal analysis of PCB needs to use SEM or slice.

Considering that the proportion of such defects is as high as 3%~10%, if AI is simply applied on the basis of existing equipment, the result will inevitably be that although AI technology reduces labor costs, it will also seriously affect production quality.

Prospects of AI for real defect processing

So what is the prospect of AI being used to deal with real flaws? When the CVR operator confirms that the defect is a true defect, it usually divides the true defect into repair and scrap, and these defects can be processed online or offline. Online processing refers to the online repair of repair defects during the confirmation process, and the scrap information is marked on the board for scrap defects. Offline processing refers to only marking defects or entering defect codes during the confirmation process, and then sending them to other personnel or equipment for repair or scrapping.

Whether AI is used to filter false defects or classify true defects depends on the PCB factory’s false point rate, defect distribution, and PCB board structure, and it is necessary to formulate corresponding artificial intelligence application rules (just like designating lanes for driverless cars) ), or provide AI with a more powerful perception system to ensure that AI can clearly perceive the defects, such as optimizing the lighting system or using 3D imaging, which requires the addition of new hardware subsystems.

After adding the VVS system and AI technology to the CDB database system, the CVR only needs to deal with the shortcomings that cannot pass the judgment. When the AI has undergone sufficient learning and model iteration (especially for the continuous update of new materials and new processes), it will be possible to basically achieve full artificial intelligence confirmation.

Conclusion

Although AI technology has been widely used in AOI image processing, there are still huge application scenarios and space in the maintenance link, especially in today’s rapid iteration of AI algorithms, the future maintenance system will be an intelligent maintenance system integrating various AI algorithms. Of course, these AI algorithms can be simple regression and classification algorithms, or complex reinforcement learning and structured learning.

Leave a comment

Your email address will not be published. Required fields are marked *