Available via license: CC BY-SA 4.0
Content may be subject to copyright.
Bulletin of Electrical Engineering and Informatics
Vol. 13, No. 4, August 2024, pp. 2566~2585
ISSN: 2302-9285, DOI: 10.11591/eei.v13i4.7601 2566
Journal homepage: http://beei.org
Printed circuit board and printed circuit board assembly
methods for testing and visual inspection: a review
Nikolay Petkov1, Malinka Ivanova2
1Department of Electronics and Energy Engineering, Technical College of Sofia, Technical University of Sofia, Sofia, Bulgaria
2Department of Informatics, Faculty of Applied Mathematics and Informatics, Technical University of Sofia, Sofia, Bulgaria
Article Info
ABSTRACT
Article history:
Received Sep 25, 2023
Revised Oct 20, 2023
Accepted Oct 24, 2023
Testing and visual inspection of printed circuit boards (PCBs) and printed
circuit board assemblies (PCBAs) are important procedures in the
manufacturing process of electronic modules and devices related to locating
and identifying possible defects and failures. Earlier defects detection leads
to decreasing expenses, time and used resources to produce high quality
electronics. In this paper an exploration and analysis about the current
research regarding methods for PCB and PCBA testing, techniques for
defects detection and vusial inspection is performed. The impact of machine
and deep learning for testing and visual inspection procedures is also
investigated. The used methodology comprises bibliometric approach and
content analysis of papers, indexed in scientific database Scopus,
considering the queries: “PCB and testing” and “PCB and testing”, “printed
circuit board assembly and testing” and “PCBA and testing”, “PCB defect
detection” and “PCBA defect detection”, “PCB and visual inspection”, and
“PCBA and visual inspection”. The findings are presented in the form of a
framework, which summarizes the contemporary landscape of methods for
PCBs and PCBAs testing and visual inspection.
Keywords:
Artificial intelligence
Automation
Machine learning
Printed circuit board assemblie
testing
Printed circuit board testing
Visual inspection
This is an open access article under the CC BY-SA license.
Corresponding Author:
Malinka Ivanova
Department of Informatics, Faculty of Applied Mathematics and Informatics
Technical University of Sofia
Sofia, Bulgaria
Email: m_ivanova@tu-sofia.bg
1. INTRODUCTION
Testing of printed circuit boards (PCBs) and printed circuit board assemblies (PCBAs) is an
important procedure in manufacturing of electronic modules and devices that guarantee timely check for
quality of the performed operations and finally the quality of the product. Nowadays, the manufacturing is
“smart” as quality control is often performed through automatic optical inspection (AOI) to verify and
validate important production tasks [1], to prevent appearance of defects in electronics production and to
notify an operator [2], even for identification of micro-size defects on PCB [3]. Statistics approaches are also
applied such as failure mode and effects analysis for quality control improvement [4] and statistical process
control for monitoring the quality of the manufacturing process and for obtaining quality of the end product
[5]. Artificial intelligence is increasingly entering for supporting conductance of different tasks such as
optimization of the components position [6], as well as machine learning for detecting patterns or anomalies
on PCB [7] and deep learning for defects classification in PCB production [8].
PCBs are used to provide mechanical support of mounted electronic components. They are produced
from non-conductive material with conductive parts in the form of traces, pads, and conductive planes that
are printed or engraved. Electronic components are mounted on the circuit board and conductive traces
Bulletin of Electr Eng & Inf ISSN: 2302-9285
Printed circuit board and printed circuit board assembly methods for testing and … (Nikolay Petkov)
2567
connect them to create a working electrical circuit or PCBA mainly through surface mount technology [9],
[10]. PCBs possess important meaning for the final electronic product, because they provide electrical
connections between components, a rigid support to hold components, and a compact package that can be
integrated into a final product.
PCBs testing is a crucial step of their development cycle that leads to saving time and resources and
to prevent problems that could occur at the final production. These are the reasons some techniques for
analysis to be performed during the early stages of PCBs manufacturing and several testing methods to be
applied. These tests, conducted on prototypes or small scales, examine the potential shorts, solder joints, and
solder functionality, ensuring that each tested board will not possess defects and damages.
Testing of PCBAs is also a key step addressing the check whether all components are placed and
mounted correctly on the PCB and whether the assembly functions are as it is expected [11]. A wide variety
of testing methods and methodologies exists in electronics as some of them are well accepted in practice,
others are only in theoretical development. Anyway, new approaches also emerged to respond to the new
specific characteristics of contemporary technologies and requirements to the PCBs and PCBAs. Recently,
the role and contribution of artificial intelligence, machine and deep learning in PCBs and PCBAs testing is
explored as solutions for minimizing the time for testing, reducing the cost of the manufactured electronic
products, reducing efforts and resources and increasing the reliability of the final products [12], [13].
The goal of visual inspection is related to ensure PCBs and PCBAs quality control. The old way is
based on operators nake-eyes as recently are applied several modern approaches that are classified in three
groups: reference inspection methods, non-reference methods and hybrid inspection methods [14]. The
research questions in this work are defined as:
− What are the most investigated methods and techniques for PCB and PCBA testing and visual inspection
in the last five years? and
− What is the impact of machine and deep learning on methods for PCB and PCBA testing and visual
inspection?
The answers of these questions will lead to drawing and understanding where the researchers’
efforts are mainly directed recently, what are the main contemporary problems they solve and what are future
trends and directions. The aim of the paper is to summarize and analyze the actual and current methods and
techniques for PCBs and PCBAs testing and visual inspection as well as to outline the meaning of artificial
intelligence, machine and deep learning in the testing process and visual inspection of PCBs and PCBAs.
This will point out challenging problems, tendences and possible future lines for research. The contributions
in this work are related to:
− Conductance of an investigation and analysis of contemporary methods for PCB and PCBA testing and
visual inspection through applying bibliometric approach and examination the content of full-papers.
− Outlining the role of machine and deep learning for solving different problems in PCB and PCBA testing
and visual inspection.
− Creating a framework that summarizes the current research landscape regarding the most investigated
methods for PCB and PCBA testing and visual inspection.
The rest of the paper is organized as follows. In section 2 the used methodology is described.
Section 3 presents an investigation regarding methods for PCB testing. The contemporary methods for PCBA
testing are explored in section 4. Methods and techniques for PCB and PCBA defect detection are examined
in section 5. Section 6 includes investigation concerning methods for PCB and PCBA visual inspection. In
section 7 the main findings are summarized as a framework with contemporary approaches for PCB and
PCBA testing and visual inspection is created. Section 8 is the conclusion.
2. USED METHOD
Analysis of the current achievements in PCBs and PCBAs testing and visual inspection, including
the utilization of artificial intelligence and machine learning is performed following the several procedures
and the used methodology is presented in Figure 1.
a. Queries forming and search conducting: eight different queries are formed: “printed circuit board and
testing” and “PCB and testing”, “printed circuit board assembly and testing” and “PCBA and testing”, “PCB
defect detection” and “PCBA defect detection”, “PCB and visual inspection” and “PCBA and visual
inspection” and submitted for searching in article title, abstract and keywords in Scopus abstract and citation
scientific database. Scopus is chosen, because it includes a big amount of bibliometric data of high quality
articles in engineering and informatics domains. The bibliometric data are taken on 31.07.2023.
b. Applying filters and sorting: the first applied filter concerns the investigated term that settles a five-year period:
from 2018 to 2022, including the indexed papers during 2023 year. The second filter sets only scientific articles
in English language to be examined. The returned results are ordered according to their relevance.
ISSN: 2302-9285
Bulletin of Electr Eng & Inf, Vol. 13, No. 4, August 2024: 2566-2585
2568
c. Bibliometric analysis: bibliometric techniques are used to outline the whole picture in the investigated
area and to give an overall view on the current research. Bibliometric analysis is chosen, because it is
accepted and preferred approach that is utilized in a wide variety of fields for scanning the research
landscape and pointing out trend topics [15], [16] as well as for finding the most influenced journals,
authors, countries and researched themes [17], [18]. For conductance of bibliometric analysis, R studio
software and biblioshiny application are used [19]. The accent is given on: i) annual scientific production
to understand the interest to the investigated topics, ii) the most relevant sources that publish articles in
the researched topics, iii) the most frequent keyword used by authors to describe the articles content, iv)
trend topics that direct the contemporary research, and v) co-occurrence network to understand the
relationship between two terms and how terms are thematically organized in clusters.
d. Full-text papers analysis: analysis of the most relevant full-text papers is performed considering the
obtained results after submitting eight queries. Also, the open access articles are considered for further
examination, when it is possible. The analysis summarizes the used investigative approaches and obtained
findings.
e. Framework development: a framework that summarizes the revealed findings is created to demonstrate
the most common and contemporary researched methods and techniques for PCB and PCBA testing and
visual inspection.
Figure 1. Used method
3. EXPLORATION REGARDING TESTING OF PCBs
For bibliometric analysis, the search is performed in Scopus scientific database as queries “PCB and
testing” and “PCB and testing” are submitted. The returned results from the first query “PCB and testing”
comprises 876 documents and from the second query “PCB and testing” 875 documents considering the
described methodology in the previous section.
(a) Queries forming and search
conducting in Scopus
“printed circuit board and testing”
“printed circuit board assembly and
testing”
“PCB and testing”
“PCBA and testing”
“PCB defect detection”
“PCB and visual inspection”
“PCBA defect detection”
“PCBA and visual inspection”
(b) Applying filters and
sorting
First filter: Investigated period: from 2018 to 2023
Second filter: Articles in English language
Sort criteria: Relevance
(c) Bibliometric analysis
Annual scientific production
The most relevant sources
The most frequent author’ keywords
Trend topics
Co-occurrence network
(d) Full-text papers analysis
The authors aim of a given investigation
Used method/methodology/model by authors
Achieved results and obtained findings
(e) Framework development
Methods for PCBA testing and visual inspection
Methods for PCB testing and visual inspection
Bulletin of Electr Eng & Inf ISSN: 2302-9285
Printed circuit board and printed circuit board assembly methods for testing and … (Nikolay Petkov)
2569
The annual scientific production is depicted on Figures 2(a) and (b) as the both curves,
corresponding to the both queries, are characterized by irregularities. The published and indexed articles for
2022 are 174 and 167 for the first and second query, respectively. However, it can be noted that interest in
the topic has not decreased in the last five years as the number of published and indexed articles are nearly
similar. Documents so far for 2023 year are 86 for the first query and 81 for the second query.
(a)
(b)
Figure 2. Annual scientific production concerning the query: (a) “printed circuit board and testing” and
(b) “PCB and testing”
The articles are most often published in Proceedings of the Conference Electronic Components and
Technology Conference, IEEE Transactions on Components, Packaging and Manufacturing Technology,
Microelectronics Reliability, Sensors, Science of the Total Environment, Intersociety Conference on Thermal
and Thermomechanical Phenomena in Electronic Systems, IEEE ACCESS, IEEE Transactions on
Instrumentation and Measurement, Journal of Physics: Conference Series as details about the publishing
sources are presented through Figures 3(a) and (b). It can be seen that the papers are mainly submitted in
journals and scientific conference proceedings in the field of electronics, which is understandable since
testing is about ensuring quality of electronics production.
(a)
(b)
Figure 3. The most relevant sources for the query: (a) “printed circuit board and testing” and (b) “PCB and
testing”
The most used keywords by authors in their papers are related to: PCB and testing and different
methods for analysis like: deep learning, machine learning, failure analysis, defect detection, finite element
analysis, image processing (Figures 4(a) and (b)). Another the most utilized keywords are: reliability, solder
joint, eddy current, flexible PCB. In both requests, the authors describe the article content through applying
almost the same keywords in their articles. Author’s keywords analysis shows that in recent years the interest
ISSN: 2302-9285
Bulletin of Electr Eng & Inf, Vol. 13, No. 4, August 2024: 2566-2585
2570
related to the use of machine and deep learning in testing has grown, and this approach has developed in
parallel with conventional methods.
(a)
(b)
Figure 4. The most frequent keywords used by authors considering the query: (a) “printed circuit board and
testing” and (b) “PCB and testing”
Considering the both queries the trend topics for 2022 years are described with the keywords: deep
learning, machine learning, sensors, printed circuit board, reliability, testing, automotive, anand viscoplastic
model, conformal coating (Figures 5(a) and (b)). Trend topics also confirm and outline the increased research
that includes machine and deep learning at PCB testing.
(a)
(b)
Figure 5. Trend topics regarding the query: (a) “printed circuit board and testing” and (b) “PCB and testing”
The created co-occurrence networks consist of eight and seven different clusters considering
respectively the queries “PCB and testing” and “PCB and testing” (Figures 6(a) and (b)). It is noticeable that
the clusters are separately formed and unconnected or weakly connected to other clusters. The biggest cluster
considering the query “PCB and testing” is formed around the term reliability, which is connected with the
terms solder joint, PCBs, corrosion, electrochemical migration, humidity, electronics, thermal cycling. A
Bulletin of Electr Eng & Inf ISSN: 2302-9285
Printed circuit board and printed circuit board assembly methods for testing and … (Nikolay Petkov)
2571
separate cluster with linked terms: printed circuit board, deep learning, defect detection, and AOI also stands
out. A small cluster is organized around the terms PCB and failure analysis. The biggest cluster regarding the
query “PCB testing” is constructed around the core term PCB, which is connected with the terms reliability,
finite element analysis, testing, FPGA, and design. Another cluster brings the terms printed circuit board,
deep learning and defect detection together as this cluster is linked to a smaller cluster with the terms: PCB
inspection, computer vision, image processing. It can be said that the most mentioned term at PCB testing is
reliability. Also, at the both queries, clusters indicating relatedness of terms printed circuit board, deep
learning and defect detection are distinguished from other clusters.
(a)
(b)
Figure 6. Co-occurrence network regarding the query: (a) “printed circuit board and testing”and (b) “PCB
and testing”
The bibliometric analysis outlines the overall picture in the PCB testing and it can be summarized
that the number of articles published and indexed in Scopus during the studied period is close in value, which
speaks of the continuous interest in the topic by researchers. The articles are mainly published in scientific
proceedings of electronics conferences and journals, which is understandable since PCB testing is a research
field in electronics. It is interesting to note that among the most frequently used keywords by the authors are
not only terms specific to testing and electronics, but also terms from the field of informatics, such as
machine and deep learning. This leads us to the conclusion of the increased importance in our modern age of
machine and deep learning for testing. This conclusion is also confirmed by the terms involved in the trend
topics, as along with sensors, printed circuit board, reliability, testing, are also included deep learning and
machine learning. Clusters stand out in the examined co-occurrence networks, the largest of which include
terms from electronics, such as solder joint, PCBs, corrosion, electrochemical migration, humidity,
electronics, thermal cycling in one cluster and printed circuit board, deep learning, defect detection, and AOI
in another cluster. The meaning of the term reliability, which has the greatest frequency of usability by
authors, should be emphasized. A smaller cluster is also formed, uniting the terms: PCB inspection, computer
vision, image processing. It outlines the entry of machine and deep learning, as well as technologies from
computer vision and image processing into PCB testing. Further exploration of the topic is performed
through content analysis of relevant articles, indexed in Scopus and mainly with open access.
Azin et al. [20] propose a method for non-destructive testing, which combines acoustic emission
technique for finding design defects on PCB and X-ray tomography for identification of the dimension of the
found critical defects. Solder joints with identified defects are analyzed for understanding their stress-strain
state. The method is capable of identifying PCB design defects, to analyze whether they are critical and to
estimate the PCB residual usage.
The work of Elliot and Brown [21] is focused on investigation regarding the meaning of PCB traces
and their angle bends (no bend, 45°, and 90°) for the maximal current conductance, location and time for
possible failure. The presented methodology for testing is destructive and the findings outline that the
maximal current passes PCB traces in a similar way and there is no significant difference when different
angle bends are used (no bend, 45°, and 90°). The authors point out as future work an investigation regarding
the influence of width, thickness, and angle bends (different of 45° and 90°) of traces on failure and
reliability of PCBs.
ISSN: 2302-9285
Bulletin of Electr Eng & Inf, Vol. 13, No. 4, August 2024: 2566-2585
2572
Wileman et al. [22] present an approach for virtual testing of PCB and PCBA considering computer
aided design (CAD) model and techniques like computational fluid dynamic and finite element analysis.
Then, through simulations are tested different PCB characteristics and an assessment is performed regarding
the reliability of the electronic system. The used method is called physics of failure, showing PCB behavior
over time and how different mechanisms (physical and thermal) could lead to failure. This allows the risk
problems to be identified during design of the electronic product and some failure issues to be removed.
Li et al. [23] develop a new sensor for defects detection on metal surface, which functionality is
related to identified changes in frequency or magnitude. The testing is non-destructive and the sensor is
capable through electrode probe to locate notch damages, their dimension and orientation.
Oliveira et al. [24] propose a methodology to prevent excessive bending strain in important PCB’s
points when in-circuit test is performed. It is based on finite element analysis and is verified through
experimentation. This methodology gives the possibility PCB maximal strain at in-circuit test to be predicted
and to understand whether such test could damage PCB.
Volkau et al. [25] present an approach for usage of unsupervised deep learning and transfer learning
through deep convolutional neural network (CNN) for detecting PCB defects on images (scratch, broken
PCB edge, and hole). Defects on PCB images are found considering the distance to clusters with normal PCB
features as the recognition rate is high.
Nguyen and Bui [26] apply algorithms for feature extraction from images and supervised deep
learning to detect defects on PCB surface in real time. The proposed system combines brute force matching
technique, oriented FAST and rotated BRIEF (ORB) and random sample consensus (RANSAC) algorithms
for defect detection and for testing the PCB quality. The advantages are robustness against noise, high speed
computation (usage of ResNet-50), visual inspection in real time and with high precision.
Silva et al. [27] apply principles of transfer learning and VGG16/ResNet-50 pre-trained models for
identifying the defective PCBs and also non-referential method for inspection. Findings related the queries
“printed circuit board and testing” and “PCB and testing” are summarized in Table 1 as the
method/methodologies for testing and their aim are presented.
Table 1. Contemporary methods for PCB testing
Paper
Method/methodology
Aim
Azin et al. [20]
− Non-destructive testing
− Acoustic emission technique
− X-ray tomography
− Finding design defects on PCB
− Analyzing whether defects are critical
− Estimate the PCB residual usage
Elliot and Brown [21]
− Destructive testing
− PCB bend angle traces influence on the maximal
current conductance, location, and time
− Possible failure
Wileman et al. [22]
− Virtual testing of PCB and PCBA through
computational fluid dynamic and finite
element analysis (physics of failure)
− Risk problems to be identified during design of the
electronic product
− Failure issues to be removed
Li et al. [23]
− Sensor identified changes in frequency or
magnitude
− Non-destructive
− Defects detection on metal surface
− Locate notch damages, their dimension, and orientation
Oliveira et al. [24]
− Finite element analysis
− To prevent excessive bending strain in important PCB’s
points when in-circuit test is performed
− PCB maximal strain at in-circuit test to be predicted
− To understand whether such test could damage PCB
Volkau et al. [25]
− Unsupervised deep learning and transfer
learning
− Detecting PCB defects on images (scratch, broken PCB
edge, and hole)
Nguyen and Bui [26]
− Algorithms for feature extraction from
images and supervised deep learning
− PCB defect detection
− Visual inspection in real time
Silva et al. [27]
− Transfer learning and VGG16/resnet-50
pre-trained models
− Identifying the defective PCBs
− Non-referential inspection
4. INVESTIGATION CONCERNING PCBA AND TESTING
The queries “PCBA and testing” and “PCBA and testing” return respectively 51 and 43 documents,
which is significantly less than the previous request. Scientific production is characterized with irregular
curves with a big distance between the minimum and maximum values (Figures 7(a) and (b)). The published
and indexed in Scopus papers for 2022 year for the both queries are respectively 10 and 7. The maximal
values are 14 produced and indexed in Scopus documents in 2019 year for the first query and 11 documents
in 2018 year for the second query.
Bulletin of Electr Eng & Inf ISSN: 2302-9285
Printed circuit board and printed circuit board assembly methods for testing and … (Nikolay Petkov)
2573
(a)
(b)
Figure 7. Scientific production for the query: (a) “printed circuit board assembly and testing” and (b) “PCBA
and testing”
Among the most used keywords by authors are: printed circuit board, reliability, contactless testing,
humidity and accessibility (Figures 8(a) and (b)). The terms augmented reality (AR), defect detection and
automated optical inspection are also part of the keywords palette. Trend topic for the both queries is pointed
out to the keyword reliability.
(a)
(b)
Figure 8. The most frequent keywords regarding the query: (a) “printed circuit board assembly and testing”
and (b) “PCBA and testing”
Co-occurence networks are presented on Figures 9(a) and (b) as they are very similar for the both
queries. Two bigger clusters are observed respectively around the term contactless testing and reliability. The
cluster with the core term contactless testing also includes the terms: accessibility, testability, in-circuit test,
principal component analysis, magnetic sensors, high density PCBA testing, design for testability, PCB
assembly production test. A smaller cluster is formed around the keywords PCBA and smart manufacturing.
Another small cluster connects terms PCBA testing, thermal signatures and defect detection. A cluster that
connects the terms optical see-through and AR is also part of the networks.
If we summarize it can be said that the scientific output is significantly less for queries “PCBA and
testing” and “PCBA and testing” compared to research related to PCB testing. This, however, opens up a
wide field for future research and possibilities for proposing new testing approaches. Among the most
frequently used keywords by the authors and related to PCBA testing are contactless testing, defect detection
and automated optical inspection, which suggests that the research in the relevant articles is focused on these
ISSN: 2302-9285
Bulletin of Electr Eng & Inf, Vol. 13, No. 4, August 2024: 2566-2585
2574
themes. AR is also seen as an applicable technology in PCBA testing. The keyword reliability is indicated as
a trend, which points out the main goal of the conducted research in the field of testing. The keywords around
which the two largest clusters are formed in the co-occurrence networks are contactless testing and
realiability, which again confirms the extremely increased importance of these terms.
(a)
(b)
Figure 9. Co-occurrence networks regarding the query: (a) “printed circuit board assembly and testing” and
(b) “PCBA and testing”
The above analysis of the full-text articles reveals in more detail the specifics of contemporary
methods for PCBAs testing. It is done mainly on articles with open access.
Alaoui et al. [28] introduce usage of the technique infrared thermal signatures for PCBA testing
regarding defects detection. This approach is capable of performing contactless diagnoses of faulty capacitors
when the physical access to PCBA is limited. Other defects that could be identified are: component presence
and polarity, shorts and opens. The components mounted on PCBs as well as the PCBAs are classified in
three groups: reliable, less reliable and faulty. A drawback of the method is mentioning the need for more
time for testing.
Qiu et al. [29] are developed a capacitive resonator sensor for defect detection (shorts and opens on
metallic traces) on PCBAs. The structure and functionality of the sensor is explained, showing its potential
for usage in manufacturing. The future work will address preparation of a sensing array with this sensor with
the goal the measurement speed to be increased.
Alaoui et al. [30] are developed a technique for contactless diagnostics of faulty components on
PCBAs applying electromagnetic signatures. Electromagnetic field probes are used for field distribution
detection over PCBA components and magnetoresistance sensor for identifying changes in these components
that work with low frequency. Electromagnetic signatures are taken from PCBA with correct components
and are compared with PCBAs with wrong ones (with other values or removed components). According to
the amplitude of a given harmonic could be identified changes in components values.
In work of Liu et al. [31] is talking about the application of time domain reflection (TDR)
technology for failure location on PCB and PCBA. TDR is a non-destructive method and is used in different
areas for failure analysis as the authors prove its suitability in the electronic industry.
Tsenev et al. [32] conduct functional test to measure deformation of PCBA as testing standards of
mounting surface are applied. Practical experimentation is performed through an intelligent measurement
system and some results are presented.
Wanchun et al. [33] discuss usage of resistance strain test in a PCBA process and whether strain can
introduce damage of PCBA. The work shows effective identification of PCBA strain distribution and
measures for strain reduction.
Runji and Lin [34] argue that AR technology can support the PCBA inspection process and present
the developed markerless AR-based PCBA inspection system with features for defects identification and for
safe work of inspectors.
Bulletin of Electr Eng & Inf ISSN: 2302-9285
Printed circuit board and printed circuit board assembly methods for testing and … (Nikolay Petkov)
2575
Bonaria et al. [35] propose an approach for optimizing the workability of in-circuit testers through
re-arranging the probes movement on PCBA and reducing testing time. It proved the effectiveness of the
presented solution.
Le et al. [36] use image processing technique for identifying missing components on PCBAs and for
inspection of solder joints. The developed AOI system is characterized with high accuracy and speed.
Findings extracted from some of the most relevant papers regarding the PCBAs testing, found in Scopus for
the term 2018-2023 year, are summarized in Table 2.
Table 2. Methods for PCBA testing
Paper
Method/methodology
Aim
Alaoui et al. [28]
− Infrared thermal signatures
− Contactless diagnose of faulty capacitors
− Defects detection at PCBA testing: component presence
and polarity, shorts, and opens
Qiu et al. [29]
− Developed capacitive resonator sensor
− PCBAs defects detection (shorts and opens on metallic
traces)
Alaoui et al. [30]
− Contactless diagnostics of faulty components on
PCBAs applying electromagnetic signatures
− Identification of changes in components values
Liu et al. [31]
− Non-destructive
− Time domain reflection technology
− Failure location on PCB and PCBA
− Failure analysis
Tsenev et al. [32]
− Functional test
− Measure deformation of PCBA
Wanchun et al. [33]
− Resistance strain test
− Identification of PCBA strain distribution and measures
for strain reduction
Runji and Lin [34]
− AR technology
− PCBA inspection
− Defects identification and for safe work of inspectors
Bonaria et al. [35]
− Re-arrangement the probes movement on PCBA
− Reducing testing time
− Optimizing the workability of in-circuit testers
Le et al. [36]
− Image processing technique
− Identifying missing components on PCBAs
− Inspection of solder joints
5. RESEARCH RELATED TO DEFECTS DETECTION ON PCB AND PCBA
The returned result regarding the query “PCB defect detection” comprises 92 documents and
4 documents for the query “PCBA defect detection” for the investigated period 2018-2023 year as the
graphics of scientific production is presented on Figures 10(a) and (b). It is characterized with an increasing
curve, which is an indicator of the increased interest in the subject.
(a)
(b)
Figure 10. Annual scientific production concerning the query: (a) “PCB defect detection” and (b) “PCBA
defect detection”
The research articles are published mainly in Communications in Computer and Information
Science, IEEE Access, Proceedings of SPIE-The International Society for Optical Engineering, ACM
International Conference Proceeding Series, IEEE Transactions on Instrumentation and Measurement,
Journal of Physics: Conference Series, Sensors, Yi Qi Yi Biao Xue Bao/Chinese Journal of Scientific
Instrument as it is shown on Figures 11(a) and (b). It should be noted that the articles аre published not only
in journals and scientific collections in the field of electronics, but also in the area of computer and
information technologies. This suggests that new computer and information technologies are also involved in
the PCBs and PCBAs defect detection.
ISSN: 2302-9285
Bulletin of Electr Eng & Inf, Vol. 13, No. 4, August 2024: 2566-2585
2576
(a)
(b)
Figure 11. The most relevant sources taking into account the query: (a) “PCB defect detection” and (b)
“PCBA defect detection”
The most used keywords by authors for better description of the paper’s content are related to defect
detection and PCB and methods used for this detection: deep learning, attention mechanism, object detection,
CNN, and you only look once (YOLOv5) (Figures 12(a) and (b)).
Trend topics extracted from the author’s keywords considering the query “PCB defect detection” for
2023 years are: PCBs and deep learning (Figure 13). For the query “PCBA defect detection” trend topics
cannot be extracted, because of the extremily small number of returned documents.
(a)
(b)
Figure 12. The most frequent author keywords for the query: (a) “PCB defect detection” and (b) “PCBA
defect detection”
Figure 13. Trend topics for the query: “PCB defect detection”
Bulletin of Electr Eng & Inf ISSN: 2302-9285
Printed circuit board and printed circuit board assembly methods for testing and … (Nikolay Petkov)
2577
Figures 14(a) and (b) visualizes co-occurrence networks for the results from the both passed queries.
Documents obtained after the query “PCB defect detection” are characterized with a co-occurrence network
with two clusters, which are closely connected. The bigger cluster is formed around the terms defect
detection and deep learning. Other connected terms in this cluster are: printed circuit board, object detection,
CNN, image processing, machine vision, and residual network. The second cluster links author’s keywords
that are often used together: PCB defect detection, attention mechanism, clustering algorithm, yolo, swin
transformer, and component. The co-occurrence network for the query “PCBA defect detection” connects
two terms deep learning and object detection. From the formed clusters and used terms entire them can
conclude that the role of deep learning plays important role at PCB and PCBA defect detection.
(a)
(b)
Figure 14. Co-occurrence network for the query: (a) “PCB defect detection” and (b) “PCBA defect detection”
It can be seen that more often the articles are devoted to research related to PCB defect detection
and very few of the studies are focused on PCBA defect detection. Articles are published not only in
conference proceedings and scientific journals in the field of electronics, but publications in those devoted to
information and computer technologies are already observed. Among the most frequently used keywords by
the authors are not only those in the field of electronics and defect detection topic, but also those from
informatics such as deep learning, attention mechanism, object detection, CNN, and YOLOV5, which shows
their application in the detection of defects. The trend is outlined through the keywords PCBs and deep
learning and reaffirms the growing importance of deep learning for defect detection topic. The same is found
by observing the co-occurrence networks, in which terms from defect detection topic in electronics and terms
from informatics like deep learning, computer vision, and image processing together participate. Detailed
examination of papers content regarding methods and models for PCB and PCBA defects detection is
proposed in Table 3. The analysis is performed mainly on open access articles.
Park et al. [37] present an analysis regarding deep learning models and training specifics for defect
detection on PCB. The characteristics of PCB images and the factors that influence on PCB image quality in
the industrial environment are discussed. Methods for PCB defect detection are classified in three groups
according to test data and predictions: i) test data are cropped PCB images and prediction is related to the
class of the cropped image, ii) test data are whole PCB images as the prediction is the image class, and
iii) direct defect detection-test data are whole PCB images and the predictions concern defects location, size,
and class.
Wang et al. [38] propose a PCB defect detection model with a few-shot learning that gives
possibilities for usage of small datasets and to achieve good performance. The model includes modules for
feature enhancement and multiscale fusion with the goal the model precision to be improved and small object
defects to be detected. Defects that could be recognized are: missing hole, mouse bite, open, short, spur, and
spurious copper.
ISSN: 2302-9285
Bulletin of Electr Eng & Inf, Vol. 13, No. 4, August 2024: 2566-2585
2578
Yang and Kang [39] present a method for PCBs defect detection, which is based on the improved
YOLOv7 network. The method uses the SwinV2_TDD module for PCB feature extraction and magnification
factor shuffle attention mechanism for improving the attention mechanism adaptability. The applied
activation function is Mish. The method is tested after experimentation of six PCB defects (open and short
circuit, spur and spurious copper, mouse bite, and missing hole) and high average precision is achieved.
Chen et al. [40] investigate the usage of deep learning algorithm in AOI process of PCBs and
especially for effective re-inspection of defects with possibility to classify them. The following defects are
recognizable: missed, flipped, shifted and sideward component, tombstoning, non-wetting, and insufficient
solder on images taken through AOI machines. Such approach is proved to be accurate, faster, and decreases
the rate of AOI machine misjudgment.
The work of Li et al. [41] is focused on how efficiently to be detected defects on PCBAs in an
automated visual inspection process as an improved variant of the YOLOv7 model is presented. The main
solved challenges are related to increase the defects detection ratio considering some environmental
conditions (luminance, color) and taking into account different types, sizes, and density of mounted
components.
Jeon et al. [42] combine analysis of thermal images and deep learning to present a contactless
method for PCBAs defect detection that supports visual inspection. Thus, in real time the defects could be
located and identified with high accuracy.
Deng et al. [43] propose a new method for PCB defect detection based on finding the contour
unconformity (anomaly) and energy transformation (edge-guided energy-based PCB defect detection
(EEDD)). Its advantages lead to possibility for tiny defects detection and flexible definition of defect criteria
taking into account the requirements of a given production.
Zakaria et al. [44] summarize and discuss some approaches for PCB defect detection such as based
on multi-frequency Moiré technique, machine vision, X-ray imaging, and probabilistic techniques, as well as
such as driven by machine learning and deep learning. The authors conclude that AOI successfully adopts
machine and deep learning and probabilistic approaches in the field of PCB defect detection. Machine
learning is also applicable for prediction purposes at defects detection on PCBs.
Table 3. Methods and models for PCB and PCBA defect detection
Paper
Method/model
Aim
Park et al. [37]
− Deep learning
− Defect detection on PCB images
Wang et al. [38]
− Model includes modules for feature enhancement
and multiscale fusion
− Model precision to be improved and small object
defects to be detected
− Defects identification (missing hole, mouse bite, open,
short, spur, and spurious copper)
Yang and Kang
[39]
− Improved YOLOv7
− PCBs defect detection (open and short circuit, spur and
spurious copper, mouse bite, and missing hole)
Chen et al. [40]
− Deep learning on images
− PCBs AOI
− Re-inspection of defects with possibility to classify
(missed, flipped, shifted and sideward component,
tombstoning, non-wetting, and insufficient solder)
Li et el. [41]
− Improved variant of YOLOv7 model
− Defects on PCBAs in an automated visual inspection
process
Jeon et al. [42]
− Contactless method
− Analysis of thermal images and deep learning
− PCBAs defects location and detection
− Visual inspection
Deng et al. [43]
− Edge-guided energy-based PCB defect detection
(EEDD)
− PCB defect detection (contour anomaly)
Zakaria et al.
[44]
− Discussion on variety of methods: multi-
frequency Moiré technique, machine vision, X-ray
imaging, probabilistic techniques, as well as such
as driven by machine learning and deep learning
− PCB defect detection
6. VISUAL INSPECTION AT PCBs AND PCBA
Two queries “PCB and visual inspection” and “PCBA and visual inspection” are submitted to
Scopus scientific database as the returned results for the first query are 51 documents and for the second
query the found documents are only 8. The annual scientific production, presented on Figures 15(a) and (b),
is characterized with irregular curves with a tendency for increasing the number of indexed in Scopus papers.
This is confirmed by comparing the number of indexed documents at the beginning and at the end of the
investigated period as for 2018 year the obtained documents respectively for the first and the second queries
are 8 and 2 and for 2022 year the returned documents are 10 and 4.
Bulletin of Electr Eng & Inf ISSN: 2302-9285
Printed circuit board and printed circuit board assembly methods for testing and … (Nikolay Petkov)
2579
(a)
(b)
Figure 15. Annual scientific production for the query: (a) “PCB and visual inspection” and (b) “PCBA and
visual inspection”
Papers are most often presented at conferences with topics in electronics and visual inspection,
signal, information and image processing, artificial intelligence in engineering, computer, and information
technology (Figures 16(a) and (b)). This confirms the strong penetration of new technologies and artificial
intelligence in the field of PCB and PCBA visual inspection.
(a)
(b)
Figure 16. Annual scientific production for the quiery: (a) “PCB and visual inspection” and (b) “PCBA and
visual inspection”
Figures 17(a) and (b) depicts the most utilized by author’s keywords. For the query “PCB and visual
inspection”, the authors most often describe the papers’ content with keywords like: deep learning, visual
inspection, PCB, automatic visual inspection, defect detection, bill of materials, computer vision, image
processing, and machine learning. The authors keywords for the query “PCBA and visual inspection”
concern attribute gauge repeatability and reproducibility, automation, c3, ccga, cleanliness, co-solvent,
contamination, corrosion failure, cross-section, and decision-making. Trend topics are presented on Figure 18
and includes four terms: printed circuit board, defect detection, deep learning, and visual inspection.
In the co-occurrence network is observed one main cluster, formed around the core terms: deep
learning, visual inspection, defect detection, and computer vision (Figure 19). Other clusters are smaller and
connect the terms image processing and quality control in one cluster, the terms automatic visual inspection,
bill of materials, PCB assurance, and hardware assurance in different cluster and the terms PCB and machine
learning are organized in the third cluster.
ISSN: 2302-9285
Bulletin of Electr Eng & Inf, Vol. 13, No. 4, August 2024: 2566-2585
2580
(a)
(b)
Figure 17. Author’s keywords for the query: (a) “PCB and visual inspection” and (b) “PCBA and visual
inspection”
Figure 18. Trend topics for the query: “PCB and
visual inspection”
Figure 19. Co-occurrence network for the query: “PCB
and visual inspection”
It can be said that the main research addresses visual inspection on PCB rather than visual
inspection on PCBA, which is confirmed with the number of published and indexed in Scopus articles.
Anyway, visual inspection is an important part of electronics manufacturing that guarantees high quality
production. In this way it will continue to evolve and to grasp new technologies as it is seen through the
name of sources that publish such papers. Considering the author’s keywords and trend topics it is obvious
the big role of deep learning for visual inspection. Other important terms are computer vision, image
processing, and machine learning. The co-occurrence network once again confirms the application of
contemporary technologies for the purposes of visual inspection. More detailed view is created after
analyzing the content of relevant papers, mainly with open access.
Cao [45] presents a real-time visual inspection system to facilitate the identification of missing
component on PCB. The solution is complex and consists of hardware and software part. The created
software framework is based on image processing technique, analysis of region of each component,
cross-correlation technique, and consideration of production rules. The author confirms the workability of the
system after experimentations and points out its advantages for manufacturing quality control.
Glue control process in PCB manufacturing through applying visual inspection is discussed by
Iglesias et al. [46]. In this way the human operator is notified when to change the glue tube. Machine learning
algorithms random forest, polynomial regression and neural network are used for solving a regression task at
estimation of the glue level. Robustness of the presented approach is proved even at adding noise on
collected images.
Zhang [47] proposes a method for automatic inspection of PCBs that comprises YOLO deep
learning algorithm, image processing technique and an algorithm for position correction. The recognition rate
Bulletin of Electr Eng & Inf ISSN: 2302-9285
Printed circuit board and printed circuit board assembly methods for testing and … (Nikolay Petkov)
2581
of three defect types: wrong component, missing component and multiple parts is very high. It is argued that
this method is effective, automating visual inspection, and decreasing manufacturing costs.
Sathiaseelan et al. [48] are developed the electronic component localization and detection network
(ECLAD-Net) for identification and classification of resistors and capacitors on a PCB and in this way to
detect malicious and reused components. The proposed novel method is compared with existing ones and the
results point out that the learning model has to possess more than one hidden layer in a neural network, the
number of layers in CNN has to be not more than in VGG-16 and it should be non-linear.
The work of Adibhatla et al. [49] is focused on anomaly detection (defects) with different sizes on
PCBs through unsupervised machine learning (deep learning), which advantages in comparison to supervised
machine learning algorithms, are usage of less data and time reduction for pre-processing. A student-teacher
feature framework is used for image classification and for the distribution learning of images without defects
as ResNet-18 for speeding inference is utilized. The effort of human operators for labeling images is
decreased. The method is characterized with high accuracy and efficiency.
According to Chiun and Ruhaiyem [50] through automated visual inspection of PCBAs
components, time and cognitive load of operators can be reduced and the process of quality control can be
improved. Experimentation with three deep learning algorithms: R-CNN, YOLOv3, and SSD FPN is
performed as the last one is seen as the best solution for objects (components from different types)
localization and detection on images. Deep learning models are created through ResNet-50 (for R-CNN and
SSD FPN) and Darknet-53 (for YOLO network).
Malin et al. [51] compare several techniques: near infrared wavelengths, X-ray and visible light for
image analysis of PCBAs and defects detection in a process of visual inspection. The authors conclude that
X-ray technique should be avoided, because of health risks. Instead, it is good to use more health-safe
approaches as this investigation identifies features of the compared techniques regarding detect defects at
given wavelength. A summarization of contemporary methods and models utilized in PCB and PCBA visual
inspection is presented in Table 4.
Table 4. Methods and models in PCB and PCBA visual inspection
Paper
Method/model
Aim
Cao [45]
− Complex solution with hardware and software part
− Software framework with image processing technique,
analysis of region of each component, cross-correlation
technique and consideration of production rules
− Identification of missing component on PCB
Iglesias et al. [46]
− Machine learning algorithms: random forest, polynomial
regression and neural network
− Regression task at estimation the glue level in
PCB manufacturing
Zhang [47]
− YOLO deep learning algorithm, image processing
technique, and an algorithm for position correction
− Recognition of three defect types wrong
component, missing component and multiple
parts on PCBs
Sathiaseelan et al.
[48]
− ECLAD-Net
− Identification and classification of resistors and
capacitors on a PCB
− Detect malicious and reused components
Adibhatla et al.
[49]
− Unsupervised machine learning method (deep learning)
− Student-teacher feature framework
− Resnet-18 for speeding inference
− Defects detection (anomalies with different
sizes) in PCBs
− Image classification
Chiun and
Ruhaiyem [50]
− Deep learning: R-CNN, YOLOv3, and SSD FPN
− Objects (components from different types)
localization and detection on images of PCBAs
Malin et al. [51]
− Near infrared wavelengths, X-ray and visible light
techniques comparison
− Image analysis of PCBAs and defects detection
7. RESULTS AND DISCUSSION
The findings from performed investigation of relevant documents, indexed in Scopus, regarding
eight different queries “PCB and testing” and “PCB and testing”, “PCBA and testing” and “PCBA and
testing”, “PCB defect detection” and “PCBA defect detection”, “PCB visual inspection” and “PCBA visual
inspection” are summarized through created framework, presented on Figure 20.
The framework outlines the current landscape of the most researched methods, methodologies, and
models for PCB and PCBA testing and visual inspection. It can be seen that alongside conventional methods that
are continuously being improved, new sensor solutions have recently been created to conduct non-destructive
and contactless tests. The role of machine, deep and transfer learning in automating testing and visual
inspection tasks is growing significantly in our contemporary ages.
The most investigated recently trending methods for PCB testing and visual inspection are classified in:
− Destructive methods that are less often investigated, perhaps due to the destructive nature and removing a
part of the prototype. Anyway, such methods are applied for identification of possible PCB failure.
ISSN: 2302-9285
Bulletin of Electr Eng & Inf, Vol. 13, No. 4, August 2024: 2566-2585
2582
− Non-destructive methods that include a wide variety of techniques for PCB testing like acoustic emission,
X-ray tomography, frequency/magnitude sensing, time domain reflection as often they are used in
different combinations to locate and evaluate PCB defects and damages.
− Virtual methods, which are based on simulations applying some analytical techniques such as finite
element analysis to identify risk problems and ro remove failure issues.
− Algorithms from deep learning and transfer learning for PCB and PCBA defects detection and real time
visual inspection are the most utilized contemporary approaches.
Figure 20. Framework, summarizing the most investigated methods for PCB and PCBA testing and visual
inspection
Methods for PCB testing and visual inspection
destructive
PCB bend angle
influence
non-destructive
possible failure
PCB design defects
finding crutical defects
defects
PCB usage evaluation
acoustic emission
X-ray tomography
Methods for PCBA testing and visual inspection
contactless
infrared thermal
signatures
PCBA defects detection
(component presence and
polarity, shorts and opens)
virtual
fluid dynamic
finite element
analysis
risk problems identification
removing failure issues
frequency/magnitude
sensing
defects detection on metal
surface
locate notch damages
electromagnetic
signatures
PCBA faulty components
(identification changes in
components values)
unsupervised deep
learning
machine, deep and
transfer learning
transfer learning
features extraction
supervised deep
learning
detecting PCB defects on
images (scratch, broken PCB
edge, hole)
PCB defects detection
real time visual inspection
time domain
reflection
PCB failure analysis
time domain
reflection
non-destructive
failure analysis
visual
augmented reality
defects identification
deep learning
deep learning
defects identification
PCBA visual inspection
virtual
physics of failure
risk problems identification
removing failure issues
Bulletin of Electr Eng & Inf ISSN: 2302-9285
Printed circuit board and printed circuit board assembly methods for testing and … (Nikolay Petkov)
2583
The most often researched methods for PCBA testing in the last five years could be classified in:
− Contactless methods for defects detection on PCBAs such as: infrared thermal signatures, electromagnetic
signatures.
− Non-destructive methods for failure analysis.
− Virtual methods, which are based on performed simulations for identification of risk problems in PCBAs
and for solving some failure issues.
− Visual methods based on AR technology, which is used in support of defects detection on PCBAs and in
assistance of quality control inspectors.
− Machine and deep learning algorithms as mainly are discussed deep learning algorithms for PCBAs
defects detection and in facilitation of visual inspection.
The created framework answers to the posed research questions: what are the most investigated
methods and techniques for PCB and PCBA testing and visual inspection in the last five years? and what is
the impact of machine and deep learning on methods for PCB and PCBA testing and visual inspection?
non-destructive and non-contact testing methods are seen to be preferred for examination and
implementation. Testing in a simulation environment is also a preferred method for investigating defects and
failures. Bibliometric analysis and article content analysis confirm the growing role of machine and deep
learning to detect defects and to support visual inspection.
8. CONCLUSION
The findings of this paper reveal the most often recently investigated topics related to PCB and
PCBA testing and visual inspection methods as these methods are classified in groups for better
understanding of their nature. The data are taken from scientific database Scopus for a term from 2018 to
2023 year. The most relevant papers mainly with open access are explored and on this basis a framework that
summarizes the contemporary landscape in the PCB and PCBA methods for testing and visual inspection is
developed.
Bibliometric analysis is also applied to understand: i) the interest to the researched topics, which has
been steady for the past 5 years; ii) the sources that published these papers, which are recently not only in the
domain in electronics, but also in the field of computer and information technologies; iii) the most frequent
author’s keywords that describe in the best way the papers content as in this case they are related to
contemporary technologies, methods, and models for PCB and PCBA testing and visual inspection; and iv)
trending topics, which point out the growing impact of machine and deep learning algorithms for improving
the reliability of printed circuit board assembly and PCBA testing and visual inspection.
The limitation of this study concerns the following issues: i) the used literature sources and their
bibliometrics are taken from Scopus scientific database and ii) mainly are examined papers with open access.
Anyway, we believe that these limitations in no way diminish the importance of the study, as Scopus is a
large scientific database indexing high-quality articles. Open access papers are published in scientific
collections or journals with an impact rank/impact factor, which also does not decrease their significance.
The future work is focused on further exploration of the application of machine learning and deep learning in
the processes of PCB and PCBA testing and visual inpection.
ACKNOWLEDGEMENT
This research is supported by Bulgarian National Science Fund in the scope of the project “Exploration
the application of statistics and machine learning in electronics” under contract number КП-06-Н42/1.
REFERENCES
[1] Á. M. Sampaio et al., “Design and development of an automatic optical inspection (AOI) system support based on digital
manufacturing,” Procedia CIRP, vol. 119, pp. 15–20, 2023, doi: 10.1016/j.procir.2023.03.080.
[2] A. A. Dzyubanenko and G. I. Korshunov, “Quality control in cyber-physical systems of smart electronics manufacturing,”
Journal of Physics: Conference Series, vol. 2094, no. 4, p. 042066, Nov. 2021, doi: 10.1088/1742-6596/2094/4/042066.
[3] T. Parakontan and W. Sawangsri, “Development of the machine vision system for automated inspection of printed circuit board
assembl,” in Proceedings of 2019 3rd IEEE International Conference on Robotics and Automation Sciences, IEEE, Jun. 2019, pp.
244–248, doi: 10.1109/ICRAS.2019.8808980.
[4] R. Gupta, “Failure mode and effects analysis of PCB for quality control process,” Mapan-Journal of Metrology Society of India,
vol. 38, no. 2, pp. 547–556, Jun. 2023, doi: 10.1007/s12647-022-00619-5.
[5] J. Li, “Application of statistical process control in engineering quality management,” IOP Conference Series: Earth and
Environmental Science, vol. 831, no. 1, p. 012073, Aug. 2021, doi: 10.1088/1755-1315/831/1/012073.
[6] J. He, Y. Cen, S. Alelaumi, and D. Won, “An artificial intelligence-based pick-and-place process control for quality enhancement
in surface mount technology,” IEEE Transactions on Components, Packaging, and Manufacturing Technology, vol. 12, no. 10,
pp. 1702–1711, Oct. 2022, doi: 10.1109/TCPMT.2022.3215109.
ISSN: 2302-9285
Bulletin of Electr Eng & Inf, Vol. 13, No. 4, August 2024: 2566-2585
2584
[7] M. Mirzaei, M. H. Sadat, and F. Naderkhani, “Application of machine learning for anomaly detection in printed circuit boards
imbalance date set,” in 2023 IEEE International Conference on Prognostics and Health Management, IEEE, Jun. 2023, pp. 128–
133, doi: 10.1109/ICPHM57936.2023.10193957.
[8] A. Bhattacharya and S. G. Cloutier, “End-to-end deep learning framework for printed circuit board manufacturing defect
classification,” Scientific Reports, vol. 12, no. 1, p. 12559, Jul. 2022, doi: 10.1038/s41598-022-16302-3.
[9] D. Kim, J. Koo, H. Kim, S. Kang, S. H. Lee, and J. T. Kang, “Rapid fault cause identification in surface mount technology
processes based on factory-wide data analysis,” International Journal of Distributed Sensor Networks, vol. 15, no. 2, p.
155014771983280, Feb. 2019, doi: 10.1177/1550147719832802.
[10] I. Parviziomran, S. Cao, H. Yang, S. Park, and D. Won, “Data-driven prediction model of components shift during reflow process
in surface mount technology,” Procedia Manufacturing, vol. 38, pp. 100–107, 2019, doi: 10.1016/j.promfg.2020.01.014.
[11] C. Houdek, Caltronics Design and Assembly, Inc. [Available Online]: http://www.claire-e-
cunningham.com/uploads/8/0/4/5/8045165/inspection-and-testing-methods-for-pcbs-an-overview.pdf.
[12] P. Chen and F. Xie, “A Machine learning approach for automated detection of critical PCB flaws in optical sensing systems,”
Photonics, vol. 10, no. 9, p. 984, Aug. 2023, doi: 10.3390/photonics10090984.
[13] B. Bártová and V. Bína, “A novel data mining approach for defect detection in the printed circuit board manufacturing process,”
Engineering Management in Production and Services, vol. 14, no. 2, pp. 13–25, Jun. 2022, doi: 10.2478/emj-2022-0013.
[14] M. Kumar and M. Kumar, “A survey on various approaches of automatic optical inspection for PCB defect detection,” International
Journal of Computer Sciences and Engineering, vol. 7, no. 6, pp. 837–841, Jun. 2019, doi: 10.26438/ijcse/v7i6.837841.
[15] M. Jafir et al., “The global trend of nanomaterial usage to control the important agricultural arthropod pests: A comprehensive
review,” Plant Stress, vol. 10, p. 100208, Dec. 2023, doi: 10.1016/j.stress.2023.100208.
[16] A. Dirpan et al., “Trends over the last 25 years and future research into smart packaging for food: A review,” Future Foods, vol.
8, p. 100252, Dec. 2023, doi: 10.1016/j.fufo.2023.100252.
[17] M. J. Akhtar, M. Azhar, N. A. Khan, and M. N. Rahman, “Conceptualizing social media analytics in digital economy: An
evidence from bibliometric analysis,” Journal of Digital Economy, vol. 2, pp. 1–15, Dec. 2023, doi: 10.1016/j.jdec.2023.03.004.
[18] D. Lee, “Bibliometric analysis of Asian ‘language and linguistics’ research: a case of 13 countries,” Humanities and Social
Sciences Communications, vol. 10, no. 1, p. 379, Jul. 2023, doi: 10.1057/s41599-023-01840-6.
[19] M. Aria and C. Cuccurullo, “bibliometrix: An R-tool for comprehensive science mapping analysis,” Journal of Informetrics, vol.
11, no. 4, pp. 959–975, Nov. 2017, doi: 10.1016/j.joi.2017.08.007.
[20] A. Azin, A. Zhukov, A. Narikovich, S. Ponomarev, S. Rikkonen, and V. Leitsin, “Nondestructive testing method for a new
generation of electronics,” MATEC Web of Conferences, vol. 143, p. 04007, Jan. 2018, doi: 10.1051/matecconf/201814304007.
[21] J. Elliot and J. Brown, “An Investigation into the failure characteristics of external PCB traces with different angle bends,” Journal of
Electronic Testing: Theory and Applications, vol. 39, no. 1, pp. 103–110, Feb. 2023, doi: 10.1007/s10836-023-06043-0.
[22] A. Wileman, S. Perinpanayagam, and S. Aslam, “Physics of failure (PoF) based lifetime prediction of power electronics at the
printed circuit board level,” Applied Sciences (Switzerland), vol. 11, no. 6, p. 2679, Mar. 2021, doi: 10.3390/app11062679.
[23] L. Li et al., “An interdigital electrode probe for detection, localization and evaluation of surface notch-type damage in metals,”
Sensors (Switzerland), vol. 18, no. 2, p. 371, Jan. 2018, doi: 10.3390/s18020371.
[24] R. Oliveira et al., “A systematic analysis of printed circuit boards bending during in-circuit tests,” Machines, vol. 10, no. 2, p.
135, Feb. 2022, doi: 10.3390/machines10020135.
[25] I. Volkau, M. Abdul, W. Dai, M. Erdt, and A. Sourin, “Detection defect in printed circuit boards using unsupervised feature
extraction upon transfer learning,” in Proceedings-2019 International Conference on Cyberworlds, IEEE, Oct. 2019, pp. 101–
108, doi: 10.1109/CW.2019.00025.
[26] V. T. Nguyen and H. A. Bui, “A real-time defect detection in printed circuit boards applying deep learning,” EUREKA, Physics,
and Engineering, vol. 2022, no. 2, pp. 143–153, Mar. 2022, doi: 10.21303/2461-4262.2022.002127.
[27] L. H. D. S. Silva, G. O. D. A. Azevedo, B. J. T. Fernandes, B. L. D. Bezerra, E. B. Lima, and S. C. Oliveira, “Automatic optical
inspection for defective PCB detection using transfer learning,” in 2019 IEEE Latin American Conference on Computational
Intelligence, IEEE, Nov. 2019, pp. 1–6, doi: 10.1109/LA-CCI47412.2019.9037036.
[28] N. E. B. Alaoui, P. Tounsi, A. Boyer, and A. Viard, “Detecting PCB assembly defects using infrared thermal signatures,” 26th
International Conference “Mixed Design of Integrated Circuits and Systems,” IEEE, Jun. 2019, pp. 345–349, doi:
10.23919/MIXDES.2019.8787089.
[29] T. Qiu, C. K. Andrew Tek, and S. Y. Huang, “A compact high-resolution resonance-based capacitive sensor for defects detection
on PCBAs,” IEEE Access, vol. 8, pp. 203758–203768, 2020, doi: 10.1109/ACCESS.2020.3036884.
[30] N. E. B. Alaoui, P. Tounsi, A. Boyer, and A. Viard, “New testing approach using near electromagnetic field probing intending to
upgrade in-circuit testing of high density PCBAs,” in 27th North Atlantic Test Workshop, IEEE, May 2018, pp. 1–8, doi:
10.1109/NATW.2018.8388867.
[31] J. Liu et al., “A Novel Non-destructive Failure Location method for electronic components based on TDR technology,” in 2022 23rd
International Conference on Electronic Packaging Technology, IEEE, Aug. 2022, pp. 1–3, doi: 10.1109/ICEPT56209.2022.9873325.
[32] V. Tsenev, V. Videkov, and N. Spasova, “Measurement of PCBA (Printed circuit board assembly) deformation during functional
testing of electronic modules,” in 2021 35th International Conference on Information Technologies, IEEE, Sep. 2021, pp. 1–5,
doi: 10.1109/InfoTech52438.2021.9548409.
[33] T. Wanchun, X. Hui, and L. Chaohui, “Application of strain test technology in PCBA process,” in 2020 21st International
Conference on Electronic Packaging Technology, IEEE, Aug. 2020, pp. 1–5, doi: 10.1109/ICEPT50128.2020.9202988.
[34] J. M. Runji and C. Y. Lin, “Markerless cooperative augmented reality-based smart manufacturing double-check system: Case of
safe PCBA inspection following automatic optical inspection,” Robotics and Computer-Integrated Manufacturing, vol. 64, p.
101957, Aug. 2020, doi: 10.1016/j.rcim.2020.101957.
[35] L. Bonaria, M. Raganato, G. Squillero, and M. S. Reorda, “Test-plan optimization for flying-probes in-circuit testers,” in
Proceedings-2019 IEEE International Test Conference in Asia, IEEE, Sep. 2019, pp. 19–24, doi: 10.1109/ITC-Asia.2019.00017.
[36] H. N. Le, T. V. Nguyen, and N. C. Debnath, “A machine vision based automatic optical inspection system for detecting defects of
PCBA,” in Advances in Intelligent Systems and Computing, 2020, pp. 480–489, doi: 10.1007/978-3-030-44289-7_45.
[37] J. H. Park, Y. S. Kim, H. Seo, and Y. J. Cho, “Analysis of training deep learning models for PCB defect detection,” Sensors, vol.
23, no. 5, p. 2766, Mar. 2023, doi: 10.3390/s23052766.
[38] H. Wang, J. Xie, X. Xu, and Z. Zheng, “Few-shot PCB surface defect detection based on feature enhancement and multi-scale
fusion,” IEEE Access, vol. 10, pp. 129911–129924, 2022, doi: 10.1109/ACCESS.2022.3228392.
[39] Y. Yang and H. Kang, “An enhanced detection method of PCB defect based on improved YOLOv7,” Electronics (Switzerland),
vol. 12, no. 9, p. 2120, May 2023, doi: 10.3390/electronics12092120.
Bulletin of Electr Eng & Inf ISSN: 2302-9285
Printed circuit board and printed circuit board assembly methods for testing and … (Nikolay Petkov)
2585
[40] I. C. Chen, R. C. Hwang, and H. C. Huang, “PCB defect detection based on deep learning algorithm,” Processes, vol. 11, no. 3, p.
775, Mar. 2023, doi: 10.3390/pr11030775.
[41] Z. Li, J. Yan, J. Zhou, X. Fan, and J. Tang, “An efficient SMD-PCBA detection based on YOLOv7 network model,” Engineering
Applications of Artificial Intelligence, vol. 124, p. 106492, Sep. 2023, doi: 10.1016/j.engappai.2023.106492.
[42] M. Jeon, S. Yoo, and S. W. Kim, “A contactless PCBA defect detection method: convolutional neural networks with
thermographic images,” IEEE Transactions on Components, Packaging, and Manufacturing Technology, vol. 12, no. 3, pp. 489–
501, Mar. 2022, doi: 10.1109/TCPMT.2022.3147319.
[43] S. Deng et al., “EEDD: edge-guided energy-based PCB defect detection,” Electronics (Switzerland), vol. 12, no. 10, p. 2306, May
2023, doi: 10.3390/electronics12102306.
[44] S. S. Zakaria, A. Amir, N. Yaakob, and S. Nazemi, “Automated detection of printed circuit boards (PCB) defects by using
machine learning in electronic manufacturing: current approaches,” IOP Conference Series: Materials Science and Engineering,
vol. 767, no. 1, p. 012064, Feb. 2020, doi: 10.1088/1757-899X/767/1/012064.
[45] X. Cao, “A real-time automated visual inspection system for printed circuit boards missing footprints detection,” International Journal of
Advanced Computer Science and Applications, vol. 14, no. 5, pp. 350–358, 2023, doi: 10.14569/IJACSA.2023.0140537.
[46] B. P. Iglesias, M. Otani, and F. G. Oliveira, “Glue level estimation through automatic visual inspection in PCB manufacturing,” in
Proceedings of the 18th International Conference on Informatics in Control, Automation and Robotics, SCITEPRESS - Science
and Technology Publications, 2021, pp. 731–738, doi: 10.5220/0010540807310738.
[47] K. Zhang, “Using deep learning to automatic inspection system of printed circuit board in manufacturing industry under the internet of
things,” Computer Science and Information Systems, vol. 20, no. 2, pp. 723–741, 2023, doi: 10.2298/CSIS220718020Z.
[48] M. A. M. Sathiaseelan, O. P. Paradis, S. Taheri, and N. Asadizanjani, “Why is deep learning challenging for printed circuit board
(PCB) component recognition and how can we address it?,” Cryptography, vol. 5, no. 1, p. 9, Mar. 2021, doi:
10.3390/cryptography5010009.
[49] V. A. Adibhatla et al., “Unsupervised anomaly detection in printed circuit boards through student–teacher feature pyramid
matching,” Electronics (Switzerland), vol. 10, no. 24, p. 3177, Dec. 2021, doi: 10.3390/electronics10243177.
[50] O. Y. Chiun and N. I. R. Ruhaiyem, “Object detection based automated optical inspection of printed circuit board assembly using deep
learning,” in Communications in Computer and Information Science, 2023, pp. 246–258, doi: 10.1007/978-981-99-0405-1_18.
[51] B. Malin, T. Kalganova, J. Danskins, and J. R. Gilchrist, “PCBA image analysis: a comparison of visible, infrared and x-ray
wavelengths,” in 2022 IEEE Physical Assurance and Inspection of Electronics, IEEE, Oct. 2022, pp. 1–7, doi:
10.1109/PAINE56030.2022.10014963.
BIOGRAPHIES OF AUTHORS
Nikolay Petkov possesses M.Sc. degrees in Automation, Information, and Control
Engineering from Technical University of Sofia and in Information Technology in Media Business
from the University of Library Science and Information Technology, Sofia, Bulgaria. He is an
employee in automated maintenance in the unit of Sensata Technologies Bulgaria EOOD and since
this academic year is a part-time lecturer at the Technical University of Sofia, Technical College–
Sofia. His research interests are related to electronics testing process and application of machine
learning in electronics. He is the author of several scientific publications. He can be contacted at
email: ntoshevp@abv.bg.
Malinka Ivanova is an Associate Professor in Computer Sciences and Informatics at
the Faculty of Applied Mathematics and Informatics, Department of Informatics of the Technical
University of Sofia. She holds a doctorate in Informatics and in Automation of Engineering Work
and Systems for Automated Design–both from Technical University of Sofia, Bulgaria. She
specialized in Technical University of Bratislava, Academy of Sciences of the Slovak Republic,
Johann Wolfgang Goethe University, Germany, Tokyo Institute of Technology, Japan and
participated in training in Slovenia, France, Austria and Spain. Her research interests are focused
on: exploration applications of machine learning in electronics, online security and data protection,
predictive and analytical modelling, automation of engineering work, and technology-enriched
learning. She is the author of more than 100 publications. She is a member of a number of program
committees of international and national conferences and a reviewer of internationally recognized
journals. She can be contacted at email: m_ivanova@tu-sofia.bg.
