The Printed Circuit Assembler’s Guide to...

™
SMT Inspection: Today, Tomorrow, and Beyond

Brent A. Fischthal
KOH YOUNG AMERICA

© 2021 BR Publishing, Inc.

All rights reserved.

BR Publishing, Inc.
dba: I-Connect007
942 Windemere Dr. NW,
Salem, OR 97304
U.S.A.

ISBN: 978-1-7342005-8-4
Visit I-007eBooks.com for other titles in this series.

I-Connect007.com
 Peer Reviewers
Christopher St. Mars
Process Engineer, Intervala

Chris St. Mars has more than 12 years of experi-

ence in the electronics manufacturing industry.
He currently serves as a process engineer for
Intervala, a full-service electronics manufac-
turing service provider in the Greater Pittsburgh
area. Chris is an SMTA Certified SMT Process
Engineer (CSMTPE), has completed Six Sigma Green Belt training and was
previously a Certified IPC Trainer of IPC-A-610. Chris earned his Bachelor of
Science degree in Electrical Engineering Technology from California Univer-
sity of Pennsylvania.

Michael Ford
Senior Director of Emerging Industry Strategy,
Aegis Software

Working for Aegis Software provides Michael

the opportunity to apply his experience
with software for electronics assembly
manufacturing to further drive technology
solution innovation, satisfying evolving business
needs in modern digital manufacturing.

Starting his career with Sony, including eight years working in Japan,
Michael has been instrumental in creating and evolving revolutionary
software solutions for assembly manufacturing that meet the most
demanding expectations.

Today, Michael is an established thought leader for Industry 4.0 and

digital smart factories and an active contributor to industry standards.
In 2020, Michael was given an IPC President’s Award as recognition for
contributions including the CFX, traceability, secure supply chain and
Digital Twin standards. Michael also regularly contributes articles, columns
and blogs in several leading industry publications.
 About the Author
Brent A. Fischthal
Senior Manager for Americas Marketing and
Regional Sales, Koh Young America

With a progressive career in marketing

and sales leadership roles within the
global electronics assembly industry,
Brent Fischthal brings industry-focused
experience in marketing strategy, sales
management, business intelligence, and
product management from positions
held previously at Panasonic Factory Solutions, Philips Assembléon, Quad
Systems, and Universal Instruments.

Brent is currently the Senior Manager for Americas Marketing and Regional
Sales with Koh Young America. Brent earned a Bachelor of Science degree
in marketing from West Virginia University and a Master of Business
Administration from the University of Florida Hough Graduate School of
Business.

In support of the electronics manufacturing industry advancement, Brent

serves as the president of the SMTA Carolinas Chapter and is a member of
the SMTA Marketing, IPC-CFX, IPC-DPMX, and IPC trade show committees.
Additionally, he serves on the MESA International Global Board of Direc-
tors and is the published author of several technical articles and white
papers focused on assembly and inspection solutions.
CONTENTS
2 Foreword
CHAPTER 1
5 It All Starts With Vision
CHAPTER 2
11 Performance Anxiety
CHAPTER 3
15 The Data Dilemma
CHAPTER 4
23 A Vision of the Factory of the Future
CHAPTER 5
29 Advanced Process Control and Inspection
CHAPTER 6
39 How Far Can AI Take Us?
CHAPTER 7
47 Smarter Manufacturing Enabled by Inspection

51 Acknowledgments

53 About Koh Young Technology

 Foreword

Currently, surface mount

technology (SMT) lines
are monitored by two key
inspection processes. The
first is solder paste inspec-
tion (SPI), which reviews
the printed circuit board
after solder paste has been
applied to the pads and
before components are placed. The second is automatic optical
inspection (AOI), which reviews the printed circuit board assembly
(PCBA) after the board has been assembled with components and
reflowed (or soldered). Now technologists are also using AOI prior to
reflow to ensure parts are placed correctly before they are soldered
into place.

The two processes are complementary in that the first, SPI, ensures
expensive components are not placed on a PCB which does not have
the right amount of solder paste in the correct locations. The second
process, AOI, ensures that those boards that are assembled have
good solder joints and all the components are placed and orientated
correctly.

Beyond their job of identifying errors and stopping flawed boards

from continuing along the line where value is added to bad boards,
or worse still, bad boards are shipped, these inspection processes
provide the data needed to improve the performance of the entire
line. SPI and AOI can both be used to make live line adjustments to
improve quality without stopping the line, and to provide data that
helps identify the root causes of failures or variances.

SPI and AOI use a combination of optics and software to assess

the quality of the processes they inspect, and the quality of these

2
two elements along with speed and ease of operation are the real
measure of success in the SMT inspection world. Fundamentally, a
good inspection solution generates the best possible data in the form
of images and measurements and uses that data to identify good
or bad boards, as well as information related to process control to
improve the performance of the overall line.

An additional dynamic impacting the SMT line—and indeed, the

whole manufacturing world—is the use of data to manage processes
and target the drive toward the smart factory, often referred to as
Industry 4.0. This trend has driven manufacturers to connect their
equipment and processes so that data, such as the inspection data
from SPI and AOI, can be used holistically. Tools such as artificial intel-
ligence (AI) are being deployed so that the large amounts of data can
be used to drive better, faster decision-making that augments the
operator, protects them from the huge complexity of the big data that
has been generated, and even offers a path to autonomous manu-
facturing where the line is able to self-manage and improve its own
performance.

For companies to succeed in the development of Industry 4.0, and

perhaps even beyond as we drive toward manufacturing autonomy,
they will need inspection solutions, and inspection partners, that can
combine domain expertise in optical inspection (vision and software)
with the ability to connect to larger systems and contribute to process
improvement utilizing tools like artificial intelligence.

3
Figure 1.1: Newspaper Rock State Historic Monument has hundreds of petroglyphs
believed to be from about 2000 years ago. (Image Source: AdobeStock)
 Chapter 1

It All Starts With Vision

Today, optical inspection systems are the preferred solution for in-line quality
control in the SMT industry. Systems such as solder paste inspection (SPI)
or automated optical inspection (AOI) systems for pre- and post-reflow are
almost standard in every production facility.

In the early days of optical inspection, SPI and AOI systems were based on 2D
inspection technology. These 2D systems were looking at different grey levels
for solder paste, solder joint, and component detection. Most decisions were
made by a “good/bad” comparison to reference images, or “golden boards.”
But for the solder paste inspection, the most important process indicator for
good solder joint quality was missing—namely, paste volume. Consequently,
2D technology for SPI became redundant when 3D measurement solutions
were introduced. Years later, the same situation began to occur for AOI.

While many 2D AOI systems still remain in use, the effort needed to keep this
technology at a low level of escape and false-call rates is tremendously high.
This is because the concept of comparing reference images is still the main
technique applied in 2D systems. To help, some equipment manufacturers
have added additional cameras and projectors to create quasi 3D or 2.5D
technology, but it is still based on the same antiquated, inadequate concept of
comparison or color assessment.

Now, after the introduction of 3D AOI, a transition to true 3D measurement is

taking place. The benefits are clear: rock-solid threshold and inspection toler-
ances provide a low escape and false-call rate with minimum debugging effort.

We see the world in 3D, and it is with that data that we are able to see what
is good and what is not, what is fit for purpose and what should be inspected
out. The same is true of inspection equipment. Only exceptional 3D vision can
provide the image data and measurements needed to make the right decision
about the fitness of a PCB to continue along the line while providing the data
needed to make holistic process improvements.

5
With increasing board complexity, more components, more joints, higher
density, and new package technologies such as 01005 (0402M) and even
008004 (0201M) microchips, 2D AOI technology using grey-scale image anal-
ysis or angled camera views of color images are no longer a viable option. Most
decisions made are based on “good/bad” comparison of reference images or
color assessment, which can easily be affected by variables such as lighting
consistency, component surface finish, board condition, and component prox-
imity.

The benefits of 3D inspection are clear:

ƒ Total faith in inspection tolerances
ƒ Reduced effort to debug inspection programs
ƒ Measurement data that provides meaningful insights about
the process and helps identify and eliminate the root causes
of defects

3D SPI, together with 3D AOI, enables manufacturers to accurately control and

monitor the solder printing, the component placement, and the solder reflow
processes.

The requirements of 3D AOI to go from inspection to measurement, and ulti-

mately to process control and optimization, is measurement data that is reli-
able, repeatable, and relatable.

Full 3D Inspection
Some AOI solutions offer a limited level of 3D coverage for components and
solder joints. Full 3D coverage is needed to identify boards with defects and
help manufacturers monitor their performance. These 3D AOI systems not

Figure 1.1: Full 3D inspection coverage examples.

6
only inspect common defects, such as missing or wrong components, but
accurately identify coplanarity and lead-bridging issues, which are prevalent
in small packages and BGAs.

Some 3D AOI solutions use multi-frequency shadow moiré technology to

measure the Z-axis profilometry of the entire board. By measuring compo-
nents and solder joints, and then offering critical height information to the
inspection algorithms, manufacturers can identify errors such as pad overhang
and insufficient solder. Unlike 3D AOI systems using a brightness threshold,
the moiré technology uses the height threshold to extract critical body and the
lead tip information. This data allows the AOI to provide reliable body and joint
measurements with CAD dimensions, since it can locate the body properly in
the very first stage of the inspection process.

Shadow Problems
Will accurate body-finding guarantee accurate 3D measurement data? Not
necessarily. There are several inspection challenges even with high quality 3D
inspection. One major challenge is the shadow effect. When various compo-
nents are assembled on a board, some short parts, such as resistors, may be
shadowed by tall neighboring components such as connectors, which leads to
inaccurate 3D measurement results. By using 8-way projection, machines can
eliminate blind spots and can construct a true 3D image using the multiple
projections available.

Figure 1.2: An 8-way projection system overcomes the shadow problem.

7
Inter-reflection Problem
Another critical issue is inter-reflection. The soldered and tinned compo-
nents on assembled boards may have specular surfaces, which will reflect
some lighting back to the camera, while creating strong inter-reflection with
other lighting reflections. Since some of the reflected light does not reach
the camera, the reflections
can generate false signals
that cause height measure-
ment value corruption. This
specular reflection issue is
becoming more trouble-
some as board density is
increasing and compo-
nent spacing is decreasing.
Figure 1.3: Example of how a single projection system Certain AOI systems over-
faces problems with specular areas. come this with superior
multi-view sensors and eliminate measurement inaccuracies with advanced
3D image fusion algorithms. If done properly, the AOI meticulously identifies
and filters the multiple reflections caused by shiny components and reflective
solder joints with multi-projections and intelligent algorithms.

Solder Joints
Measuring a solder joint is much more complicated than assessing a typical
component body. This is because the solder deposit varies depending on
board design, printing conditions, and they may have specular reflections.
Such differences can affect the optical inspection of defective solder joints. To
help, an algorithm can be applied to the data to generate robust and stable
measurement data for resistors, capacitors, and other component leads,
regardless of solder joint quantity or shape.

Figure 1.4: Example of a 3D solder joint measurement.

8
Moreover, the 3D AOI helps to verify solder joint quality with measurement
conditions adhering to the various IPC-A-610 standards. IPC-A-610 standards
are designed to satisfy quality standards. Consider that for Class 3, the highest
standard of the IPC classes, the standard for rectangular chip components
calls for a minimum fillet height that is less than the solder thickness, plus 25%
component thickness, or the solder thickness plus 0.5 mm [0.02 in], which-
ever is less. This quality standard translates into a Z-dimension suited for a 3D
algorithm. However, due to specular reflections of solder joints, only true 3D
inspection can offer reliable solder joint results.

Figure 1.5: AOI solder joint inspection under IPC-A-610 standard.

Board Warpage
Board warpage also creates a challenge for 2D and even some 3D inspec-
tion. On a warped board, pad positions defined by board CAD or Gerber files
appear distorted, thus confusing conventional inspection systems and gener-
ating incorrect data. Board warpage becomes more pronounced during the
reflow process due to the high temperatures required for processing. Figure
1.6 shows how board warpage results in different height levels for the same
component due to the variable base plane difference—even when using high-
quality components. Suppliers should have created an advanced algorithm to
actively compensate for warpage.

Figure 1.6: Board warpage can create inaccurate

measurements and false calls if not accurately captured.

9
 Chapter 2

Performance Anxiety
A limitation of many 3D optical inspection systems is the cycle time typically
associated with processing millions of pixels to reconstruct a full 3D image
using data captured from multiple channels. There should not be a compro-
mise between 3D inspection and throughput. A successful inspection deploy-
ment should provide oversight for the process, not compromise, interrupt or
slow that process.

The performance or speed of inspection equipment falls into two camps.

One is the time taken to switch from one inspection program to another. The
second is the speed of the individual inspection. The former should be as close
to zero as possible and, at worst, not longer than the changeover time of the
line. The latter should be at the cadence of a line itself or better. The inspec-
tion system should never be the bottleneck on the line or the process that
dictates the speed at which boards are assembled.

Processes Optimized With Reliable Data

Some systems capture dozens of unique measurement data sets for each field
of view. This produces a lot of data, and equipment makers need to manage
the big data processing with parallel computing to satisfy even the most
demanding applications.

Equipment suppliers can improve 3D AOI speed thanks to hardware and

optical technology that generates reliable and repeatable data. These inspec-
tion results are then stored in a central database, which can be used in
deploying the optimal inspection program for multiple operations and reduce
programming and setting condition times.

Process optimization is desired by manufacturers and equipment suppliers,

including manufacturers of inspection equipment. However, this optimization
has been difficult to achieve with 2D systems, as most do not offer height
information, and cannot accurately measure and quantify shape, coplanarity
and solder amount. True 3D AOI systems measure every aspect of the compo-
nent and solder joint in accordance with the IPC-A-610 standard, and then a
significant set of reliable measurement data.

11
 Figure 2.1: A centralized database is an ideal solution
for efficient data management and analysis.

This data can be shared, stored and processed in a central server where it
will help with dynamic decision-making. Using various module applications,
manufacturers can monitor performance to help the real-time monitoring of
line performance with the highest level of traceability and transparency.

Other modules can optimize the printing and mounting processes by auto-
matically recommending optimal printing parameters in short, closed loops.
These closed loops are the building blocks of the smart factory. They provide a
feedback loop for a specific process and can be used to make real-time adjust-
ments that improve performance and reduce waste while also providing data
to the central nervous system of the factory.

No Escapes, No False Calls

Ultimately, we want to provide inspection solutions that perform at the highest
level. This means that they catch every defect accurately and do not call out
defects when there are none. Inspection utopia means no escapes and no
false calls, all derived at optimal line speed. This desire to drive down both
of these key performance Indicators is paramount for those developing and
building inspection tools.

Many users would say that these two data points are the only real perfor-
mance indicators for the success of an inspection implementation. Frustra-
tions with 2D inspection relate substantially to the large number of false calls;
switching from a 2D to a 3D inspection solution often eliminates 90% of false
calls. False calls do nothing but slow the process down, frustrate the operator,
and massively undermine confidence in the inspection process.

12
Escapes can be even worse, and sometimes it is difficult to set tolerances that
balance the two. The cost of escapes can range from fully assembled boards
failing final inspection and having to be reworked or scrapped, to boards with
latent defects reaching the customer or the field where they have a much
larger impact. Escapes can cause reputation damage, expensive recalls, field
service problems, and even undermine a brand’s position in the market.

Meanwhile, boards continue to become more complex with smaller compo-

nents that are packed closer together and joints that are harder to inspect.
These challenges and limitations of optical inspection systems cannot be
overlooked, because they can be the cause for false calls and escapes.
Those engaged in developing inspection solutions need to continuously
drive improvements in quality and performance while keeping pace with the
constant drive for smaller, smarter, and faster products.

13
 Chapter 3

The Data Dilemma

Data-driven analytics, using inspection data, can be used to overcome produc-
tion challenges and to enable the smarter factory of today and of the future.
Industry 4.0 and its associated benefits will doubtlessly advance the industry,
so equipment suppliers need to continue to work diligently to accelerate M2M
communication standards to aid the use of machine data to accelerate this
trend.

To this end, the role of inspection becomes two-fold. The first is to ascertain
if boards are good or bad at the point of inspection. The second is to provide
data for the greater good of the process, the line and—ultimately—the factory.

Connectivity is the Foundation of the Smart Factory

Initiatives like the IPC Connected Factory Exchange (CFX) and IPC-Hermes-9852
underpin efforts within the industry to develop standards and help create
a smart factory. These M2M communication standards, guided in part by
Industry 4.0, are altering the manufacturing process by improving metrics
such as first pass yield and throughput by applying autonomous process
adjustments.

Far beyond an automatic line changeover, this two-way communication allows

equipment to automatically adjust production parameters to increase board
quality and lower costs by eliminating manual labor, rework, and scrap. As
part of this mission, advanced process control with interconnected PCBA
equipment will revolutionize process optimization and lay the foundation for
the smart factory.

Measurement-based Inspection
Every electronics manufacturer—and most equipment suppliers, including
automated inspection providers—is looking to optimize the assembly process.
However, this is difficult, or even impossible, with the limitations of 2D imaging,

15
the former industry standard. Not only is it difficult for 2D AOI systems to iden-
tify defects on curved and reflective solder joints, but 2D AOI systems simply do
not generate data that is reliable enough to actually deliver consistent results.

Every aspect of the 2D inspection process relies on contrast—not quantitative

measurement. As such, 2D users must either suggest the repair or scrapping
of defective boards, which increases costs and eliminates the potential for
process improvements.

Finding the Component Body

The introduction of 3D imaging to the inspection market solved many of these
problems. By measuring components and solder joints, and then offering
critical height information to the inspection algorithms, users could identify
errors such as pad overhang and insufficient solder.

The 3D data is only valid when 3D technology is used for all component types
to extract the exact body dimensions. Systems that use “blob detection,” which
may be susceptible to external factors such as board warpage and component
proximity, are less reliable. The combination of measurement and process
data piles collected from its SPI and AOI systems, as well as from printers and
mounters, allows delivery of advanced AI features with reliable “big data.”

To overcome this, true 3D technology must extract the component body data.

While typical 2D and even some 3D inspection technologies combine real-

time board warp compensation to increase inspection data accuracy, newer
Figure 3.1: Only true 3D measurement
technology can locate the component body.
3D systems go much further and
can apply shadow-free 3D moiré
technology, thus, providing superior
results by measuring every aspect
of the component and solder joint
according to IPC-A-610 standards. The
ability to generate a significant set of
reliable measurement data gives rise
to industry-leading, measurement-
based analysis and optimization.

16
Ground-breaking Transparency
In a competitive market, manufacturers want to monitor and adapt the
process to drive toward zero defects while accessing data anytime, anywhere.
To cope with shorter life cycles, inspection solutions must be able to collect
and analyze large amounts of data to produce full traceability.

Likewise, leading-edge analysis solutions need to collect all inspection and

measurement data from every piece of equipment and then provide the data
anywhere within the network with an intuitive user-interface. This powerful
analytical solution provides the best possible level of traceability, transpar-
ency, and visibility.

Data is a fuel for Industry 4.0, so advanced inspection systems must evolve
from simply judging “pass/fail” into highly intuitive, dynamic decision-making
systems, emphasizing the need for reliable, traceable data. These systems
must ensure the highest levels of transparency and traceability by showing
the line conditions, including machine configuration and software version, all
while providing the required documentation for changes to the job files, pack-
ages, parts, and much more.

Users can quickly verify whether all lines are within the ideal condition. If a vari-
ance occurs, the user can instantly upload optimized programs and inspection
conditions with a library manager.

This software module illustrated in Figure 3.2 provides a complete central

management solution for component libraries, programs, inspection condi-

Figure 3.2: Examples of a library software module web interface.

17
tions, and more. It should combine all equipment into a single centralized
library. Managers can then trace and control all changes at the user level. This
control allows continuous data analysis and helps guide engineers in the right
direction.

Eliminating Bottlenecks
The system must instantly analyze and chart the data with relevant indicators
including yield rate, NG (No Good) analysis, DPMO (defects per million oppor-
tunities) analysis, Gage R&R (repeatability and reproducibility), offset analysis,
and additional metrics that allow users to compare board performance and
identify process deviations or potential problems.

By using real-time statistical process control (SPC) solutions, users can locate
the exact defect origin by checking false calls and NG components from the
dashboard, as well as evaluating and optimizing default settings.

For instance, if thickness is the major problem, users can click the compo-
nent to view the analysis result and identify the root cause of the issue. The
system will provide a measured component thickness X-bar chart captured
with average, minimum, and maximum values, plus tolerance levels.

Figure 3.3: X-bar chart for thickness of R1005 component.

If values often deviate from average values and tolerances are too tight, users
can adjust the tolerance levels to minimize false calls. On the other hand, if
the process is stable, operators can tighten the tolerance to prevent future
escapes. The measurable data helps improve the process.

18
In the meantime, operators can review defects with real data from all lines
with an offline program optimizer. Applications such as these simulate the
results of any adjustments without disturbing production, ensuring reliability
by allowing users to simulate the adjusted setting. Then, a library manager tool
can seamlessly deploy the optimized program(s) and inspection condition(s)
to all connected production lines.

These iterative actions can achieve continued process optimization based on

real measurement data, rather than user experience. The applications show
the process improvements with statistics and charts. The benefits are real,
reliable, and repeatable. Some manufacturers have achieved a 98% reduction
in false call rate by optimizing their inspection programs offline.

Manufacturers can also monitor line performance in real-time using real-

time monitoring software. This software module tracks all the machines in
the factory, with metrics such as the best performing machine yield and top
five defects, plus charts reflecting yield, availability, X-bar, and sigma. This
all happens on a single page so an individual engineer could easily monitor
the performance of many lines. Real-time monitoring software that signifi-
cantly reduces false calls and eliminates unnecessary handling can be lever-
aged while allowing one operator to monitor multiple lines with dozens of
machines. This frees up other operators to focus on other critical line tasks
and process improvements.

Connecting Big Data

A single inspection system has limits and cannot manage and optimize a
complete line in isolation. Knowing this consideration, some companies have
been working with printer and mounter partners to achieve total communica-
tion and streamline the surface mount lines, aiming for a zero-defect future.

Working with these partners, they can leverage the true power of connectivity.
M2M connectivity optimizes the process by exchanging real-time SPI and AOI
measurement data with other machines in the production line. The systems
feed real measurement data such as offset, volume, height, area, and warn-
ings to other systems, while identifying and analyzing trends to optimize the
process. Combining multiple modules can deliver unsurpassed performance.

19
For example, when the system combines multi-machine data results with M2M
connectivity, the connected inspection systems automatically define correla-
tions between the processes.

Figure 3.4: Example of a system using true 3D process

measurement data to help realize a smart factory.

Using this advanced communication, the AOI can feed corrected mounting
position values to mounters, which ensures the pick-and-place machines
mount the components in the correct position. This feature improves process
repeatability by automatically adjusting placements and identifying trends to
make further positional corrections. Some are working with industry leading
printer and mounter suppliers to empower connectivity and realize a smart
factory.

Autonomous Process Optimization

Understanding the increasing importance of networked intelligent systems
in the smart factory, suppliers must continuously test platforms for future
growth and expansion. When releasing new software modules, a manufac-
turer can remotely implement the software upgrades as needed. Harnessing
the power of connectivity extends beyond automated adjustments toward a
comprehensive infrastructure for autonomous process optimization, making
a smart factory well within reach.

Some collaborations have led to process optimizers. The software exercises

complex algorithms to develop closed-loop print process recommendations.
Some are powered by AI for adaptive learning, which evolves the systems far
beyond the roots as a closed-loop print process tool to become a system that
continuously monitors and adjusts the process.

20
An enhanced solution could combine real-time print process data with SPI
measurement data. For instance, there could be three modules:
1. An advisor module automatically performs a design of experi-
ment (DoE), which is created to perform a detailed SPI result
analysis using advanced diagnostic algorithms and noise filtering
models, and then recommends the ideal print parameters.
2. A diagnostic module that uses multiple anomaly detection algo-
rithms to actively optimize the print process and further reduce
false calls.
3. The third module would be focused on optimization with adap-
tive learning to generate models and fine-tune parameters
based on the diagnostics results.

While each module could provide standalone process benefits, the combined
power of the three modules would ensure the highest process reliability and
production flexibility without dedicated resources and expertise.

Increased Efficiency, Reduced Costs

There is always a drive to improve profitability. Manufacturers can remove
the associated scrap or repair costs from defective boards by identifying the
root cause and implementing process improvements with intelligent software
solutions. Manufacturers can further improve the return on investment by
deploying stored programs and eliminating the fine-tuning typically required
when changing over to a new product.

Moreover, since operators do not need to fine-tune or monitor processes,

managers can deploy them to other production areas. For example, instead
of five operators running five SMT lines, one operator can handle all five lines
using a remote monitoring software. Intelligent software solutions can offer
tangible benefits that satisfy the needs of real-world businesses. There is a
clear vision and focus on next-generation cooperative efforts that expand
process capabilities and improve factory performance.

21
 Chapter 4

A Vision of the Factory of the Future

Intelligent software solutions can help customers analyze and optimize the
production process by managing process data from connected SPI and AOI
systems to speed the arrival of the 4th Industrial Revolution and perhaps
even offer a path to the autonomous manufacturing future some are calling
Industry 5.0.

So, What Would the Ideal Smart Factory Solution Suite Look Like?
Industry 4.0, or smart manufacturing, should represent a revolution in
process optimization that ensures the highest standards of quality and reli-
ability on the SMT line and beyond. The software should collect inspection
and measurement data from all equipment into a central repository for defect
detection, real-time optimization, enhanced and accelerated decision making,
and detailed traceability. Once the data is collected and analyzed, the software
can be used to improve metrics, increase board quality, and lower costs by
eliminating waste, variance, false calls, and escapes.

The software should include multiple complementary software modules for

extensive data collection, management and analysis, as well as a compre-
hensive set of optimization solutions that would integrate 3D SPI and 3D AOI
systems to provide the power of connectivity and pave the way for Industry 4.0
with smart, data-driven analytics.

Using the right 3D measurement technology will deliver accurate measure-

ment data to help manufacturers streamline the SMT line and realize a zero-
defect production environment. The ability to generate reliable measurement
data has given rise to the creation of measurement-based analysis and optimi-
zation solutions. With quality control capabilities and a full lineup of integrated
inspection systems, the ideal software suite should tie everything together:
SPI, AOI and other production machines, all supported by AI-powered analysis
for fully automated control that boosts productivity and minimizes costs.

23
Focus on Data Management
Processed data is information, analyzed data is insight; data management
modules provide the key to efficient production and high standards of quality
and reliability.

Realizing a smart factory means taking a practical approach to process and

systems while examining areas to improve productivity. Using real 3D measure-
ment data generated during inspection helps manufacturers define inefficien-
cies and boost line efficiency. For example, a “Library Manager” module can
simultaneously deploy programs and inspection conditions across multiple
lines to enhance productivity and data integrity with consistent performance.

Operators can further improve line maintenance with real-time monitoring

where the system instantly displays relevant process parameters to remote
locations for immediate analysis and action. Adopting multi-point inspection
from SPI (post-print), pre-reflow AOI (post-placement), and post-reflow AOI
systems collects, manages, and monitors real data, so operators can deter-
mine actionable insights to optimize processes.

Suggested Software Modules

Library Manager
ƒ Deploys optimized programs and conditions to all inspec-
tion machines to seamlessly program, deploy, move, delete,
compare, and back up programs for all connected machines
ƒ Offers traceable management of changes via user level identi-
fication
ƒ Maximizes productivity by centralizing and optimizing all
programming resources in a single management system
Multi-point Connectivity
ƒ Connects all inspection systems within a line to consolidate
inspection results for review to find correlations between
processes
ƒ Streamlines production with regular data-based review, diag-
nosis and optimization of printing, pick-and-place and reflow
processes via linked inspection results
ƒ Provides review of SPI, pre- and post-AOI images, trend charts,
and inspection results to correlate printing, placement, and
reflow data

24
Real-time Monitoring
ƒ Displays real-time statistics so operators can monitor line
performance in real time
ƒ Enables a single operator to monitor multiple lines and dozens
of machines, which allows operators to focus on other critical
line tasks
ƒ Tracks various machines in the factory, showing the best
performing machine yield and top five defects list, plus charts
reflecting yield, availability, X-bar, and sigma all on one page

Focus on Data Analysis

The strength of the ideal smart factory software suite will exist in its analyt-
ical power. A statistical process control module can provide a straightforward
visualization of real-time production and inspection results, including configu-
rable charts for user-specific parameters. Users can identify the exact defect
origin by checking false calls and NG (no good) parts from the dashboard,
as well as evaluate and optimize default settings by comparing actual results
with average, minimum, and maximum values. For example, if the process is
stable, operators can further tighten the tolerances to prevent escapes.
Statistical Process Control
ƒ Provides a cross-platform multi-line process analysis applica-
tion that includes production performance analysis by product
ƒ Delivers detailed inspection results with visual graphics and
accumulated statistics through deep inspection data analysis
tools, including charts, histograms, and more
ƒ Instantly analyzes data visualized with relevant indicators to
compare board performance and identify process deviations
ƒ Comprehensive and extensible cross-platform modules make
it easy to maintain, control and upgrade new features via the
SPC page
ƒ Supplies a straightforward visualization of real-time production
and results for all lines
ƒ Offers charts and graphs relating to refined analysis results
from production, including configurable charts for user-specific
parameters
ƒ Drills down to the component level to identify root cause, to
minimize false calls, or to tighten tolerance levels to prevent
future escapes, as well as to evaluate and optimize default
settings

25
Focus on Process Optimization
Above all, the software suite should deliver the opportunity to realize a fully
automated and advanced manufacturing facility through a comprehensive
infrastructure for autonomous process optimization.

Several modules could work in tandem to harness the power of accurate

measurement results. In this scenario, an “Offline Programming Optimizer”
provides an intuitive graphical
simulation tool to review identified
defects with accumulated histor-
ical real data from all lines, which "In short, it will be a
avoids unnecessary down time.
This will help the manufacturer reli- quantum leap toward
ably foresee the impact of changing
process parameters without stop- the realization of a
ping production on the line.

Moving forward, it should be able

smart factory and
to autonomously render complex along the path
process optimization decisions
typically reserved for dedicated to autonomous
process engineers. In short, it will
be a quantum leap toward the real- manufacturing."
ization of a smart factory and along
the path to autonomous manu-
facturing. Overall, the system should be designed as a modular platform for
future development and expansion. As new modules are available, or when
a process requires additional smart factory modules, the manufacturer can
simply and easily implement the necessary modules.
Remote Monitoring System
ƒ Removes defect reviewers from the production floor while
providing real-time remote access to each piece of equipment
throughout the network
ƒ Provides direct control from a web browser at a remote loca-
tion to review all data, machine status indications including
yield and availability, and systems

26
Offline Program Optimizer
ƒ Allows operators to load and verify identified defects, while
enabling measurement-based simulations using real accumu-
lated historical data from the production lines
ƒ Simulates the results of any adjustments without affecting
production
ƒ Deploys new settings to all machines in the production line
simultaneously

Creating Smarter Factories

Machine-to-machine (M2M) is the foundation that drives the smart factory
vision one step further, enabling automatic SMT line maintenance and
management. The aforementioned software should be designed to work with
printer, mounter, conveyor, and other equipment suppliers to simplify commu-
nication across the entire SMT line. Initiatives such as IPC-CFX, IPC-Hermes,
and IPC-DPMX all help to pull the factory of the future into the present.

27
 Chapter 5

Advanced Process Control and Inspection

Reliable AOI methods have become powerful, economical complements to
traditional test strategies. AOI can be used successfully as a process moni-
toring tool for measuring printing, placement, and reflow performance. Some
advantages include:
ƒ Detecting and correcting SMT defects during process moni-
toring is less expensive than after final test and inspection,
where repairs are typically 5 to 10 times more costly.
ƒ Detect trends in process behavior, such as placement drift or
incorrect mounting, earlier in the overall process. Without early
inspection, more boards with the same defect could be rejected
during functional test and final inspection.
ƒ Identify missing, skewed, or misplaced components with incor-
rect polarity earlier in the assembly process when component
placement is verified before reflowing.

Yet, a single inspection system has limitations, especially when there is limited
or no communication with the balance of the line. In this setup, it simply cannot
optimize a printed circuit board assembly process. Equipment suppliers must
cooperate to achieve communication for a zero-defect future. M2M connec-
tivity can optimize the process by exchanging real-time SPI and AOI measure-
ment data with other machines in the production line. This real-time feedback
includes measurement data such as offset, volume, height, area, and warnings
to other systems, while analyzing trends to optimize the process and identify
trends. The connected systems can automatically define correlations between
the processes.

For instance, the electronics manufacturing industry has created numerous

studies and reports detailing how the solder reflow process can help position
surface mount components normally on the pads, even if component place-
ment is off pad. However, the trend to shrink components to 0.3 mm bumps
or 0201M microchips is opening doors to explore how process controls can
improve yields in high-density placements.

29
Enter advanced process control (APC), a proven control and optimization tech-
nology delivering measurable and sustainable improvements in production
yield. Most engineers will agree that stabilizing control loops, with underuti-
lized or ineffective process time and strong process interactions, is exceed-
ingly difficult. APC has become a standard solution for realizing stable control
processes. Quite simply, APC is the added value upgrade to a process automa-
tion system. APC collects and analyzes solder and component location data
from an inspection system, and then sends the recommendations across
the line to printers or mounters for automatic implementation, as Figure 5.1
shows.
Figure 5.1: Theoretical
SMT lines using Advanced
Process Control active
feedback between solder
paste printer, solder
paste inspection (SPI),
component mounter, and
automated optical inspec-
tion (AOI) systems.

APC Printer Feedback

The continued demand for smaller, lighter, and smarter electronic devices
has led to an increasing use of smaller components. These shrinking pack-
ages force smaller aperture designs and finer solder paste, which have made
stencil printing a highly sophisticated process with a tremendous impact on
production yields. It has been reported that 70% of all PCB assembly defects
are primarily due to problems directly associated with the solder printing
process. Manufacturers must ensure the optimal printing parameters are
consistently applied during production.

An enhanced solution, formed of interlinking software modules, can actively

optimize the printing process by combining real-time printing information
with SPI measurement data. More advanced software automatically performs
a Design of Experiment (DOE) intended to complete a detailed SPI result anal-
ysis using advanced diagnostic algorithms and noise filtering models, and then
recommends the ideal print parameters.

In the testing, the software conducted an automatic DOE by changing the

screen-printing parameters. When handling two parameters, it needed 11
prints. For three parameters (printing speed, squeegee pressure, and sepa-

30
ration speed), it would require 17 prints, as was used in the results shown in
Figure 5.2.

Figure 5.2: Three printer parameters (printing speed, squeegee pressure,

and separation speed) were all considered in this test run of 17 PCBs.

Figure 5.3: Automatic print process data analysis by SPI machine with feedback to printer.

31
The software triggers the SPI to send the information to the screen printer to
automatically adjust the parameters. The maximum and minimum values for
squeegee pressure, printing speed, and separation speed can be set when
adjusting the printer parameters as shown in Figure 5.3.

Once the system defined the optimum values for the printer parameters, 10
boards were printed and analyzed. The volume histogram and the X-bar and
sigma chart of 10 boards show that the screen printer quality between the
engineer and the KPO module is almost identical. The green-colored histo-
gram in Figure 5.4 represents the optimization completed by an engineer with

Figure 5.4: KPO test results comparing human (green) and AI (blue) printer optimization.

25 years of intensive print process expertise. The blue-colored histogram in

the same figure represents the optimization completed autonomously by the
software. The results are virtually identical. Other software can use multiple
anomaly-detection algorithms to actively optimize the print process and
further reduce false calls. Ideally, the software will evolve to autonomously
generate optimized models and fine-tune the process parameters in real-time
using actual results. While the software modules would each provide stand-
alone process benefits, the combined power of an AI-powered software suite
would ensure the highest process reliability and production flexibility without
dedicated resources and expertise.

32
APC Mounter Feedback Results
With the benefits of printers communicating with solder paste inspection
machines known, what about mounters and AOI? Connecting mounters with
AOI delivers obvious benefits such as improved yields, especially in high density
boards. The mounters use the received data to update the placement program;
thereby, ensuring the components are placed onto the solder deposits rather
than onto the substrate pads. This approach to placing components on the
printed solder uses the self-alignment principle to increase production yields
and reduce defects. When solder is off-pad and components are placed to
the predefined placement location in the program, self-alignment is not effec-
tive. During reflow, components will shift off pad, or bridge with other pads,
causing rework or scrap.

Figure 5.5: (L) Traditional chip placement onto pads before and after reflow.
(R) APC controlled chip placement onto solder deposit before and after reflow.

Alternatively, corrected placements will maximize the self-alignment prin-

ciple. Mounting these microchips onto the solder paste instead of the pad will
increase yields and quality. Figure 5.5 shows a set of test results. Using this
advanced communication, the 3D AOI can feed corrected mounting
position values to mounters, which ensures the mounters place the compo-
nents in the correct position. This improves process repeatability by automati-
cally adjusting placements and identifying trends to make further positional
corrections.

Figure 5.6: Examples of pre-reflow 3D inspection results identifying

missing, offset, or rotated parts, which is fed back to the mounter.

33
Using real data from quantitative measurements, certain 3D AOI systems
feed correct component position values to component mounters and ensure
components are mounted in the targeted position. This type of feature
improves process repeatability by automatically adjusting component place-
ment to the paste, rather than to the pad location (Figure 5.6). Moreover, the
software will identify the shift trend and conduct further position correction by
using true 3D measurement data from the AOI system.

Connecting inspection systems with mounters can help achieve complete line
communication and further enhance the value of the inspection process. For
example, M2M connectivity optimizes the process by exchanging real-time
measurement data between the printer, SPI, mounters and AOI systems. The
systems feed offset and warning data to other systems, while analyzing trends
for process optimization and traceability. Combined, this process provides
unsurpassed performance.

Communication between equipment will improve process repeatability by

automatically adjusting component placement to the solder deposit, rather
than to the pad location. This advanced process further improves microchip
mounting reliability. Figure 5.7 charts dramatic improvement across five
different defect types when a manufacturer uses advanced process control in
production compared to a conventional placement approach with no commu-
nication between systems. Networked intelligent systems that allow real-time
results to be correlated, calculated, and visualized will become even more
essential in the smart factory.

Figure 5.7: Impementation of APC sharply decreases post-reflow defects across the board.

34
M2M communication standards, guided in part by Industry 4.0, are quickly
altering the manufacturing process by improving metrics such as first pass
yield and throughput by applying autonomous process adjustments.

Component Size and Increasing Density

Are Driving Innovations and Collaboration
We have learned that APC can increase production yields and reduce defects.
For example, APC improves process repeatability by automatically adjusting
component placement to the paste, rather than to the pad location. Moreover,
APC can identify a shift trend and implement further position correction by
using true 3D measurement data from the AOI system. APC has the potential
to completely change process optimization.

As today’s board complexity is increasing with more components and joints,

higher density, and shrinking package technologies such as the 0201 metric
(008004 Imperial) microchips (Figure 5.8), basic AOI technologies using blob
analysis or high megapixel cameras may no longer be practical. Most deci-
sions made are based on a “good/bad” comparison of reference images, which
can easily be affected by variables such as component surface finish, board
condition, component proximity, and more.

Although 2D AOI is still in the market, more manufacturers are adopting 3D

AOI to increase board quality. The benefits are clear: using clearly defined
thresholds—backed by accurate data—will eliminate the need to constantly

Figure 5.8: Using a ballpoint pen for reference, the image compares 0201 metric
(008004 Imperial) component with an 0402 metric (01005 Imperial) and an 0603 metric
(0201 Imperial) component. (Image courtesy of ASM Assembly Systems GmbH)

35
debug inspection programs. Moreover, measurement data generated from
some 3D AOIs provides meaningful insights about the process and helps elim-
inate the root causes of a defect. Combining a 3D SPI with 3D AOIs enables
manufacturers to accurately control and monitor the board assembly process.

With so much data, engineers are hard pressed to collect, process, and imple-
ment all the data using traditional techniques and software. Artificial intelli-
gence and deep learning allow machines to learn from the vast amounts of
process data collected and adjust outputs based on the data inputs, performing
tasks to help engineers operate more intelligently. The examples that get the
most media attention, such as computers playing chess or autonomous (self-
driving) vehicles use deep learning to learn tasks by processing large amounts
of data and recognizing patterns in the data.

AI is ideal for volume PCB production and helps create a data set for a smart
factory. From statistical process control to instant program refinements,
AI-powered platforms can intelligently apply real-time data to improve
production processes.
Going beyond smart
factory solutions, manu-
facturers can use the
same technology to opti-
mize a process and adjust
parameters by utilizing
machine-learning algo-
rithms.

Realizing a smart factory

means taking a practical
approach to process and
systems, while examining
areas to improve productivity. Combining machine learning with 3D measure-
ment data generated during inspection, helps manufacturers define ineffi-
ciencies and boost line efficiency. Machine learning incorporates programmed
algorithms that receive and analyze input data to predict output values within
an acceptable range. As new data is fed to these algorithms, they learn and

36
optimize their operations to improve performance, developing intelligence
over time.

For example, some tools allow manufacturers to simultaneously deploy

programs and inspection conditions across multiple lines, which enhances
productivity and, more importantly, data integrity with consistent perfor-
mance. Operators can further improve line maintenance with other tools
for real-time monitoring to instantly display relevant process parameters at
remote locations for immediate analysis and action. Combining multi-point
views from SPI and pre- and post-reflow AOI with real data management and
monitoring allows operators to determine actionable insights to optimize the
processes. However, the adaptation of AI-powered process tools takes optimi-
zation to a higher level.

Converting all the data requires a simulation tool to review identified defects
with accumulated historical data from PCBA lines, while avoiding unneces-
sary downtime. Software tools can reliably allow manufacturers to predict
the effects from fine-tuning without stopping the line. Moving forward, an
AI-powered platform can autonomously render complex process optimization
decisions typically reserved for dedicated process engineers.

Embracing the connectivity will help create a smart factory. For instance,
software modules can exercise complex algorithms to develop closed-loop
process recommendations. The machine-to-machine connectivity drives the
smart factory vision one step further by enabling automatic and predictive
maintenance. Finally, combining inspection with printers and mounters can
enable the network tools to connect and simplify communication across the
entire PCBA line.

Defining the correct process parameters often requires a high degree of

expertise because of the various environmental considerations affecting the
process. Using AI-powered systems and M2M connectivity, manufacturers can
link inline inspection systems with the associated printer and mounters in the
line to overcome the challenges. As previously shown in Figure 5.4, automated
machine learning can already match the results from process experts—and
this will only improve.

37
 Chapter 6

How Far Can AI Take Us?

Artificial Intelligence is more than a buzzword associated with business and
industrial technologies. In recent times, it has demonstrated promising poten-
tial across a variety of industries, from self-driving vehicles to virtual doctors,
and it has far-reaching applications within the manufacturing sector.

By using the right mixture of AI technologies, manufacturers can boost their

efficiency, improve flexibility, speed up processes, and even achieve self-opti-
mizing processes. The SMT industry is no exception, and while it is facing a
chronic shortage of skilled labor, AI seems a natural fit. Equipment providers
are enabling the smart factory of the future by adopting AI to generate “knowl-
edge” from “experience.”

Machine-to-machine communications are quickly changing the manufacturing

process by aggregating process data such as first pass yield and throughput.
By strengthening large-scale data sets using the highest quality data captured
through industry-leading 3D metrology, AI algorithms necessary for smarter
manufacturing processes can be trained by utilizing the data sets. Equipment
manufacturers are adapting and adopting AI in the inspection technology.

What Is AI?
Artificial intelligence is a multidisciplinary field of science with the goal to create
intelligent machines by making them “smarter.” Historical applications of this
goal include natural language processing and translation, visual perception,
pattern recognition, and decision making, but the number and complexity
of applications have been quickly expanding. Out of all the advancements
researchers have made, the current driver of the AI inflection point is thanks to
major advances in deep learning. Deep learning, a subset of machine learning,
is an algorithm with a hierarchy of “deep layers” of large neural networks fed
by data without feature engineering. Koh Young is also applying deep learning
with its AI engine, solving problems that are difficult for experts to model. From
practical solutions to improve measurement quality and inspection accuracy,

39
equipment vendors are utilizing AI to meet the rising demand and challenges
in the SMT industry.

Value of Accurate 3D Data

Data is crucial to the success of the AI solutions. Deep learning effectiveness is
linked to the quality and quantity of the input data to address many different
requirements from a large number of fields.

The use of AI to provide smarter inspection systems is desired by every

inspection provider, but has been difficult to realize due to the limitations of
2D imaging, which was the de facto standard. Not only is it difficult for 2D
AOI systems to identify defects on curved and reflective solder joints, 2D AOI
systems cannot generate reliable data. Every aspect of 2D inspection relies
on 2D features such as contrast; thus, it is extremely challenging to correlate
it to the quantitative measurement of 3D objects. The use of 3D information
related to components and solder joints is essential.

Measurements Get Ever More Accurate

It all begins with solving inspection challenges of SMT assemblies. As previ-
ously mentioned, finished boards with reflowed solder joints and components
invariably present inter-reflection challenges. As such, the lighting reflections
will cause false calls that may generate erroneous measurement values. As
board density increases with smaller components and tighter interspacing,

Figure 6.1: 3D measurement improvement with Koh Young AI engine.

40
the reflection becomes more troublesome. AI is used to prevent measure-
ment errors by incorporating learning in the inspection solution. The hybrid
fusion of an analytical approach utilizing a mathematical measurement model
and AI used for learning abnormal symptoms from the combination of good
and bad measurement data, allows the detection and elimination of abnormal
measurements, which reduces false calls and escapes as illustrated in Figure
6.1. Through the hybrid fusion approach, the measurement accuracy only gets
better regardless of many different challenges.

Share Knowledge Between Lines

Another area AI can be proactively applied is with AOI programming. With
the help of deep learning methods using true 3D data, the assignment of
components on a PCB is gradually becoming autonomous. An AI-powered
auto-programming system proposes the recommended inspection conditions
based on 3D measured data, which not only simplifies inspection condition
programming, but also makes it faster and more consistent with the best
mapping conditions. Figure 6.2 describes the general process. This has been
seen to reduce job preparation time by up to 70%, which makes it an ideal
solution for high-mix, low-volume, or time-sensitive applications.

Moreover, prediction and decomposition engines allow components to be

categorized by lead count, component type, measurement score, and more.
These elements further help cleaning, parsing, enriching, and shaping the
data, which plays a major role in AI empowerment. Going further, AI will be
applied to pad grouping and inspection condition tuning, while incrementally
learning new packages at new sites.

Figure 6.2: An automatic programming system process concept example from Koh Young.

41
Improve Yield and Process Optimization
With increasing component miniaturization, improving inspection quality
and increasing inspection system throughput programming is paramount in
the manufacturing industry. These needs are increasingly fulfilled with more
computational power, which then yields even better inspection solutions. AI
allows such improvements to be fulfilled more quickly with machines that
continuously learn to solve new problems.

The main challenge is complete end-to-end optimization, harnessing the

power of AI solutions. As an example of the application of AI and inspection,
Koh Young has developed a process optimizer (KPO) solution outlined in
Figure 6.3.

Figure 6.3: An example from Koh Young showing an application combining AI with inspection.

KPO is the Koh Young smart factory solution driven by AI to control and opti-
mize the printing and mounting operations. KPO heavily relies on accurate 3D
measurements data and error detection from SPI and AOI machines, which
sets the stage for smart factory solutions.

The KPO printing solution includes three interlinking modules that exercise
complex algorithms to develop closed-loop print process recommendations:
Printer Advisor Manager (PAM), Printer Diagnosis Manager (PDM), and Printer
Optimizer Manager (POM).

The enhanced AI engine actively optimizes the printing process by combining

real-time printing and SPI measurement data. PAM automatically performs

42
DoE’s designed to perform a detailed SPI result analysis using advanced diag-
nostic algorithms and noise-filtering models, and then recommends the ideal
print parameters. PDM uses multiple anomaly detection algorithms to actively
optimize the print process and further reduce false calls. The final module,
called POM, uses an adaptive learning engine to generate models and fine-
tune process parameters. While each module provides inherent standalone
process benefits, the combined power of the three modules ensures the
highest process reliability and production flexibility while reducing dedicated
resources and expertise.

The KPO mounting solution

includes three modules called
Mounter Advisor, Mounter Diag-
nosis, and Mounter Optimizer. "Better inspection
Mounter Advisor studies and
analyzes the offset distribution of solutions, along
the major mounter elements and
differentiates these offsets from with AI and its
the component offsets measured
by the pre-reflow AOI system. The
associated benefits,
system can automatically identify
mechanical and software mounter
will help advance
failures, as well as the root causes.
The Mounter Diagnosis and
the manufacturing
Mounter Optimizer modules are
currently being developed in collab-
industry..."
oration with our mounter partners.
These solutions will recommend
ideal component placement positions by utilizing SPI, pre-reflow AOI, and
post-reflow AOI data, all the while automatically notifying technicians and
engineers about mounter issues in real-time during production.

AI and Inspection Mitigates Skill Shortages

Better inspection solutions, along with AI and its associated benefits, will help
advance the manufacturing industry as it confronts challenges, including the
lack of skilled manpower and cost minimization.

The electronics industry is facing a chronic skilled employee shortage. Job-

hopping and employee misalignment lead to high turnover, which further

43
compounds the challenge. While industry organizations such as IPC and SMTA
are tackling the issue head-on with education and training programs, a need
exists for equipment suppliers to cooperate with other industry leaders and
organizations to adopt machine-to-machine communication standards and
to ensure that the equipment they produce reflects the skills available in the
market.

One such tool being used by equipment and software developers is AI. The
truth is this does so much more than mitigate skill shortages, because there is
so much more to do.

What AI does exceptionally well is to protect the operator or SMT engineer

from the complexity provided by the huge amount of data currently being
generated. The reason AI is so good at chess is because it has seen every
scenario possible, witnessed and learned from every strategy and can, in a
fraction of a second, consider thousands of options, exploring the potential
outcome of each. This means that better decisions can be made much faster
and much more consistently.

And that consistency is hugely important for any manufacturing company. One
of the key challenges with manual operation is the variability of humans. What
we need is for every operator to look at the same data in the same way, make

(Source: Adobe Stock © Gorodenkoff)

44
the same decision and deliver the best outcome. What currently happens is
that some shifts do better than others, and some operators, often those with
greater experience, make better decisions. Shifting that decision making to
an AI-enabled software system means that it is not only consistent, but it is
constantly learning and improving.

What AI and well-designed inspection systems can also do to mitigate the skill
shortages, is provide faster operator learning and more intuitive assisted user
interfaces. This means less complexity for the programmer or line operator,
and a shorter time to get up to speed for new operators. For the line or factory
manager, this again means consistency.

The goal of the inspection equipment and software system is to augment the
operators, to allow them to do the best job they possibly can. This means
protecting them from complexity, creating a system that assists them and
learns from them, provides them with a simply intuitive interface, has the data
they need available instantly when they need it, and ensures they work consis-
tently.

Manufacturing needs to be able to operate without relying on tribal knowl-

edge, or that operator who just seems to be able to get the best out of the
line. Those skills need to be systemized and shared in a way that makes every
operator, new or experienced, as good as the best expert in their field.

45
 Chapter 7

Smarter Manufacturing Enabled by Inspection

There is much that inspection can offer the world of manufacturing, and in this
book, we have focused on the current state and the near-term future. Right
now, inspection plays an increasingly important role in making manufacturing
smarter. Inspection allows us to deliver the quality we know is essential for
the products being manufactured right now, and in the future. It is providing
the data that allow companies to improve their own performance and effi-
ciency. Data from inspection is contributing to almost every process on the
SMT line, providing real insight into faults and their root cause. This inspection
data is also providing immediate feedback to other processes in the line, often
in real time, making on-the-fly adjustments that reduce scrap, downtime, and
even the use of consumables such as solder paste and cleaning materials.
Inspection is picking up the slack in terms of skill shortages, allowing lower-
skilled operators to manage lines or parts of lines and, thanks to intuitive soft-
ware, allowing fewer operators to manage more machines and lines. All in all,
inspection systems and the exceptional data they produce are making a huge
contribution to the digital transformation of the factory floor and the drive to
Industry 4.0.

Figure 7.1: Data quality is more important than data quantity

to create effective and reliable solutions.

47
But what of the future beyond the short- and medium-term goals of digitally
optimized factories? How far can inspection and data take us?

We believe it can take us so much further than anyone has even imagined.
Some are already talking about what comes after Industry 4.0, a vision of
Industry 5.0 that many refer to as autonomous manufacturing. Some compare
the journey to autonomous manufacturing to the journey to self-driving vehi-
cles: we can see the future but there are a whole lot of steps along the way.
Right now, the manufacturing industry is at the assisted vehicle stage. We
have lane control, assisted cruise and much more besides. But to have SMT
lines that can completely take care of themselves, reconfiguring for optimum
performance and for every new product, is as far away as vehicles driving
themselves around cities without colliding or needing any human interven-
tion.

However, as with smart cars, smart factories are not fueled by gasoline, they
are fueled by data. And the quality and availability of that data will define how
well the factory runs. Smart cars will operate in smart cities, where data is
shared at speed and is sufficiently robust to trust in mission critical applica-
tions, like choosing to stop or go at a pedestrian crossing. Likewise, smart
machines need smart factories and they, in turn, need smart manufacturing
ecosystems and supply chains.

What is common to every “smart”

application, be that a car,
home, city, or a factory,
is the need to rely on
constant robust access
to high quality reliable
data. Put simply, the future of
manufacturing relies on the constant
development of higher and higher
quality data from every channel. It relies
on that data being available on-demand
and immediately. And it relies on that data
being captured by robust quality machinery.

Right now, we are all working hard to drive toward Industry 4.0, to
ensure that we can provide that data where it is needed and when it is

48
needed. This journey will continue. And we continue to push the boundaries
of what manufacturing companies can do, and to drive for greater efficiencies
and performance benefits of Industry 5.0, Industry 6.0, and whatever comes
after that!

We passionately believe that the inspection data is pivotal to the future of

manufacturing and will help propel the industry beyond even our wildest
expectations.

49
 Acknowledgments
Many thanks to the following colleagues who assisted with content and
research:
ƒ Brian Kang, Director of Research Center at Koh Young
Research America
ƒ Jenny Yuh, Marketing Specialist at Koh Young Technology
ƒ Joseph Park, Team Leader of Applications Engineering
Group at Koh Young Technology

Additionally, the author would like to thank Denis Kang, formerly of Koh
Young America, for his contributions.

Above all, the author would like to express his appreciation of Dr.
Kwangill Koh, whose passion and vision revolutionized the electronics
assembly industry by laying the foundation for True3D™ measurement-
based inspection technologies.

51
 About Koh Young Technology
Established in 2002, Koh Young pioneered a new market by launching the
first 3D Solder Paste Inspection (SPI) system using a patented dual-projec-
tion Moiré technique. Since then, we have become the global leader in 3D
measurement-based inspection equipment for the electronics industry.
Based on our True3D™ measurement-based inspection technology, Koh
Young has developed 3D Machining and Assembling Optical Inspection (MOI),
3D Dispensing Process Inspection (DPI), Semiconductor Packaging Inspec-
tion (MEISTER Series), and Medical Robotics for brain surgeries (KYMERO), in
addition to our pioneering SPI and AOI systems for circuit board assembly.
Through our technological innovations, we have secured thousands of global
customers, and maintain the largest global market share in the SPI and AOI
markets. Additionally, with our user-centric R&D activities, we continue to
leverage our core competencies, develop innovative solutions, and enter
new markets. Learn why so many electronics manufacturers trust us for reli-
able inspection at kohyoungamerica.com.

53