Professional Documents
Culture Documents
™
SMART DATA
Using Data to Improve Manufacturing
BR Publishing, Inc.
dba: I-Connect007
942 Windemere Dr. NW,
Salem, OR 97304
U.S.A.
ISBN: 978-1-7342005-7-7
Visit I-007eBooks.com for more books in this series.
I-Connect007.com
Peer Reviewers
Happy Holden
Consulting Technical Editor, I-Connect007
Marius Stepanescu
Technical Director, ICCO EMT
Zac Elliott
Technical Marketing Engineer, Siemens
Digital Industries Software
1 Introduction
CHAPTER 1
Whenever we discuss data, keep in mind that people have been collecting
data, verifying it, and translating it into reports for a long time. And if data is
collected and processes are changed automatically, people still will be inter-
preting and verifying the accuracy of the data, creating reports, making recom-
mendations, solving problems, tweaking, improving, and innovating. What-
ever data collection system is used, any effort to digitalize needs to engage
and empower the production team at the factory. Their role is to attend to
the manufacturing process but also to act as the front line of communications
and control.
When operations are not performing as expected, they need to be able to:
1. Act as the first line of issue containment so that they can minimize the
effect of the problem on the final product or process through problem-
solving and corrective actions in real-time.
To help your factory take the next steps in the journey to digitalization, in this
book, we’re going to look at some of the major hurdles that your teams face in
collecting manufacturing data that then will be useful—not only for improving
processes but also for improving materials and supply chain management,
tracing the sources of problems and defective or counterfeit parts, and
providing trends analysis for business forecasting and reporting.
In our previous book The Printed Circuit Assembler’s Guide to… Advanced Manu-
facturing in the Digital Age, we looked at what needs to be done to create a
smart factory that is able to collect data—including how to remove barriers in
an organization, protect the data, and create a system for managing it.
1
In Chapter 1, we will look at the challenges of collecting good data and where
collecting it makes sense for the factory and improving business. In Chapter
2, we will examine what makes data smart, meaning the difference between
data on its own and analytics. In Chapter 3, we will cover how data can be
distributed from the underlying infrastructure for external use. We also detail
some of the tools available today to help you put these principles into practice
and look at a real-world example of how companies are reaping the benefits
of putting their data to good use with analytics.
The requirements for product quality and reliability contribute to the growing
need for meaningful analytics in manufacturing. With growing demands from
quality-sensitive industries—such as aerospace, automotive, smartphones,
and medical—manufacturers need to ensure their factory operations work
properly. Analyzing data simply is not enough. Company managers need to
use analytics to create knowledge that can positively affect manufacturing.
Figure 1.1: Newspaper Rock State Historic Monument has hundreds of petroglyphs
believed to be from about 2000 years ago. (Image Source: AdobeStock)
2
Advanced analytics refers to the application of statistics and other mathemat-
ical tools to business data to assess and improve practices. In manufacturing,
operations managers can use advanced analytics to take a deep dive into
historical process data, identify patterns and relationships among discrete
process steps and inputs, and then optimize the factors that prove to have the
greatest effect on yield.
3
4
Chapter 1
5
The Benefits of Using Data
One of the key benefits of using data effectively is boosting prediction accu-
racy, and this requires a deep understanding of the data and domain exper-
tise. It also requires access to the data in the first place, and not all companies
are willing to give that up. When this is the case, many companies commit R&D
resources to figure out how to do it.
6
Data is key to companies unlocking the ability to optimize their processes,
reduce costs, and accurately measure ROI—all by being able to make good
decisions about products, services, employees, and strategy.
Companies that lack the ability to use data in an intelligent manner to drive
informed business decisions are at a competitive disadvantage and will miss
actionable insights about their customers and their products.
Not knowing the best way to read, understand, and apply data can actually be
costing your business. Those costs could take the form of lost revenue oppor-
tunities, lower efficiency and productivity, quality issues, and more.
Businesses that fail to analyze the data from production equipment and
adequately predict machine performance and downtime put their manufac-
turing capabilities at risk. Breakdowns happen suddenly and cause a huge hit
to productivity.
For this sharing of language and data to work best, the data has to be visible,
accurate, detailed, and live to enable real-time communication and better
agility when reacting to changes.
7
These changes made it clear that smarter and leaner processes are necessary
for Industry 4.0 to take hold in manufacturing. Addressing such challenges
while maintaining and optimizing smart and lean processes requires software
and hardware solutions that can make use of data.
Figure 1.2: The stages from raw data to useful and actionable analytics.
Figure 1.3: Various machine interfaces, consistent and complete data delivery,
bandwidth, and integration efforts are all challenges to moving data from the
machine lines to applications that can provide actionable information.
8
The third-party software vendors that were supporting mixed-vendor lines
were often replaced with the machine vendor’s own software. This enabled
better balance and optimized the line in a more efficient manner, given the
complexities of this new generation of automatic placement machines.
Things started to change again when places like Europe and North America
brought an increasing trend of higher-mix production, which changes the
number-one challenge for manufacturers in these regions. Rather than
needing to focus on speeding up the placement for a single board, the new
trend created a need to focus on better changeover time between products to
optimize line utilization. Given that PCB manufacturers are still working with
more than one machine vendor, part-data creation, and management are put
in higher significance.
The stencil apertures can be a percentage reduction in the size of the original
copper pad, a specific distance inset from the copper pad, or some form of a
custom aperture that is not related to the copper pad underneath but more a
function of the package that will be placed on the land pattern.
Jet solder-paste printers, such as the Mycronic MY-600, are becoming more
popular for high-mix, low-volume designs, as well as for prototype runs,
replacing the stencil and screen-printing steps. Very small drops of paste
are deposited individually on each pad to build up the correct paste profile,
depending on shape and volume requirements.
9
Quick programming in the software will prevent having to cut a stencil and will
bring a quick turnaround of PCB assembly. Here, it is less about using data
to analyze the cost differential between jet printing and screen-printing but
more about the time-to-market aspect of getting the small number of boards
built in quickly as possible.
Solder Paste Inspection Data
Solder paste inspection (SPI) is used to detect any issues once the paste has
been laid on to the PCB. At this point, the little incremental cost has been
added to the PCB in the form of expensive components, but it is widely known
that ~70% of the scrap is caused by the solder paste printing process.
Traditionally, Gerber data is used as a starting point to create the SPI program.
However, in these circumstances, the information about which pad is associ-
ated with which component and component pin is lost. This makes aggrega-
tion of defect data challenging unless formats that are more intelligent are
used, or additional component and pin data is provided and tagged with the
component references.
SPI machines are available today that can perform the inspection of the solder
on the PCB autonomously. Based on the results, the data can be used to make
a decision regarding depositing additional solder paste in the case of insuf-
ficient paste on certain pads. This way, instead of having to clean the paste
off the defective board, the correction is made while the board remains in the
production line.
Automated SMT Placement Equipment
Any individual component on the board will only be physically placed by one
machine during the PCB manufacturing process. Therefore, decisions need to
be made that determine which machine will place each component onto the
PCB to maximize throughput or minimize line configuration changes.
With the trend to adopt a single vendor for SMT component placement, a
single machine vendor will typically provide multiple modules of flexible
placement machines to place the parts from the feeders and trays in the
10
correct positions on the board automatically. Smaller components are typi-
cally placed first, and then larger and taller components follow later. Given the
single machine vendor for placement equipment, it is more common for the
vendor’s software to perform the machine optimization and balancing across
the placement modules.
Smart production software must understand the work orders that need to
be scheduled at a higher level with respect to the finite limitations of the SMT
production environment. Enterprise resource planning (ERP) systems are
11
typically scheduling in an infinite planning environment. They consider a due
date, product, quantity, and material. If they have enough material, and based
on the beat rate of a line, they can determine that they can achieve a due date.
Unfortunately, SMT line capacity and material reels can affect these assump-
tions, which results in due dates missed and significant time needed between
product changeovers. Smart scheduling software that considers the line
capability and part supply form data can determine a much more achievable
schedule that works in conjunction with ERP to achieve a higher on-time due-
date delivery.
Once the PCB has been through the oven, post-reflow AOI can inspect both the
component and the joints because the solder has set. Hidden joints cannot be
inspected with AOI because this technique relies on a visual line of sight to the
12
joint. Components can be repaired if defects are found, although the repair
cost will be higher now because the existing component needs to be removed
and replaced with a new one.
In both cases, AOI measurements can be analyzed, and the results can indi-
cate if actions should be taken to improve the manufacturing process.
Automated X-ray inspection (AXI) can be used to detect component- and pin-
level issues and even find defects within the solder joints or hidden defects
under components because of its ability to look through the board or compo-
nent. However, AXI testing is time-consuming and costly, and machine-learning
algorithms can be used to predict which AXIs can be skipped to optimize the
cost.
The go-to solution for high volumes has been in-circuit test for more than
30 years. It uses electrical test techniques to isolate each component on the
board, confirm that the connectivity of the design is correct, and deduce that
the board should operate correctly. In reality, certain circuit configurations
and functional defects limit the detection capability that, in turn, will reduce
the overall fault coverage achieved with these techniques. Because it is fast,
it is used in high-volume environments; but because of the expense and time
involved in creating the in-circuit fixture, it is less applicable to low volume
situations.
Boundary-scan test is a specific test technique that uses the IEEE 1149 standard
to perform manufacturing defect detection via additional circuitry built into
13
certain types of components that provide simple access to complex compo-
nent circuitry. It can also be used for built-in self-test or in-system program-
ming. Provided the test access port of the boundary scan chain is exposed to
an edge connector, it can be straightforward to run.
Figure 1.4: SMT data format, consistency, application integration, and machine control.
Flying probe test uses similar techniques to in-circuit test, but instead of using
a custom fixture interface, it uses a limited number of robotic probes that
move and then press down to contact the PCB. Because of the physical move-
ment of the probes, the test times associated with flying probe test will be
longer than in-circuit test, but for small batches and quick turnaround, it can
be used when in-circuit test is not appropriate.
More intelligent test creation for flying probe allows the test time to be reduced
to align with the SMT line beat rate. Using board sampling techniques and only
Figure 1.5: Post-SMT process traceability, rigid test plan, complete line performance, and data loss.
14
running flying probe test if functional tests fail can be employed to align with
the overall production and quality goals. However, as the product mix and
manufacturing variants increase, non-fixturing test methodologies will likely
become more popular.
Data collected from the test machines can be used to improve the design for
the test process and reduce the cost and time of both the test programming
and the test duration and quality.
Normalizing, Interpreting, and Exchanging Data
Beyond the connection and exchange of data, normalized data is required to
integrate disparate automated processes and computerize human decision-
making. Normalized data is data that is expressed in a single language with a
consistent meaning regardless of the source of information.
To make the data useable from point to point and throughout the systems, it
needs to be understood from machine to machine, hardware to software, and
system to person. The point of normalization is to make variables comparable
to each other.
For example, you can measure temperature in both Fahrenheit and Centi-
grade in Seattle and Paris on a given day, and one is 68°F, and the other is
25°C. Saying 68 is bigger than 25, so Seattle is warmer, is wrong. Instead, the
measurements need to be reduced to the same scale and then compared.
Normalization is the process of reducing measurements to a neutral or stan-
dard scale.
Figure 1.6: In the entire machine network, the data can be normalized so that it can be used
throughout the factory to control and improve processes.
15
Normalized relations, and the relationship between one normalized relation
and another, mirror real-world concepts and their interrelationships.
In the factory, process applications are responsible for exchanging raw data
with equipment and operators. When connected using a gateway platform,
they can normalize the events and information into a single language for
external consumption. Internally, each application functions optimally for the
given equipment or process but can be set up to use the same interface to
describe the manufacturing operations being performed.
For the purposes of the adoption of computerized control over existing auto-
mated processes, and to enable more intelligent and flexible automated
processes in the future, the new internet of manufacturing communication
standard must represent all events of significance on the shop floor, not only
those from machines.
Figure 1.7: Maintaining a digital thread with open, intelligent data exchange formats.
16
large or small, can use the specification for communication projects custom-
ized to each operation, knowing there is full compatibility with any future proj-
ects and new machines.
The raw data captured from a machine or process, using whatever proprietary
interfaces the machine may have, or other data capture methods, will usually
simply state that the machine is, for example, stopped because no PCB has
arrived. As an event, this has little information other than the obvious. It can be
used to calculate machine uptime and downtime, but without understanding
the cause of the event, little else. Most messages from processes are similar to
these, which are very much dependent on— yet ignorant of— external issues.
In the normalized ODB++M environment, the system can receive the raw event
data and immediately investigate the cause of the non-arrival of the next PCB
to be processed. The preceding process may belong to a different platform,
be from a different vendor, or may be a completely different type of process
altogether. As ODB++M normalizes the data into a standard language, the
system can easily determine that the preceding machine had been stopped
for another reason, such as an error when attempting to pick up a material for
placement. The true location of the issue behind the event is clarified.
17
material had been delivered, an MSD issue had occurred with the material, or
a feeder jam happened.
Only by putting together all of the information from all of the relevant sources
can the original event be fully qualified, valuable event data created, and the
data used immediately for managing the live operation. The qualified event
can be stored in a cloud-based application, where subsequent analysis can
search for patterns that can yield significant improvement opportunities.
ODB++M makes all of this possible.
ODB++M enables value to be created from data from all kinds of machines
and processes, where individual vendors are open to share information and
to create value in their own products from the others’ use of ODB++M data.
Perspective and Qualification
Site applications are responsible for adding perspective from the complete
line and qualifying the data collected to identify root causes and bottlenecks.
When using a networked data collection and management system, the process
applications can all produce the same type of normalized events in a single
language; thus, minimal effort is required to connect processes for the highest
level of detail and perspective. In Chapter 2, we will look at what makes data
smart; that is, the difference between data on its own and analytics.
18
Distributing and Using the Data
In addition to the existing enterprise-level resources, such as ERP and MES, a
gateway combined with a standardized language provides the capability for
the discovery of resources in the factory and the events supported by those
resources. We also look at some real-world examples of how data can be
used to improve manufacturing and bring ROI to big data. In Chapter 3, we
will look at how data can be distributed from the underlying infrastructure for
external use.
19
Chapter 2
Companies have been collecting data in large volumes. Highly varied data
from manufacturing operations comes in quickly that needs to be validated,
and its value prioritized so that it can be turned into something useful—trans-
formed from big data to smart data.
The amount of data available has grown exponentially into big data. Twenty
years ago, a PCB work order resulted in 100 data records, megabytes of data;
today, it is 10 billion records, terabytes of data. The investment in collecting
this data and storing it is high. However, without a way to analyze the data,
without analytics, it will not result in ROI.
Using analytics provides the best results with high-quality data. However, with
bad or non-validated, and inappropriately prioritized data, the most advanced
analytics still can be misleading and a waste of time.
What is high-quality data? It is data that has been normalized, formatted, and
validated so that it can be translated and read to provide insights and fore-
sight. It is data that can be understood at the point of consumption and be
immediately acted on.
21
Data analytics is the process of exam-
ining data sets to draw conclusions
about the information they contain.
Manufacturing intelligence means
taking big data, applying advanced
analytics, and putting it back into
manufacturing processes.
What kind of data is most useful for improving and modifying manufacturing
processes?
Useful data comes from managed material handling across the factory,
including:
• Incoming material management and labeling
• JIT Kanban material delivery and FIFO
• Material setup and verification for the entire process
• MSD, AVL and LED management, and feeder maintenance
management
• Material consumption, replenishment, and traceability
• Asset utilization and monitoring
22
What is needed to make this data actionable?
• Automated capture of the full build record from machines and
material verification processes (including events, tests, inspec-
tion, results, symptoms, causes, and resulting actions)
• Support for all materials, including PCB, SMT, bulk, hand-placed,
plated through-hole (PTH), solder paste, and serialized and
assembly parts
• The ability for process control
• Consistent data across all platforms and processes
• Efficient and practical data storage and fulfilling long-term
storage requirements
• Comprehensive reporting tool
We talk a lot about the digital twin. The whole point of having a digital twin for
an electronics manufacturing operation is to be able to analyze and under-
stand what is happening, why it is happening, and how to improve it. Research
shows that while the amount of data generated by manufacturing operations
increases exponentially, a small portion is analyzed.
23
Further, it optimizes operation and labor by measuring and analyzing how
resources are spent and by tracking WIP in real-time. In addition, it ensures
quality and drives improvement by identifying and analyzing process defects
and material and process failures and increases design-to-manufacturing effi-
ciency by detecting factors affecting yield and areas for improvement.
24
At the diagnostic and predictive levels, data incorporates all collected param-
eters at corporate and site levels. It can be used to identify root cause, yield,
utilization, quality and performance issues, and for trend analysis. A good
analytics tool can be used to perform a root-cause analysis and discover why
the KPIs behave the way they do. At this point, it is useful to be able to connect
multiple sources of data.
Here, a report shows two alternate components from different vendors. When
factoring the quality data with the cost of the components, we see that the
cheaper component actually ends up costing more because of the scrap mate-
rial resulting from it.
25
sioning and tolerancing, identifying critical product and manufacturing infor-
mation.
PCB assembly process planning creates a digital twin of the production process,
enabling a “design anywhere, build everywhere” approach. Box-build planning
prepares process plans for NPIs, identifying the effects of design changes on
box-build lines, and delivering updated work instructions.
26
Figure 2.4: Examples of a sophisticated matrix. The clusters show group suppliers by mate-
rial characteristics. Regression shows the influence of material age on the potential for
defects. This image also shows the correlation matrix of the four KPI parameters.
27
Figure 2.5: A report can become the driver to make the necessary process changes by analyzing
the correlation between different parameters. For each component, we can see the different
manufacturer/supplier and unit price. A certain manufacturer may be more expensive for a
specific component, but its yield is higher, so its factored unit price will be lower than the other
and recommended to be used. The analytics shows how much money you can save if ordering
10,000 units from this component.
Figure 2.6: The distribution shows us all times when a machine stopped for changeover.
Thresholds can be defined to determine which changeover times were within the norm and
which deviated from it.
Not Just About Collecting Data, but Making the Data Smart
Industry 4.0 is not just about connecting another machine and collecting the
data. It is really about the mindset of understanding how problems can be
solved—looking at improving and moving toward excellence. Manufacturers
can buy a digital solution out-of-the-box from most companies in this area,
but they really must focus on how to use the knowledge in the data and the
software solutions implemented to improve the business.
28
Many of our customers are looking at overall equipment efficiency or overall
line efficiency only from the placement point of view, not the overall line. It can
be that the bottleneck is in the oven. They also do not consider that change-
over is actually a plant stop. This simple example shows the importance of
data.
Let us say, in a given week, we have 10,080 minutes, and we need to run main-
tenance, a few changeovers, inventory count, and shift changes—all planned
stoppages. Next, we have weekends and holidays. Actually, we already lost a
chunk of the available production time. In the following example, you can see
there is no third shift (we need to reduce 2,400 minutes), so by design and
plan, we would never get 100% overall equipment efficiency (OEE).
29
We have a line waiting upstream. Let us say the printer had to wait for the
pick-and-place machines, for example. We also have a line waiting down-
stream—the oven is waiting. From the total time, we “wasted” 76% on the
planned factors, and out of the total availability, the performance was not
good enough.
If the printer is really slowing down upstream, maybe we should take the
SP+SPI offline. However, to make a case and prove it, we need a baseline to
compare it with. With such a simple step, we can increase the line perfor-
mance by 10%, and we will never have soldering issues on the line itself.
In this example, we can see the overall picture. The real numbers factored into
the OEE/OLE that is presented with the current utilization rate (CUR)—that is,
how efficiently we used the actual available production time. To summarize,
making the data smart means taking the previous example and automating
the process so that this data is continuously monitored and analyzed for
better decision-making. Digitalization has to start by turning data into smart
data.
30
31
32
Chapter 3
With all of this complexity, a bottleneck is often caused by some external force
that is not being measured. A machine may not be operating because of an
actual malfunction in the equipment, or it may be waiting for some upstream
or downstream process. Perhaps the operator is on break, or there is a
shortage of materials causing the downtime. To identify the root cause of a
problem and provide for an actionable response, these external forces must
be considered.
33
With information about process performance and the external constraints
influencing production, many optimization opportunities are possible using
an intelligent application. The process-specific layer would be able to optimize
based on external knowledge from other processes and higher-level appli-
cations, while the site-application layer would benefit from detailed process
information from each individual equipment.
Feedback
A prime example of applying these layers in a smart factory is in closed-loop
feedback. In this scenario, measurements taken at one process are used to
adjust the operation of another process automatically to maintain a consis-
tent result. For example, the SMT machine could adjust placements based on
drift data measured at the AOI.
Because a normalized interface is used at the AOI and the SMT machine, this
application functions across varying platforms while allowing each individual
equipment to take the optimal action for its technology.
Finite Planning
Finite planning of the SMT schedule can be significantly improved and opti-
mized through automation and computerization. In the typical situation, the
ERP system manages the customer demand and material requirements in
coarse granularity with little detail of the resources used in manufacturing.
Once the work order demand is generated in the ERP system, diligent work is
put into developing a production plan to satisfy the orders.
34
If we use a smart factory topology, automation of the finite planning process
functions at each layer of the factory. The ERP system manages the customer
requirements and the high-level site calendar. A digital model of the production
process is generated based on constraints in the factory. All lines, machines,
processes, materials, transactions, and resources are taken into account in the
model to create a simulation of the manufacturing process. It supplies real-
time performance information from the manufacturing equipment through
the IoT infrastructure. The model also supplies the means to simulate produc-
tion for the given process.
The first step to developing the lean material management engine is accessing
information held in various systems. The ERP system provides the work-order
demand that defines the sequence and schedule of products to run. The
warehouse management system provides the detail of individual components
that are available for production. At the process-specific layer, the equipment
system provides the machine program information, performance information,
and material consumption details.
35
Next, using the production schedule, the current machine setup, and the
live IoT data stream from the equipment, the lean material engine can work
out when individual components will need to be replenished—either on
the current order as reels are exhausted or on an upcoming order during a
changeover. With the connection to warehouse management, the lean mate-
rial engine can determine the ideal location from which to move components
and automatically initiate the movement transactions. The consumption data
reported by the individual machines can be aggregated, and accurate reports
made to the ERP system.
Finally, the large line-side buffer stocks are unnecessary. Material is ordered
from the warehouse or from Kanban storage only when it is needed on the
machine. With the automated reporting of consumption and wasted mate-
rials from the machine, the ERP inventory is as accurate as possible.
Traceability
Collecting traceability data has traditionally been a difficult requirement for
manufacturers. The complexity and cost associated with collecting detailed,
accurate data could lead to inconsistent results because individual trace-
ability requirements are negotiated between the customer and supplier on a
product-by-product basis.
The requirements for traceability data collection can be fulfilled by the existing
process applications in the smart factory infrastructure. Because the process
applications support a neutralized language and a normalized set of events,
consistent information can be aggregated by the site layer applications regard-
less of the particular machine platform. Detailed and consistent data can be
gathered regardless of the equipment platform.
36
Manufacturing Analytics and IoT Systems
Now let us look at specific applications that will allow us to turn data into busi-
ness intelligence. First, we will cover the specifics that target this specific need.
Then, we will look at a real-world example that illustrates how it is being done
today.
With business analytics, you can harness historic data to drive informed deci-
sions. Companies can unearth powerful insights by identifying patterns in
37
thousands of readings from many products over time. The ability to predict
anomalies in equipment performance can result in significant savings on
scheduled repairs, maintenance cost reductions, and the virtual elimination
of breakdowns.
To begin, you can get a full enterprise performance overview for all manufac-
turing sites, lines, and equipment, including overall equipment effectiveness,
KPI reports, and dashboards. You can compare performance between facto-
ries and production lines, view history trends, and get an intuitive display of
results using graphs and detailed text listings of search results. Other data
sources can be integrated to enrich and share your reports and dashboards
throughout the company and with customers.
Video: Collecting data is not enough for improving PCB manufacturing processes.
The challenge is to turn big data into smart data. Watch this video clip to better
understand how Opcenter Intelligence Electronics can help achieve this target.
38
It’s a big-data analytics platform focused on and optimized for electronics
manufacturing. Its capabilities cover the areas discussed earlier: descriptive
analytics that provides real-time and historical KPI data, diagnostic analytics
that provides root-cause analysis and drill-down, and predictive analytics
for advanced trend analysis and prediction. Now, let us look at a real-world
example of companies putting their data to good use.
The company introduced its own PCB assembly process using K&S machines.
ICCO serves as a beta site for new K&S software releases. In this win-win situ-
ation, ICCO has access to the most advanced, cutting-edge software, and
K&S can test their software in a real production environment. ICCO is highly
involved in K&S’s software development process and can influence it as well,
providing the feedback that is so essential in the release of stable, fine-tuned
software systems.
39
Siemens provides K&S with OEM materials-management capabilities for their-
production software. While working with K&S, ICCO was exposed to our digital
manufacturing software and evaluated the solutions for their own use.
They also use Opcenter Execution Electronics IoT software to acquire valu-
able manufacturing data that is used to analyze and improve production
processes. The software runs on any Linux-based system and easily connects
to the devices on the line, generating a real-time digital twin of shop-floor
performance.
40
trends and proactively perform maintenance when required, tracking machine
uptime/downtime and percentage of defects, for example. The ability to track
incoming material helped reduce their inventory levels.
Using these execution and intelligence electronics tools, ICCO accelerated its
shop-floor processes by 20% while cutting the number of errors nearly in half
by using automated data mining processes and analytics. Performance data,
once available only in historical reporting format, can now be accessed by
ICCO in real-time when it is needed. Problems can be fixed immediately, and
scrap is minimized.
41
Conclusion
IoT applications are gathering huge amounts of real-time, shop-floor data
constantly but collecting data simply is not enough—it needs to be used intel-
ligently. Analytics is the application of statistics and other mathematical tools
to business data to assess and improve practices. In manufacturing, opera-
tions managers today are using analytics to dive deeply into historical process
data, identifying patterns and relationships among discrete process steps and
inputs and optimizing aspects that have the greatest effect on yield.
Siemens, in its own factories and with customer partners worldwide, is putting
analytics solutions to work to enable better decision-making and ultimately
to reduce cost and waste. In the real world, manufacturers can use data to
improve factory performance, product quality, and profitability. Each manu-
facturing operation is unique, so let us help you get started down the right
path to the best solution for your factory.
References
1. Cadalyst Staff. “Closed-Loop Digital Twins are Key to Harnessing Data
Complexity,” June 2019.
3. Franke, M. and Avidan, I. “Improving the PCB Print Process Using Factory
Data: How We Transform a Buzzword into Manufacturing Excellence in
Electronics,” presented at IPC APEX EXPO 2019.
42
About Siemens Digital Industries Software
Siemens Digital Industries Software is driving transformation to
enable a digital enterprise where engineering, manufacturing and
electronics design meet tomorrow. Our solutions help companies
of all sizes create and leverage digital twins that provide organiza-
tions with new insights, opportunities and levels of automation to
drive innovation. For more information on Siemens Digital Indus-
tries Software products and services, visit siemens.com/software
or follow us on LinkedIn, Twitter, Facebook and Instagram.
43
OEE Quality
OEE Quality
OEE Quality
88.7% 97.8%
88.7% 97.8%
Availability Performance
Availability
88.7% Performance
97.8%
Availability Performance
94.6% 95.5%
94.6% 95.5%
OEE Quality
1.90%
Pickup Error Monthly
94.6% Trend 95.5%
1.90%
1.25%
Pickup Error Monthly Trend
1.10%
88.7% 97.8% 0.80% 1.90%
1.25% 0.68%
Pickup Error Monthly Trend
1.10%
Availability Performance 0.34%
1.25% 0.80%
0.18% 0.17% 0.18% 0.68%
0.10%
1.10%
0.34% 0.80%
0.68%
0.18% 0.17% 0.18%
0.10%
Defects Pareto 0.34%
94.6% 0.18% 95.5%
0.17% 0.18%
0.10%
80%
Defects Pareto
65%
Quality 100
1.90%
60%
Defects Pareto
35 35 80%
Shortsolder Shortsolder
26
18
29
Pin in air/tombstone
16
23
61
NA
100
60 60% 40%
Minimum 31%
97.8% 38 35 35
0.68%
0.80%
2060
20 12 11 11 11
6 Minimum 31% 40% 20%
38 35 35 5 5 5 3 3 3 2 1 1 1
0.34% 20 12 11 11 11 20%
20 6
0.17%Performance
5 5 5 3 3 3 2 1 1 1
Missingsolder
82
Missingcomponent
80 80
Shiftcomponent
28 28
20 20
24 24
26
18
29
Pin in air/tombstone
16
61 23
61
Wrong Component
78 67
Billboard 78
Billboard
NA
0.18% 0.18%
0.10%
Missingsolder
82
Missingcomponent
Shiftcomponent
Reverspolarit
Shortsolder
26
18
29
Pin in air/tombstone
16
23
67
NA
Defects Pareto
95.5%
Cum. Defect Percentage
80%
65%
100 60%
1.90%
60 Minimum 31% 40%
38 35 35
20 12 11 11 11 20%
20 6 5 5 5 3 3 3 2 1 1 1
1.25%
Missingsolder
82
Missingcomponent
80
Shiftcomponent
Reverspolarit
28
20
24
Shortsolder
26
18
29
Pin in air/tombstone
16
23
61
Wrong Component
67
78
Billboard
NA
1.10%
Smarter Decisions
0.80%
0.68%
0.10%
of data. 60%
However, approximately
Minimum 31% 40% 70% of the data is not
20%
11 11 6 5 used
5 5 3due
3 3 to
2 the
1 1 lack
1 of an easy platform that will
leverage the data in an intelligent manner to drive
24
Shortsolder
26
18
29
Pin in air/tombstone
16
23
61
Wrong Component
67
78
Billboard
NA
Visit
siemens.com/software website