Practical
Training on Smart Manufacturing
Based on Project Case
This article is written at the invitation of a magazine,
according to the genre of officially published papers.
I. Preface
At present, the training of
Smart Manufacturing is weak in two aspects: (1)
there are few on-site cases: many trainings still focus on the
explanation of the importance of Smart Manufacturing and
some actions of major companies in Smart Manufacturing,
that is, "what have you done"? Our training is based on our
own Smart Manufacturing cases and focuses on explaining
"how to do" in Smart Manufacturing. (2) There is less
involved in engineering modeling, machine learning and
intelligent system development: these three points are the key
differences between Smart Manufacturing and other
manufacturing, such as digital manufacturing. In the specific
historical stage of Smart Manufacturing,
although basic automation and digitization are still of great
help to the manufacturing industry, these are not intelligent
after all, but only the indispensable foundation of
intelligent. Our training focuses on the three main fields of
Smart Manufacturing, although we also make some brief
introductions to other fields, such as MES.
At present, the foundation of Smart Manufacturing
technology is relatively weak. In the transformation from
mechanization to automation, digitization and intelligence,
make-up courses account for a large proportion. At present,
the development stage of the enterprise mainly lies in
automation and digitization. In terms of intellectualization
mainly characterized by engineering problem modeling, machine
learning and intelligent system development, at present, it is
mostly concentrated in the periphery of Smart Manufacturing, such as intelligent warehousing. At present,
the optimization of the core areas of the manufacturing
process is mainly focused on several platform developers.
However, Smart Manufacturing has developed
rapidly. Many countries' governments have given greater support to
Smart Manufacturing. It can be predicted that in the next two to
three years, the Smart Manufacturing environment will be
significantly improved. From the high-quality data required for model-based
machine learning and intelligent system development, it will
gradually be better guaranteed. At present, the team is
vigorously carrying out Smart Manufacturing training
while installing the developed mature defect early warning
system for the manufacturing execution system MES to provide
production guidance and reduce defective products; At the same
time, when the conditions are ripe, develop some equipment
software (equipment version of intelligent system).
Since there are few practical cases in engineering modeling,
machine learning and intelligent system development required
for manufacturing process optimization in, the team also
relies on the cases completed in Germany, the United States
and South Korea, in addition to the cases of the team in large
enterprises in China (BYD and Nisco, etc.). Case based
teaching goes deep into the level of "how to do", which is a
step closer than that in the market. In the next few years,
many countries will enter the main battlefield of "how to do" in
Smart Manufacturing, so it is time to provide training
in this field!
II. Case and Resource Based Training
The main trainer completed a
large-scale project with a total investment of DM 1 million
from the national DFG German scientific research and
Technology Association when he was doing a doctoral project in
Germany in his early years, and wrote a book. On the eve of
graduation, the main trainer was employed by the senior
engineer of Morgan construction company (now Siemens) in the
United States to be in charge of its model development. The
main trainer has developed three groups of more than 100 sets
of models by using a large amount of data accumulated by the
company for more than 100 years. After that, the main trainer
took these models to the front line of the factory to develop
Smart Manufacturing system; As a software engineer /
Software Consultant, he completed 30 computer courses for
eight years, developed 50000 pages of large websites and
dozens of apps, and achieved breakthroughs in a series of
computer theories such as object-oriented programming, SOA and
enterprise windows, reaching the level of doctor of Computer
Science in the United States.
Considering the weakness of current technical
foundation and the desire for Smart Manufacturing
technology, we decided to use the team's more than 200 on-site
cases of Smart Manufacturing in Germany, the United
States, South Korea, and other countries, as well as the
million pages of data collected in various countries around
the world to carry out a series of technical training to
improve the level of Smart Manufacturing in
manufacturing industry.
Our team has been using Smart Manufacturing to solve
manufacturing problems for decades. Even before the term of
Smart Manufacturing became popular all over the world,
we have been doing data acquisition, engineering modeling,
machine learning and intelligent system architecture
development in high-end manufacturing, and generating the best
parameter combination in the production process through
intelligent software, Transfer to basic automation. In the
early years, it was mainly in Europe and America, and then in
Asia, especially in some large companies.
Table 1: Experience
of the main trainer as
German engineering Ph.D. and US
software Ph.D.
Country |
Time |
R & D content |
Germany |
1990-
1995 |
-
Development of finite
element artificial intelligence model (German DFG)
-
Book publication: finite element simulation
of temperature / mechanics of materials with large
deformation (1996, German Edition)
-
It creates a precedent for the application
of finite element simulation in complex section production
surface factory |
U.S.A. |
1995-
2005 |
-
Morgan, more than 100 sets of production process models
(1995-1999)
-
8 years 30 computer courses, including on-the-job training
(1998-2005)
-
Software Engineer (cascat)
+ Software Consultant + 50000 page website + intelligent
software
-
Breakthrough in computer theory: OOP + SOA + enterprise windows, etc |
Table
2: items completed by the main trainer
Table 3: projects completed by American Metal
Pass (including
projects guided or participated by the main trainer)
US team: material processing project |
Number
(116) |
Intelligent
system development
Consultation and optimization
Galvanizing, coating and finishing line
Pickling line
Heating
furnace
Continuous caster section
Finishing line
Cooling descaling system
Marking and detection system
Rail factory
Type wire rod factory
Medium and heavy plate plant equipment
Rolling equipment factory
Plate |
(11
)
(14 )
(13 )
(6 )
(13 )
(7 )
(4 )
(4 )
(9 )
(4 )
(10 )
(4 )
(10 )
(7
) |
At the same time, the team has collected and sorted out
a large number of Smart Manufacturing related materials
in major libraries in Germany, the United States and Canada in
the past 30 years, amounting to millions of pages! The case
training of integrating these materials into a large number of
Smart Manufacturing projects is an excellent learning
opportunity for students! This enables students to really
learn how to do Smart Manufacturing. Imagine a speaker
who doesn't know how to manufacture intelligently. Naturally,
he can't teach students; If the technical materials on hand
are not comprehensive enough, what students can learn will be
greatly reduced!
Table 4: main technical data used in training
Number |
Technical information |
Pages
(10000 or eqv.) |
1 |
More than 200 books related to industry 4.0 / Smart Manufacturing downloaded from Aachen University Library in
November 2019 (PDF version) |
8 |
2 |
Relevant materials downloaded / collected in a Colorado
library for four days in the summer of 2007 |
12 |
3 |
The research results of a 15 year project brought from
Germany to the United States include more than 40 doctoral
theses and related models |
3 |
4 |
Over the years, a large number of technical materials have
been collected from the Carnegie Mellon University
Library, the University of Pittsburgh library and the
famous Carnegie Library in Pittsburgh (the main trainer
can borrow the books of all libraries). For example, there
are dozens of annual World Automation annual conference
proceedings of Carnegie Mellon University |
20 |
5 |
Technical materials collected from relevant libraries of
MIT and Harvard University |
10 |
6 |
Relevant data collected from several universities in
Oregon, such as the University of Portland and ogi
research center |
10 |
7 |
Relevant technical data collected from the University of
Vancouver (UBC) in Canada for three days |
3 |
8 |
A complete set of explanations on product quality and
optimization technology in the 50000 page Metal Pass online
technical consultation part |
1 |
9 |
Main technologies of more than 3000 automatic control /
intelligent control software on 50000 page Metal Pass
network |
1 |
10 |
50000 pages of tens of thousands of technical data and
related models on Metal Pass (the data is given in the form
of model, and the error is usually less than 1%) |
4 |
11 |
50000 pages of other technical materials on Metal Pass,
such as dozens of industrial apps, dozens of European and
American papers published and three books published in
Europe and America (for example, books of the trainer
published in Germany and USA) |
5 |
12 |
Morgan's reference room has a large number of technical
data, as well as a large number of technical data
collected by the company from customers around the world
for more than 100 years |
10 |
13 |
Other total |
20 |
III. Engineering
Modeling, Machine Learning and Intelligent System Agent
Technology
At present, Smart Manufacturing (but still not
the Intelligent Manufacturing) is mainly at the
beginning and end of the manufacturing process, such as
intelligent warehousing, incoming materials and finished
product transportation, etc., while the main process of
Smart Manufacturing, such as the optimization of
production process, is not much involved. We believe that the
main body of Smart Manufacturing should first be
engineering modeling, which models the main process parameters
and main product problems in the production process, so as to
facilitate the optimization and solution; The second is
machine learning. Based on the field data, the model obtained
from engineering modeling is optimized by machine learning to
make the model fully bound with the production line; The third
is the intelligent system architecture and development. We
should integrate a large number of field engineering models
and a large number of field scene use cases into the
intelligent system. The so-called scenario use cases mainly
refer to all on-site problems to be solved by all on-site
engineers. Theoretically speaking, the production based on
intelligent system makes the field no longer need to solve
these problems manually. Of course, these engineers /
technicians will go to the background to continuously optimize
the intelligent system; The optimized system will be operated
by field personnel to complete production. At the same time,
managers pay attention to observing the production results and
possible problems of this system. In this way, everyone works
hard to make continuous optimization of production! This kind
of control system is what we call Smart Manufacturing
system. The optimization of the head and tail of Smart
Manufacturing is also important, but the core should be the
optimization of Smart Manufacturing in the production
process.
IV. Three Attempts to Explore The Basis of
Smart Manufacturing of Chinese Enterprises and Students
1. Field project
After several years of cooperation with a large enterprise,
although the first, second and third phases of the project
have been completed, it is also obvious that it is difficult
to provide high-quality data at the level of Smart
Manufacturing on site, although these data basically meet the
digital manufacturing represented by data board. For example,
the MES system of the enterprise has no tool information.
Therefore, the relevant data about the use of tools have to be
found in the film detector database, and the film detector is
often not connected to the production line; Originally, this
was a small problem, but the site said that it must be solved
by a third-party network maintenance provider, which is always
delayed for a long time, and the supplier of the film detector
is very reluctant for us to obtain data from the film
detector. Although every company claims to have a large amount
of data on the table, there is little data that really meets
the needs of high-quality Smart Manufacturing, and it
often takes a long time! In this way, the cost of intelligent
system suppliers is huge. Enterprises do not have a clear
punishment system for the problem of missing data on site,
resulting in poor integrity of on-site data. Small businesses
often simply cannot collect the high-quality data they need. I
feel that the data problem is the biggest obstacle to
Smart Manufacturing, especially in some developing countries!
2. "How to do" after five basic trainings
First of all, we have conducted a primary training of ten
lectures and five courses. The focus of this training is to
teach students what to do in Smart Manufacturing?
Because we found that a series of losses caused by
misunderstandings in the Chinese industry are quite large. For
example, when building the Smart Manufacturing
production line, because it is not clear what is Smart
Manufacturing, we will only build the production lines of
basic automatic control, MES and ERP, mistakenly believe that
it is an Smart Manufacturing production line and lack
the core intelligent system of Smart Manufacturing. This
kind of production line can only be operated by engineers,
because engineers have the intelligence of production process,
but operators, due to limited intelligence, are difficult to
turn a large amount of data into intelligence in operation,
which is quite difficult to operate. Because there are many
problems in this kind of intelligent production line, the
benefits are slow, and there is even a dilemma of fast death
if you invest and slow death if you don't invest! Therefore,
our first stage of training focuses on letting the trainees
know what Smart Manufacturing should "do"?
However, in the second stage, when we want to teach students
"how to do", we find that there are many problems to be
solved. Recalling the journey I have traveled, I have
experienced for decades! In fact, Smart Manufacturing
needs to have a strong background in a series of fields such
as process, product, equipment, automation, data acquisition,
model and software. It is best to have actually worked on
site! It can be judged from this that it is difficult for
students to reach such a high level quickly only by a few
courses. To this end, we have decided to hold dozens of
lectures on Smart Manufacturing in the next years.
3. Recruitment status
Recently, the team issued an advertisement for dozens of
Smart Manufacturing technical directors and Smart
Manufacturing experts, clearly requiring the recruitment of
personnel with engineering modeling, machine learning and
intelligent system architecture development background to
focus on the optimization of manufacturing process.
V. Main Contents of Annual Training
At present, there is not sufficient training in the subject of
Smart Manufacturing (manufacturing process optimization). This training focuses on this aspect and omits the
fields before and after the manufacturing process. Firstly,
engineering modeling is carried out for the on-site process,
equipment and product quality of the enterprise; Secondly,
machine learning is carried out based on the built model and
the online data collected on site; Thirdly, as a key point,
the architecture and development of intelligent system are
carried out based on the above two stages. The intelligent
system contains about 1 / 3 of the engineering models and
about 2 / 3 of the use cases and user scenarios. There are
many courses in each of these areas. The technical points of
all lectures are based on the completed Smart Manufacturing projects. However, in addition to the above
three parts, each benchmark project and other projects are
listed in a separate group, which is called the field problem
analysis of Smart Manufacturing. In addition, a large
number of courses belong to the background foundation of
Smart Manufacturing and do not belong to any of the
above groups, which can be included in the overview and
requirements group.
Table 5: course content details
Group |
Category |
Main contents of the course |
Num. |
I |
Overview and requirements of Smart Manufacturing |
In addition to the following specialized fields, basic
technologies in Smart Manufacturing include
multi-level computer control, basic automation, MES, ERP,
learning methods, etc |
11 |
II |
Engineering modeling of Smart Manufacturing
(production process optimization) |
Engineering problems include the coordination between
various process parameters, the calculation of main key
parameters, the influence of incoming material parameters
and various engineering parameters on product quality, and
so on. The lecture is based on the actual cases of more
than 100 sets of models in the same industry and emerging
industries, as well as the three-day modeling lecture
conducted by the main trainer for a university and a
company. |
8 |
III |
Machine learning based on model and field data, data
acquisition, digital manufacturing, etc |
Firstly, collect the dynamic data of the site based on
digital manufacturing, then machine learn the established
engineering model, and use a large amount of online data
to inverse calibrate the model coefficients in the
engineering model, so as to fully bind them with the
process, products and equipment on site; With the change
of equipment and production conditions, the model
coefficient changes, so the model is fully bound with the
production line. The lecture is based on the actual cases
on site and the solutions of self-study logic problems. |
7 |
IV |
Intelligent system architecture and development, and
factory software engineering |
Based on the formed model (about 1 / 3) and various
scenario use cases of the production line (about 2 / 3).
Scenario use cases are mainly the sum of all solutions to
the problems solved by each engineer. Scenario use cases
also include some unique problem solutions of products.
For example, when thick materials are processed into thin
materials, due to the low temperature at both ends of the
material, the head and tail are thicker than the middle
part due to the rapid temperature drop during the same
reduction, so the solution of scenario use cases is
required. Usually, the corresponding software should be
constructed first, and then programmed by the programming
team. |
12 |
V |
Field case analysis of Smart Manufacturing |
In addition to the above technical points that call on the
field case analysis, the case analysis here summarizes the
completed field projects; In addition, the team is
studying the content sharing of various intelligent
systems. |
14 |
VI. Introduction
to Smart Manufacturing Training Benchmarking Project
The thermal and mechanical finite element artificial
intelligence model completed by the main trainer in Germany in
the early years created a precedent for the application of
this model technology in the final shape prediction and
production process optimization of large and complex sections
at the factory level. In the model development of Morgan
construction company (now Siemens), three groups of more than
100 models have been completed, including material deformation
model series, material force energy and power demand model
series, and material microstructure and mechanical property
prediction model series. After that, he took the model to the
front line of the plant to develop intelligent system, and led
the development of three secondary systems (Intelligent
System) for metal smelting electric furnace, refining furnace
and continuous casting for kescat company. See Table 6.
Table 6: Benchmarking items
Benchmarking
customer |
Benchmarking
Project |
Freiberg Laboratory |
(Germany) German Research Association project, artificial intelligence
+ offline model |
Morgan/Siemens |
(US) More than 100 sets of model development / offline model
development of manufacturing process |
Cascade
Steel |
(US) three sets of Level 2 development / online
model development |
EOS |
(US) model requirements ↑, and a new generation of Level 2
has been developed |
POSCO |
(Korea) production line process / equipment optimization ← online
intelligent design software |
NISCO / TISCO, etc |
(China) Level 2 optimization: application of a new generation
of Level 2 |
Taiyuan University of science and technology, etc |
(China) manufacturing industry / research and development of lithium
battery intelligent equipment |
BYD |
(China)
Lithium battery manufacturing, Level 2 development,
soft sensing technology |
Tesla |
(US) lithium battery pole piece defect model (Nevada super factory +
developed in California) |
After that, based on the technical foundation and solid model
background of German engineering and US software
Ph.D., the main trainer served as a technical consultant for
intelligent system development in various countries all over
the world, and successively completed large-scale
international benchmarking projects such as EOS of
the United States, POSCO of South Korea, NISCO and BYD of China, as well as Tesla of the United
States.
EOS's traditional high-end material project: in the
production process of hard and thin products, there were
defective products every day. During the return visit half a
year after completion, I was told that there had been no
similar defective products in the past half a year.
Figure 1: BYD Lithium Battery burr length is predicted and measured by model
(there are more than ten influencing factors, including tool
quality and use, process parameters and incoming material
quality, etc.)
BYD lithium battery Smart Manufacturing related
projects (see above picture): burr defect is the main factor causing lithium
battery fire (Samsung mobile phone lithium battery explosion, BYD lithium battery incident). At the beginning of the
cooperation, the factory used the pole piece slitting burr
prediction model which is very difficult to model to strictly
investigate the model level of the team. It is required that
the hit rate of the model is 85%, and the team has reached
98%; At present, the second phase of the project (burr early
warning) and the third phase of the project (knife notch
measurement) have been completed. In view of the weak data
acquisition ability of some countries, the
soft sensing technology with high difficulty in the industry
has been successfully applied. It is difficult to predict the
parameters directly by using the soft measurement model; This
defect early warning system is based on BYD MES and quality
inspection data. According to the evaluation of BYD's internal
cooperation in this project, at present, only Metal Pass can
provide this process / product model system combining
equipment and automation. Figure 1 shows the predicted burr
length based on the operating parameters of lithium battery
electrode. Compared with the measured burr length, the hit
rate is more than 98%!
Electronic manufacturing related projects: more than ten
enterprises such as Skyworth, TCL and Guangye. In addition,
there are still a large number of online information projects:
In Industry 4.0 Metaverse site Meta4-0.com, Dozens of apps, tens of
thousands of pages of data and dozens of categories of
technical materials (mainly in earlier www.metapass.com).
VII. Training as An Important Way to Gather
Talents
Smart Manufacturing talents need strong technical
background not only in the fields of engineering process,
products, equipment and automation, but also in the fields of
software engineering (Architecture and development) and data
acquisition. If you need to have working experience in
relevant fields, at least 20 years of work experiences are
required for a person to be mature. Therefore, real
Smart Manufacturing talents
are in great demand. Our training focuses on gathering these
multi-disciplinary high-quality talents together and further
improving their technical level to meet the needs of
Smart Manufacturing of our enterprise and related
enterprises.
Due to the huge cost involved in the international projects
and international resource collection used in the training, we
only take 1 / 10 of the actual cost as the training cost, and
if the trainees are at their own expense, it can be halved.
VIII. General Overview
A full year of Smart Manufacturing technology training
is being carried out. The training is based on more than 200
Smart Manufacturing projects completed by myself and my
team in Germany, the United States, China and South Korea, as
well as millions of industrial technical information collected
in major libraries such as in Germany, in USA and
in Canada. The purpose of training is mainly to gather talents
and further improve their skills.
Training
General,
Paper,
Courses,
Course overview,
Resources,
Trainer,
Experience,
Study method,
project,
Case-based
=========
Contact us: Please contact Linkedin in
www.Linkedin.com/in/intelli, OR scan the figure below to
add Wechat (e.g. myQQfriend); Tel: (+1) 858 898 1288; E-mail BLiMetaverse@Gmail.com. See
Enrollment
for Current Training Series.
|