Article Type : Review Article
Authors : Sheng Pin Kuan and Shin Li Lu
Keywords : Data science; Big data; Quality trilogy; Statistical trilogy; Artificial intelligence
Statistical thinking is very important for modern management
and technological personnel. When a manager or engineer is be required for
reporting related issues, if there is an additional objective statistical data
to enhance the evidence of the report, then it will be easier to convince
readers. Realistic conjectures without statistical data basis will be regarded
as a metaphysics; the interpretation of the real status is supported by
statistical data basis, and the cause and effect can be induced. With the
deductive logic derived theory, coupled with the verification and induction of
statistical data, you can be sure of the rationale. Quality practitioners
should cultivate statistical thinking skills by themselves, and can thoroughly
understand the quality problems from the ways of three quality principles:
“essence of substance,” “process of business” and “psychology.” Traditional
quality data are nothing but variable data, attribute data, defects, internal
failure costs, external failure costs, etc., and these data are also collected
through data gathering, data processing, statistical analysis, finding root
cause, and so on. In the past, quality practitioners also relied on these
so-called professional skills for a position in the company. When these quality
data are collected, organized, analysed and monitored automatically by
computer, if quality practitioners do not to keep up with the steps of times,
it will be confused or full of difficulties. Precision tool machines are also
embedded in various IoTs due to the development of Internet and IOT technology,
collecting machine operation status, component diagnosis, machine life
estimation, consumables monitoring, power consumption monitoring, utilization
monitoring, and various types data analysis, and so on. Production site data
are transmitted to the cloud server through the Internet, such data mining and
forecasting have been gradually integrated to be the Data Science, and it will
be the future of quality field worthy of thinking.
Statistical thinking is
very important for modern management and technological personnel. When a
manager or engineer is be required for reporting related issues, if there is an
additional objective statistical data to enhance the evidence of the report, then
it will be easier to convince readers. Realistic conjectures without
statistical data basis will be regarded as a metaphysics; the interpretation of
the real status is supported by statistical data basis, and the cause and
effect can be induced. With the deductive logic deriveda theory, coupled with
the verification and induction of statistical data, you can be sure of the
rationale. Quality practitioners should cultivate statistical thinking skills
by themselves, and can thoroughly understand the quality problems from the ways
of three quality principles: “essence of substance”, “process of business” and
“psychology”. Under the guidance of Internet thinking, one is Big Data
thinking, which is written like this: In the era of Big Data, enterprises strategy
will shift from “business deriver” to “data deriver.” Massive data information
of user’s consume behaviour looks chaotic, but behind it which is the potential
logic of consumer behaviour. Big Data analysis can learn about the inventory
and pre-sales of products in various regions, time periods, and consumer
groups, and then conduct market judgments, and adjust products and operations
based on this [1]. Users generally generate data, behaviours, and relationships
on the network. The precipitation of these data helps enterprises to make
predictions and decisions. In the era of Big Data, data has become an important
asset of enterprises, even is core assets. The value of Big Data is not big,
but the ability to mine and predict. The core of Big Data thinking is to
understand the value of data, create business value through data processing,
data assets become core competitiveness, and small enterprises must have Big
Data also. Traditional quality data are nothing but variable data, attribute
data, defects, internal failure costs, external failure costs, etc., and these
data are also collected through data gathering, data processing, statistical
analysis, finding root cause, and so on. In the past, quality practitioners
also relied on these so-called professional skills for a position in the
company. When these quality data are collected, organized, analyzed and
monitored automatically by computer, if quality practitioners do not to keep up
with the steps of times, it will be confused or full of difficulties. Precision
tool machines are also embedded in various IoTs due to the development of
Internet and IoT technology, collecting machine operation status, component
diagnosis, machine life estimation, consumables monitoring, power consumption
monitoring, utilization monitoring, and various types data analysis, and so on.
Production site data are transmitted to the cloud server through the Internet,
such data mining and forecasting have been gradually integrated to be the Data
Science, and it will be the future of quality field worthy of thinking.
In November 1997, Dr. Chien-Fu Jeff Wu gave an inauguration speech when he was appointed as the H. C. Carver Chair Professor at the University of Michigan. The topic was: “Statistics = Data Science.” In this lecture, he described statistical work as a trilogy of data collection, data modelling and analysis, and decision-making [2]. In his conclusion, he coined the modern term “Data Science” and advocated that Statistics should be renamed Data Science, and statisticians should be renamed data scientists. Later, he also gave a lecture entitled “Statistics = Data Science” as his 1998 P.C. Mahalanobis Memorial Lecture. This lecture pays tribute to Prasanta Chandra Mahalanobis, an Indian scientist, statistician, and founder of the Indian Institute of Statistics [3]. The following is an excerpt from his slide:
Statistics =
Data Science? C. F. Jeff Wu University of
Michigan, Ann Arbor ·
What is
“Statistics”? ·
A
Statistical Trilogy ·
Frontier
and Beyond ·
A Bold
Proposal |
The current state of
statistical work can be described by a Statistical Trilogy: ·
Data Collection (experimental design, sample
surveys). ·
Data Modeling and Analysis. ·
Problem Understanding / Solving, Decision Making. |
Promising Current/Future Directions: ·
Large / complex data: Neural network models, data mining (of
massive data bases). ·
Empirical - Physical Approach: driven by data and
mechanistic knowledge, mechanistic: unknown state statistical: unknown state ·
Representation and Exploitation of Knowledge: Representation of
knowledge as a Bayesian prior model (possibly in
high-dimensional space) computational algorithm?interaction with
cognitive science. |
Why can neural network
modeling solve some complex / tough
problems? o
Can model complex (i.e., nonlinearity, interaction)
relationships. o
Use cross-validation and other statistical
techniques to find parsimonious models and gain predictive power. o
Good at developing simple and efficient computational algorithms, develop
problem-specific hardware.
Think Big, Learn
from Others! |
·
Tremendous progress has been made in image
reconstruction: Penalized maximum 1ikelihood, Bayesian Gibbs sampling. ·
Much less is known and much needs to be done in
computer vision: “Vision is a process
that produces from images of the external wor1d a description that
is useful to the viewers and not cluttered with irrelevant information
(Marr 1976)”. ·
Computer vision: An infusion of psychophysics,
neural physiology, statistics, engineering and artificial intelligence. |
Some
suggestions: ·
A balanced curriculum: more emphasis on data
collection, Scientific / mathematical basis for modelling,
computing for large / complex systems. ·
Interdisciplinary training: Requirement of a
cognitive minor, joint teaching by statisticians and scientists. ·
A radical idea: an applied master or doctoral
program with 30% - 50% courses
outside statistics. |
A proposal: “Statistics” ? “Data Science” “Statisticians” ? “Data Scientists” |
The authors use limited statistical knowledge to interpret the foresight of Dr. Chien-Fu Jeff Wu, 20 years ago. He foresees that if Statistics develop in the direction of the current traditional statistical trilogy, it will inevitably become one of the tools in other fields. Statistical trilogy: data collection, Data modeling and analysis and decision-making. Just as J. M. Juran advocated that there is a universal method for managing quality: quality planning, quality control, and quality improvement, which is called the Quality Trilogy. This view was put forward by J. M. Juran in 1986, and it was widely accepted by the industry and became the basis of Quality Management (Figure 1,2).
Figure 1: Statistical Trilogy.
Figure 2: Quality Trilogy.
In recent years, Data
Science has been a subject that uses data to learn knowledge. Its goal is to
produce data products by extracting valuable parts from data. It combines
theories and technologies in many fields, including Applied Mathematics,
Statistics, Pattern Recognition, Machine Learning, Data Visualization, Data
Warehousing, High-performance Computing, etc., and statistical technology would
be one of the most common knowledge. Data Science helps professionals and
non-professionals understand the problem by using a variety of relevant data.
Data Science and technology can help us to correctly process data and assist us
in research and investigation in the fields of Engineering Science, Biology,
Social Science, Anthropology, etc. In addition, Data Science is also of great
help to business competition. It is the same as Quality Management, the Quality
Management is related to the field of Science, Engineering, Management,
Sociology, Law, Education, etc. It is difficult to belong to one academic
discipline, which means that they are one of the general disciplines. Knowledge
is divided into professional knowledge and general knowledge of individual
courses. Professional knowledge needs to be proficient, and general knowledge
needs to be broad. Dr. Chien-Fu Jeff Wu suggested that in addition to the
balance of the statistical trilogy, the courses of the Institute of Statistics
need to increase cross-field courses, and even increase to 35-50%.
For the current /
future promising direction of Statistics, Dr. Chien-Fu Jeff Wu suggested this
as following:
·
Large
/ complex data.
·
Empirical
- Physical Approach.
·
Representation
and Exploitation of Knowledge.
·
Why
can neural network modelling solve some complex / tough problems?
·
Tremendous
progress has been made in image reconstruction.
·
Much
less is known and much needs to be done in computer vision.
Today, 20 years later, using Data Science as a keyword to search on Google website, whether it is in Chinese or English, there are millions of searches, a bunch of similar to Figure 3: Data Science pictures and article advertisements, especially this article: “Data Scientist: The Sexiest Job of the 21st Century”, I don’t know how many articles on newspapers and magazines have mentioned it. However, few people talk about the so-called new technologies such as Data Science, Big Data, Artificial Intelligence, etc., which are the inevitable products of the successful development and application of Information and Communication Technology after accumulaing the research of diverse knowledge in the long run. The following paragraph serves as a summary and echo of the teachings of Dr. Chien-Fu Jeff Wu (Figure 3).
Figure 3: Data Science pictures and article advertisements.
Scientific research is
an iterative learning process. Deduction and Induction are both methods we use
in inference. The purpose of statistical methods is to make this learning
process more efficient. Statistical inference is based on the results of
experiments or observed objective phenomena, that is, data, through estimation,
comparison, and prediction to draw conclusions that are closer to the truth.
This is the so-called induction method. The deductive method uses existing
conjectures, models, theories, that is, hypothesis, through a professional
deduction process that conforms to scientific logic, to deduce hypotheses that
are more connected to the truth. This iterative learning process makes
knowledge more and more complete, data generation and analysis of scientific
research [4] (Figure 4,5).
The field of Artificial Intelligence (AI) refers to the technology that artificially realizes the wisdom of human beings. However, there is no technology that can achieve the same level of human intelligence, and the vast majority of AI in the world can only solve a specific problem. From the perspective of the information science academia, the following novel scenario that is similar to Western magic and Eastern martial arts may be more convincing [5]. The field of AI can be said to be the holy grail of Information Science, or more like the Lord of the Rings. This imaginatively dazzling field has attracted generations of Information Science researchers to continue to invest in these difficult problems. I don't know whether it is because of courage or profit, Information Science researchers have been advancing courageously in this field. Many people have invested their entire lives without getting the results what they had imagined, and they have been killed without even entering the door. The theory in the field of AI has become a huge and monstrous creature, people who want to get started when they see such a huge creature, they often get lost before reaching the door, what about the others? After entering the door, there was no way out.
Figure 5: Data generation and analysis of scientific research.
Figure 6: The three waves of AI.
“Looking up is high,
drilling is strong, looking forward, suddenly behind.” this sentence is used to
describe Confucius, and I think it is more appropriate to describe “Artificial
Intelligence.” Often when a researcher thinks that he has already seen some
solution, in the end he only discovers that it turned out to be a dream. When
the problem seems so simple, in fact, there is a fairy or monster hiding in the
door, and looking at you with enthusiasm. They plan to lead you into this
field, after then killing you. People who lose their way in the field of AI are
usually not eaten by monsters, but bewildered by fairies with beauty. In the
end, they thought they were in heaven, but they went to hell without knowing
it. If you want to see the world where this fairy is, please follow me.
Scholars who study AI may be more like the characters on Xia Ke Island in Jin
Yong’s novels. After arriving at Xia Ke Island, they are fascinated by the
exquisite martial arts secrets and are no longer willing to return to the
original arena. They have studied martial arts throughout their lives, but is
in vain. The characters on Xia Ke Island can only see the words in the secret
script, but cannot see the iconic martial arts contained in the writing
strokes, so they miss the real exquisite martial arts. Perhaps, scholars who
study AI have the same problem. They are always obsessed with various seemingly
advanced algorithms, but they can't see the whole forest but trees only, so
that they can't see how the overall wisdom is formed. I am afraid this is a
researcher in the AI field, challenges that must be faced in the future. The
three waves of AI, so let the authors have a general outline of the development
of AI and realize the above-mentioned AI novel scenario [6] (Figure 6).
From the standpoint of
quality professional, after authors looking at Table 1: the methods used by AI,
we dare not intend to develop new technologies in the AI field or make any contribution,
because we know almost nothing; After we looking at Table 2: the related
application of AI, the application intention in the field of quality
professional is inspired, because we do not want to be absent from the wave of
AI [7] (Table 1,2).
Referring to quality related science research, technology development, and application promotion, in order to discuss the process of the development of modern quality management, we cite a paper, which is in line with Management Sciences, Figure 7 is an excerpt from the paper that discusses the relationships among the sciences, technologies and utilizations in Management Science.
Figure 7: relationships among sciences, technologies and utilizations in management.
The relationships among
are outlined in the following [8]:
·
Research
of body of knowledge;
·
Transfer
knowledge to be a state of the art;
·
Transfer
knowledge to be a state of the art;
·
Transfer
knowledge to be an economic usage, virtually nil;
·
Feedback
the gaps of body of knowledge;
·
Develop
a new state of the art;
·
Apply
a state of the art to economic usage;
·
Recognition
of areas for scientific activity;
·
Feedback
the gaps of body of technology;
· Improve the current application more efficiently (Figure 7).
Since the beginning of the last century, the development of the field of modern quality management has been going on for nearly a hundred years. It is also remarkably similar to the integration of sciences, technologies and utilizations in the development of “Management Science”. In the early days of the field of modern quality management, most professionals relied on the achievements of statistical science such as Control Chart, Sampling Acceptance and Orthogonal Experiments, and even developed into a branch of statistical science: Engineering and Industrial Statistics. Solving the quality problems, statistical methods and thinking way is one of the important tools, but also need to integrate the methodologies of essence of substance, process of business and psychology. After years of professional integration in different fields, the quality management is gradually formed a multi-value organization team work mode, which is continuous improvement, leadership, management by objectives, full participation, common language, technology sharing, problem solving, response to change, cultural heritage and people oriented. It is the integration of statistical science, management science, engineering science, system science, information science, psychology, etc., in order to improve the quality of human life. The development and promotion of the quality professional field is definitely not developed independently. It would be developed accompany with the state of the community, society, country, region, and global fields, such as the development of politics, economy, society, and technology. Especially the defence industry, international trade, management systems, business models, manufacturing systems, communications, computers, data processing, and people’s living standards. Nowadays, the quality professional field should also exam the future development and promotion projects in the so-called “The third wave of AI”, which something’s are worth to do?
Back to this topic: “The Applications of AI in Quality Technology”. In recent years, the authors have gradually established the “Quality Philosophy” and “Core Values”. Based on the “Quality Philosophy”, which is continuous improvement, leadership, management by objectives, full participation, common language, technology sharing, problem solving, response to change, cultural heritage and people oriented; based on the “Core Values”, which is practical application as the purpose, system integration as the means, pragmatic benefits as the inducement, and sustainable development for good fruit. The process of brewing this issue requires careful consideration and painstaking effort. Since “Industry 4.0” began to be popularly discussed in the industry, the authors can only focus on the cultivation of personnel in the quality professional field and have no countermeasures. Also only focus on productivity 1.0 ? productivity 2.0 ? productivity 3.0 ? productivity 4.0, which was in response to such global manufacturing trends, Taiwan Administration has initiated Productivity 4.0 which follows the historic long term plans of taking production automation as productivity 1.0, promoting to industrial automation as productivity 2.0, and expanding to industrial computerization as productivity 3.0, depict as Figure 8. Productivity 4.0 is based on intelligent automation and employing the internet of things, intelligent robots, and Big Data, coupled with lean management, will help domestic industrial upgrades and transformation (Figure 8).
Figure 8: The historic long term plans of Productivity 1.0 to Productivity 4.0.
Figure 9: Manufacturing management system hierarchy.
Figure 10: Smart factory operation schematic diagram.
Table 1: The methods used by AI.
Classification |
Methodology |
Search |
DFS, BFS, Best-FS, A*, Min-Max+?-? Cut, Dynamic Programming |
Optimization |
Greedy Algorithm, Simulate Annealing, Genetic
Algorithm |
Logical inference |
Boolean Logic, First-Order Logic, Probabilistic
Logic, Fuzzy Logic. |
Neural network |
Back Propagation Network, Hopfield Network |
Probability statistics |
Bayesian Network, Hidden Markov Model, EM Algorithm |
Comparison |
Pattern Matching, Regular Expression |
Table 2: The related application of AI.
Simulation
behavior |
Related
applications |
Perception |
Voice Recognition
(ears), Image Recognition (eyes), Handwriting Recognition (eyes), Fingerprint
Recognition (eyes) |
Reasoning |
Expert System,
Computer Games, Computer Chess, Medical Diagnosis (brain) |
Understanding |
Machine
Translation, Conversation System (brain) |
Learning |
Computer chess,
expert system, medical diagnosis, identification (brain) |
Action |
Robot Soccer
Game, Autonomous driving, Commercial robot, Smart Controller (hands, feet and
body) |
The author's view of
“Industry 4.0” have always been the following paragraphs as the guiding
principle: “From the perspective of micro-industry development, it will create
many kinds of innovative business model via customized design and marketing.
From supplier chain: purchasing, production controlling, incoming, production
and shipping to demand chain: ordering, logistics delivery, retail, and
maintain service, it can integrate the all processes to be a value chain
through computation, communication, controlling, collaboration and real time
response. This cross-company information system enables engineering
collaboration and logistics collaboration under the internet platform
architecture. Engineering collaboration: provide collaborative process of
two-way interaction between customers and suppliers, and customers can inquire
the engineering information needed from suppliers, to speed up customer quality
analysis, quality improvement and design improvement process, as well as
shorten the customer's time schedule from pilot run to mass production.
Logistics collaboration: provide customers and suppliers more transparent and
complete information interface, from the customer’s order to the production
schedule, from the production order to status of the lot number, from the
outgoing quality control to the shipping, customers can get the information
from the system and analyse in advance, solve the common problems of both sides
immediately.” The above situation is the vision of the industry 4.0 for
manufacturing operations management. Some advanced enterprises have already had
the ability to self-improve to this level, however, the SMEs are lack of this
kind ability. The MESA / ISA-95 standard have been defined the manufacturing
management system hierarchy [9]. In general, SMEs can achieve the vision of
Industry 4.0 operation by using Figure 10: Smart Factory Operation Schematic to
improve the quality of information of internal operations step by step (Figure
9,10).
According to the
author’s professional view, the development and promotion of the quality
technology also needs to be developed in accordance with necessity of each
enterprise, it does not need to be specific for AI. The development of AI
similar to Productivity 4.0 which follows the historic long term plans of
taking production automation as productivity 1.0, promoting to industrial
automation as productivity 2.0, expanding to industrial computerization as
productivity 3.0, and basing on intelligent automation and employing the
internet of things, intelligent robots, and Big Data, coupled with lean
management as productivity 4.0, it will help enterprise upgrades and
transformation step by step. The digital application progress model of the
enterprise must be optimized into: "From supplier chain: purchasing,
production controlling, incoming, production and shipping to demand chain:
ordering, logistics delivery, retail, and maintain service, it can integrate
the all processes to be a value chain through computation, communication,
controlling, collaboration and real time response." The quality
requirements of all processes of the value chain of productivity 4.0 would be
much more transparent. As the demand chain, product and service will be
required more accurate, speedy, reliable, safety, ecological, and
environmental. As the supplier chain, the product and service will be required
more easily to design, manufacture, change, transport, maintain, recycle and
trace. Therefore, quality practitioners should pay more attention to the
technologies of all processes of the value-added in the future, especially
system integration engineering. It focuses on system-value-added integration
between software and hardware. The process of value-added integration is
nothing more than system planning, system analysis, and system design, project
outsourcing and programming, unit testing, system and integration testing, user
acceptance testing, and system acceptance (Figure 11).
Therefore, quality
practitioners must focus on the development of relevant knowledge and
technology in the AI development process, and they must also recognize that the
development of AI in network information technology-related software and
hardware knowledge is formed through a market economy business model. It is
based on the market economy as its "core value": practical
application as the goal, system integration as the means, pragmatic benefits as
the inducement, and sustainable development of competition and cooperation as
the good result. Based on the above, the authors have established a curatorial
platform for innovation and quality management knowledge, to identify, create,
acquire, capture, apply, share, store, and map these documented information and
knowledge which can enhance quality of life, product quality, service quality,
environmental quality. In application of AI, the following are the knowledge
and technology which are the quality practitioners should have.
·
AI
contents and scientific research results understanding.
·
The
development and application of AI + IOT = AIOT.
·
Basic
knowledge of digitized documents and the Internet, and the ability to
communicate and operate.
·
Education
and training of basic statistics and probability.
·
Computer
calculation, analysis software, Big Data, and other software applications (such
as Excel, R language, Minitab, JMP...).
·
Organize
various levels of quality knowledge technology, such as CQT, CQE, CRE, CQM, and
must eliminate the waste and keep up with the times, and then increase and
modify to adapte "Industry 4.0" and "Artificial
Intelligence".
·
Project
planning and development methodology of manufacturing products.
·
Project
planning and development methodology of service industry products.
·
Application
software such as MES, ERP, PLM, CRM, SCM, KM, Promodel, etc.
·
Manufacturing
process rationalization, information process rationalization and system
integration technology.
·
Application
and promotion of Automated Optical Inspection (AOI).
Continuous improvement mechanism and promotion
knowledge (QCC, Six Sigma, Lean Production).