=Paper=
{{Paper
|id=Vol-2258/paper18
|storemode=property
|title=Methodological Basics of Creating Intelligent Quality Management Systems in Mechanical Engineering
|pdfUrl=https://ceur-ws.org/Vol-2258/paper18.pdf
|volume=Vol-2258
|authors=Georgy Burdo
}}
==Methodological Basics of Creating Intelligent Quality Management Systems in Mechanical Engineering==
https://ceur-ws.org/Vol-2258/paper18.pdf
Methodological basics of creating intelligent quality
management systems in mechanical engineering
G Burdo1
1
Department of Technology and Automation of Mechanical Engineering, Tver State Technical
University, A. Nikitin Emb., 22, Tver, 170026, Russia
Abstract. Quality management systems are becoming increasingly widespread in various
sectors of the Russian economy, in mechanical engineering in particular. Enterprises and
companies claim that they have these systems. In spite of this fact, in practice they often have a
limited set of certain procedures that describe a type of quality management actions at certain
stages of product creation and service performance. In this article, the author makes an attempt
to identify problem points that are connected to the development and implementation of these
systems, as well as to plan some ways to improve their effectiveness.
1. Introduction
The requirements of the modern market of engineering products are quite rigid. They make the
producers ensure the quality of products. At the same time, the issue of introducing products to the
market on time is always significant. It is known that one of the most effective levers for achieving
these goals is introduction of an automated product quality control system (QCS) in a company.
Therefore, it becomes clear that ensuring the effective functioning of these systems is of high priority.
There are many companies involved in development of quality management systems for
consumers. However, achieving effective functioning of QCS in practice is often impossible. This is
despite a large documentation package that regulates quality management procedures in companies
and available recommendations [1, 2, 5, 6, etc.].
We are going to try to understand the reasons for such a negative situation by answering the
following questions.
2. What should be automated quality control systems?
First of all, it should be noted that quality assessment procedures are almost an unformulated task.
Therefore, in our opinion it is necessary to provide the following points in an automated system:
a) use of an artificial intelligence device when forming knowledge bases and decision making;
b) ensuring the possibility of an automated system operation involving experts;
c) system openness for tuning to consumers;
d) a system might belong to a hybrid class with various types of artificial intelligence models.
The second important point is that the logic of system operation corresponds to a product life cycle.
However, in practice, it often happens that companies neglect an important stage of development of a
product technical specification. Therefore, QCS can only determine if a product measures up design
documentation that fixes a certain level of product quality. Nevertheless, the question is who
determines this level.
131
Hence, the need to include the stage of development of the technical task for a product into the
composition of QCS work stages becomes clear.
As an example, let us consider a QCS model developed according to this provision [7, 8, 18] as a
complex system with sub-systems R R11 , R12 ,..., R15 , R21 , R22 ,..., R212 (figure 1).
0
Figure 1 represents the QCS subsystem designated as R0 (subsystem of an upper level). Subsystems
of subsequent levels are described below.
Figure 1. A set-theoretical model of QCS.
1
Subsystem R1 performs quality control within technical specifications (TS) development phase. It
includes subsystem R21 that is quality control at the stage of technical specifications (TS)
development.
2
Subsystem R1 performs quality control within the design work (DW) phase with four life cycle
3
stages. It fulfills the following functions: R22 is quality control at the stage of research (R&D); R2 is
4
quality control at the stage of draft proposal (DP) development; R2 is quality control at the stage of
5
draft design (DD) development; R2 is quality control at the stage of preliminary design (PD)
development.
Subsystem R13 performs quality control within the work paper (WP) and maintenance
documentation (MD) development phase with two life cycle stages. It fulfills the following functions:
R26 is quality control at the stage of WP development; R27 is quality control at the stage of MD
development.
Subsystem R14 performs quality control within the product manufacture and testing phase (M&T)
with three life cycle stages. R28 is quality control at the stage of process engineering (PE); R29 is
quality control at the stage of product manufacture (PM); R210 is quality control at the stage of product
testing (PT).
132
5
Subsystem R1 performs quality control within the product maintenance and utilization (M&U)
11
phase, with two life cycle stages. It fulfils the following functions: R 2 is quality control at the stage
12
of product maintenance (Mc); R2 is quality control at the stage of product utilization (PU).
1
Control subsystem R 0 has six control functions. The first one is to control subsystem R1 , that is to
find out and adjust product quality parameters within TS phase: R01 : A b11 a11 , where A is a control
signal that is a set of requirements for product quality parameters at all the stages of PLC under the
1
enterprise management system; b1 is a set of actual product quality parameter values at the stage of
1
TS; a1 is a set of corrective actions on product quality parameters within TS phase.
The second function is to control subsystem R 21 , that is, to find out and adjust product quality
parameters within DW phase: R0 2 : A b12 a12 , where b12 is a set of product quality parameter values
within DW phase; a12 is a set of corrective actions on product quality parameters within DW phase.
1
The third function is to control subsystem R 3 , i.e. to find out and adjust product quality parameters
3
within WP and MD development phase: R03 : A b13 a13 , where b1 is a set of actual product quality
parameter values within WP and MD phase; a13 is a set of corrective actions on product quality
parameters within WP and MD development phase.
The fourth function is to control subsystem R41 , that is, to find out and adjust product quality
parameters within M&T phase: R0 4 : A b14 a14 , where b14 is a set of actual product quality parameter
values within M&T phase; a14 is a set of corrective actions on product quality parameters within M&T
phase.
The fifth function is to control subsystem R51 , that is, to find out and adjust product quality
parameters within M&U phase: R05 : A b15 a15 , where b15 is a set of actual product quality parameter
5
values within M&U phase; a1 is a set of corrective actions on product quality parameters within
M&U phase.
At last, the sixth function is to find out the product quality parameters in all the PLC phases:
R0 6 : A b11 b12 b13 b14 b15 B , where B is an output signal that is a set of end product quality
parameters.
This model can serve as a basis for development of design procedures in QCS.
The third aspect is related to the criteria for evaluating decisions at QCS work stages.
It seems that evaluation criteria can be quantitative (with boundary values), qualitative (like yes,
no) and qualitative with probabilistic nature (the probability of reaching a required level).
Therefore, there is a problem of bringing them to a single evaluation scale [7].
3. Why do we need quality control systems?
It seems a very simple question, but getting a correct answer is often very difficult. Let us try to figure
out why by answering the questions listed below.
A. Who should determine the quality of a product?
There is a common answer that it is a producer who determines it. Here you can object, “Does a
producer consumes his own products?” Obviously no. If we consider the question of quality even at
the level of common sense, it becomes clear that the one who buys products must determine their
quality.
133
Hence, there is a conclusion that an automated quality control system should provide the possibility
of analysing product quality required by consumers.
B. How to determine what a consumer wants?
This is the first difficulty for QCS implementation. There are many methods of market research
that have varying degree of uncertainty.
In our opinion, the most effective approach might be a preliminary market assessment using
marketing technologies included in the automated quality control system, and then production of trial
lots. It is clear that such methodology requires investment. Nevertheless, it can secure against big
losses when launching programs based on marketing research and advertising.
C. How a consumer mind forms a quality model?
It is very difficult to predict consumer needs without answering this question. In our opinion, the
following points determine this procedure.
First.
Unwillingness to perform labour-intensive actions that do not require intellectual expenses. For
example, a vacuum cleaner or a washing machine. This case is typical for an ordinary person.
Second.
A desire to increase productivity and improve labour quality or to make it easier. An example is the
appearance of more productive machines. This case is typical for a producer of goods and services.
Third.
Obtaining brand new opportunities for leisure or performing new functions by employees, etc., that
is, implementation of something that was impossible before.
The first three moments describe “non-imposed” models.
Fourth.
Advertising. In our opinion, the main thing here is not to overdo it, otherwise a consumer might
start to refuse an advertised model. It is also important that an advertisement recipient gets accustomed
to it. If he does not see a product or service in an advertisement, he begins to doubt its usefulness. On
the other hand, long-term advertising of the same leads a recipient to the idea “is it really good?”
Relying on this moment only does not allow long-term success, therefore the production of goods is
effective for a limited period of time, and does not guarantee their overproduction.
Fifth.
The ability to understand trends in the relevant production field. This approach is useful for high-
tech and highly organized industries with highly skilled administrative staff. Unfortunately, it is not so
frequent.
To sum up, it is possible to offer the following method of “soft” management of a consumer quality
model. Using the approach proposed in the answer to the question B, one should determine a product
quality model formed based on the first three moments. Then the prediction of this model trends
follows. Based on the forecast one can determine product modification and bring it to the market while
unobtrusively showing the advantages of new samples in advertisement.
Summarizing the above, we can conclude that QCS should include marketing technologies and an
apparatus that allows experts to build product quality models and predict their trends.
4. How to approach the development of a quality control system?
We can determine several most methodologically significant steps.
4.1. Determination of parameters that describe product quality in QCS
We assume that we have determined the product quality required by a consumer as shown above. In
the Japanese system of “lean production” this concept of quality is close to the concept of value.
It is necessary to formally describe this quality to use it in QCS. For this purpose, we need to find
out what are the most important product parameters that form its quality. After determining the
parameters, it is necessary to find out their allowed intervals.
This is a difficult task and is possible only for the most qualified specialists. Its difficulty is in the
fact that the parameters do not always determine quality in an explicit form; the parameters cannot
always be a numerical estimate [1, 2, 5, 6, 8].
134
4.2. Identifying actions that affect product quality
At the second stage, it is necessary to understand what actions in a company affect product quality.
This is necessary for understanding time points of a product life cycle in which its quality is
monitored.
The main mistake of system developers is to take into account in QCS only those actions that
explicitly form product quality (i.e., quality control systems exist on their own, as well as processes in
the organization). In industry, these are the processes of production preparation [7, 8, 9, 10, 11]. In our
opinion, if we talk about product quality, we should mean its entire life cycle. Here we come to
understanding the following facts:
1) product quality cannot be considered in isolation from all processes in a company;
2) quality control is a way of organization functioning;
3) organization activity consists of processes [3, 4, 10].
Therefore, QCS is not an extension to a company automated control system, but its essence.
4.3. Determining organization processes that affect product quality
The basis of any management system including a quality management system is company business
processes. Sometimes it is not entirely clear in practice what is the purpose for describing company
processes. In other words, “process control” is introduced.
When talking about introducing process control as an end in itself, one can assume only an attempt
at optimizing with not completely understandable criteria, ordering functions and eliminating their
duplication [11,12,15]. Such approach will not bring great benefits.
When implementing process management, it is necessary to understand what we want to manage.
Without such statement of the problem, everything will be reduced to some “improved”
administration. Therefore, in practice process management is often based on business activities
(accounting, warehouse, production, top management). This shows the problem related to the ability to
realize and represent company processes hierarchically linking them also in time.
It should be taken into account that processes are not required to copy a company administrative
structure. Moreover, different divisions may be involved in the same process.
Thus, another problem to be solved is the development of a hierarchical process system in QCS. It
implies inclusion of lower level processes in higher level ones, which agrees with the content
presented in [16, 17, 18].
4.4. Determining process performance criteria in QCS
A system of processes requires building a system of criteria at each hierarchy level aimed at ensuring
product quality [8]. These criteria should be based on the concept of product quality supplemented by
quality criteria of a particular organization process.
For example, the purchase of raw material. It is clear that the criterion connected with the concept
of product quality will be the criterion of raw material quality. However, the process of purchasing
raw material implies that it is important to specify exactly which raw material, which parameters, and
which supplier it should be (complementary quality criteria). A mandatory additional criterion should
be time and the accuracy of function execution of individual operations in the process. It is not an easy
task and requires a serious investigation of possible incorrect and inadequate actions.
The principles of forming criteria for evaluating solutions at product life cycle stages can be
formulated as follows:
a) the principle of a criteria boundary type;
b) the principle of a criteria relative type;
c) the principle of a criteria complexity;
d) the principle of clarifying criteria.
The principles are based on the fact that when creating new products at the level of the world's best
samples, the criteria should be modified in comparison with the criteria of modernization and
modification, which usually have minimax (type 1) or boundary (type 2) nature.
135
The principle a claims that type 1 criteria can form an insignificant part, since creating a new
product changes a large number of its principal parameters. Therefore, the most of criteria should be
boundary (no more, no less, in the interval) and be formed based on expert evaluation.
The principle b indicates that when evaluating decisions (performed work), relative criteria
obtained by comparing synthesized versions of technical systems or by comparing with a prototype
(prototypes) at the advanced world level become very important.
Let us consider some examples. For instance, cost supplement by 15% does not say anything as a
criterion. However, if the accuracy of measuring the parameter is increased by a factor of 2 in this
case, then the parameter is acceptable.
The principle c states that the overall product evaluation must be carried out by complex criteria,
criteria derivatives applied at each stage. For example, increasing product complexity, its design
hierarchy is not the best case if we take into account design only.
Nevertheless, it can allow keeping to the principle of building block design while simplifying
planning an industrial output, to simplify and reduce the cost of its assembly and maintenance.
The principle d determines the need for gradual criteria specification. Product design is not
completely specific at the initial design stages. Therefore, expert criteria do not always have a
quantitative form; they can be represented by expressions of fuzzy logic of the form “basically
fulfilled requirements”, “doubtfully fulfilled requirements”, etc.
Detailing design documentation provides the transition to numerical values of criteria.
4.5. Determining a person responsible for a particular process in QCS
The question who is responsible for a particular process is important. It is difficult since various
company divisions participating in it already have their leaders. Usually, the most competent division
head (who understands the essence of the process) should lead the process. The task of other division
heads is to distribute executives properly.
Here we get a logical, not always accepted though, distribution of leadership by a functional and
administrative feature.
The leader of a lower-level process is functionally subordinated to the head of the process which
includes the given one.
4.6. How to make a final decision correctly
We consider that for this purpose, all processes in the organization should be divided into two groups:
1. processes that directly affect the quality of goods;
2. processes that do not directly affect the quality of goods, but are necessary for implementing
first group processes.
The first group processes in production, for example, include the processes of design preparation,
technological preparation, manufacturing, testing, etc., i.e. the processes that are a part of the concept
of a product life cycle.
The processes of the second group are transportation, document management, financial
management, formation of contracts, etc.
Therefore, the most worthwhile way of organizing process quality assessment is the following.
The processes of the first group are the most responsible. Therefore, for them it is necessary to
introduce understandable scales translating qualitative estimates into quantitative ones by attracting
experts. For each process, it is necessary to define a minimal assessment boundary, below which the
process must be repeated anew.
If we have estimates for each process related to product life cycle stages or steps, it is necessary to
obtain a weighted (integrated) estimate of a process set that directly affect product quality.
Obviously, expert systems are necessary here since organization representatives are the experts.
For the processes of the second group, it is sufficient if a division head controls functions of a
particular process within his division based on additional criteria.
Let us consider a method of decision-making based on fuzzy sets as an example [13]. A team of
experts, who are company representatives, assesses possible control actions using the following
method.
136
1. Experts are assigned with weights (their sum is equal to 1) taking into account: a) a degree of
responsibility for a decision; b) a qualification in a solution domain.
2. Experts determine a permissible lower limit of their general confidence in the effectiveness of a
control action aimed at ensuring product quality (recommended based on professional experience ≥
0.7).
3. Experts determine a periodicity of adjustments (it is recommended as private assessments in
QCS appear).
4. Each expert discusses and estimates an option of control impact (confidence in achieving a
result by implementing the proposed control action) in numerical form (from 0 to 1).
5. Getting a weighted estimate of experts’ general confidence in achieving product quality by
implementing the discussed action.
6. A person responsible for making a decision makes a decision (to implement a particular
decision or not).
7. Go to the next assessment point. Assessment points correspond to product life cycle stages and
steps.
Another issue is a way to determine weighted confidence of all experts. For this purpose, a
knowledge base based on fuzzy sets has been developed using expert estimates [13].
The rules have the following form:
(IF confidence of the expert 1 with weight P1 is equal to A1 ) AND (IF confidence of the expert 2
with weight P2 is equal to A2 ) AND ... (IF confidence of the expert n with weight Pn is equal to An )
THEN (as a result, weighted confidence of the experts is equal to Bi ).
The linguistic variable Ai has 3 terms: sufficient, mean, insufficient as shown in figure 2.
Figure 2. The membership function for the linguistic variable Ai .
The expert assessments helped to form the following fuzzy production rules that determine the
value of the linguistic variable Bi (weighted confidence of experts). Preliminary in the production
rule, there are mutually exclusive subconditions containing opposite values of the linguistic variable
Ai (sufficient insufficient). We can see that the result is production rules that contain only
adjacent values of the linguistic variable Ai .
1. IF (all subconditions contain the value of the linguistic variable Ai – sufficient) THEN (the value of
the linguistic variable Bi is very reliable).
137
2. IF (all subconditions contain the value of the linguistic variable Ai – insufficient) THEN (the value
of the linguistic variable Bi is very unreliable).
3. IF (the number of subconditions with the value of the linguistic variable Ai – average and exceeds
the number of subconditions with the value of the linguistic variable Ai – sufficient or insufficient) OR
(no more than one less than the number of subconditions with the value of the linguistic variable Ai –
sufficient or insufficient) THEN (linguistic variable Bi is doubtful ).
The linguistic variable Bi has 5 terms: most reliable, reliable, doubtful, unreliable, and quite
unreliable (figure 3).
Figure 3. The membership function for the linguistic variable Bi .
4. IF (the number of subconditions with the value of the linguistic variable Ai – sufficient or
insufficient more than one exceeds the number of subconditions with the value of the linguistic
variable Ai – mean) THEN (the linguistic variable Bi is reliable or unreliable, respectively).
For example: (IF confidence of the expert 1 is sufficient AND confidence of the expert 3 is
sufficient AND confidence of the expert 4 is average) THEN the value of the linguistic variable Bi is
reliable.
The phasilogical conjunctions of subconditions are carried out according to the minimum rule.
Finding joint membership functions is usually according to Mamdani’s fuzzy implication.
The defuzzification of a weighted assessment of experts' confidence (result) is performed by the
centroid method.
4.7. What performers are needed when implementing quality management systems
The success of a quality management system is mostly determined by employees engaged in the
activities.
It is also important that the management is ready to implement quality management systems, that
is:
1) understanding that quality cannot be cheap, that it is necessary to invest in it;
2) it is necessary to invest in technical re-equipment of production;
3) it is necessary to invest in staff development.
138
The most difficult task is to make all process participants to understand a sense of corporate
responsibility for the result. It is important to convince everyone involved in the process that the
successful functioning of a company in general is the guarantee of well-being of each executive.
Experience shows that it is difficult to avoid forming an appropriate corporate culture and using
methods for stimulating employees’ activities, namely incentives, and not anti-incentives
(punishment).
Thus, it becomes clear that the development and implementation of effective automated quality
management systems makes it necessary to take a fresh look at company activities, to rethink the usual
activity rules.
Practical implementation of decision-making procedures in machine-building production systems is
given in [7, 8, 18].
5. Conclusion
The authors consider that the main conclusion is that quality management is a way of company
functioning. Therefore, all the processes of company activities must be subordinated to the goals of
this activity method. In other words, all processes of company activities should be built into a quality
control process.
The opponents might say that the article is partly polemical and many statements can be
objectionable. We can agree with this point of view. The paper has a conceptual nature. However, in
the opinion of the author, the main problem points are indicated.
Therefore, the authors deliberately do not consider the problems of optimization of a company
functional structure, its divisions, stages of work on introducing quality management systems, etc.
It is also necessary to mention that the introduction of a quality management system will seriously
affect not only mid-level management units, but also the activities of top management. It will also
cause silent protest and rejection of some employees.
Therefore, the effective implementation of activities related to quality control depends on the will
of top management.
The proposed approach to the creation of an automated quality management system is implemented
at a machine-building enterprise; the results of implementation and testing are adequate to the stated.
In the near future, it will be possible to summarize them and present them in the form of publications.
6. References
[1] Watson G H, Conti T and Kondö Y 2003 Quality into the 21st Century: Perspectives on Quality
and Competitiveness for Sustained Performance (ASQ Quality Press) p 261
[2] Glichev A V 2001 Basics of Product Quality Management 2nd ed (Moscow: RIA Standards and
Quality) p 420
[3] Andersen B 1999 Business Process Improvement Toolbox (Asq Pr) p 233
[4] Repin V V and Eliferov V G 2004 Process Approach to Management (Series Practical
Management) (Moscow: RIA Standards and quality) p 380
[5] Gorlenko O A and Miroshnikov V V 2002 Creation of Quality Management Systems in an
Organization (Moscow: Mashinostroenie-1) p 124
[6] Sharipov S V and Tolstova Yu V 2004 Quality Management System (St. Petersburg: Piter) p
192
[7] Burdo G B, Semenov N A, Vinogradov G P and Sorokin A Yu 2016 Decision assessment
principles in the automated system of engineering product quality control Software and Systems
2(114) pp 113–118
[8] Burdo G B, Semenov N A and Sorokin A Yu 2013 An automated quality management system
model in multi-nomenclature machine building production Software and Systems 4 pp 46–59
[9] Allan J J 1990 A Survey of Commercial Turnkey CAD/CAM Systems 2nd ed (Dallas, Tex.:
Productivity Int., Inc.)
[10] Amirouche M L 1993 Computer-Aided Design and Manufacturing (Englewood Cliffs, NJ: Co
Prentice-Hall)
[11] Womack J P and Jones D T 2003 Lean Thinking 2nd ed (NY: Free Press)
139
[12] Shingo Sh 2007 Kaizen and The Art of Creative Thinking (Enna Product Corp. and PCS Inc.)
[13] Zadeh L 1965 Fuzzy sets Information and Control 8 pp 338–353
[14] Ackoff R 2003 Ackoff's Best: His Classic Writings on Management (Wiley) p 356
[15] Rampersad H 2001 Total Quality Management 1st ed (NY: Springer-Verlag) p 179
[16] Oakland J S 2004 Oakland on Quality Management (Elsevier Butterworth-Heinemann) p 487
[17] Sakata Sh 2016 Practical Guide to Quality Management (Moscow: Mechanical Engineering) p
215
[18] Burdo G B, Palukh B V and Semenov N A 2016 Decision-making support when dispatching
technological processes in machine building. Proc. 15th National Conf. on Artificial Intelligence with
International Participation (Smolensk) vol 1 (RAAI) pp 231–238
Acknowledgments
This work has been supported by the Russian Foundation for Basic Research, project 17-01-00566.
140