New Fuzzy Model for quality evaluation of E-Training of CNC Operators

International Journal of Engineering Science Invention
ISSN (Online): 2319 – 6734, ISSN (Print): 2319 – 6726
www.ijesi.org ||Volume 5 Issue 6|| June 2016 || PP. 48-60
www.ijesi.org 48 | Page
New Fuzzy Model for quality evaluation of E-Training of CNC
Operators
Zoran Milojevic1
, Zarko Milojevic2
1
(Machine Engineering, Faculty of Engineering/University Kragujevac, Serbia)
2
(Economic History, Faculty of Economics/University Hohenheim, Germany)
Abstract: The quality of e-learning is a very important issue, especially when production technologies are
concerned. This paper introduces a new fuzzy model for e-learning quality evaluation. All uncertainties and
consequent imprecision are modeled by triangular fuzzy numbers. The quality of CNC e-learning process is
determined by using the fuzzy logic IF-THEN rules. The proposed method derives an aggregated satisfaction
value both for the participants as well as the trainers.The authors introduce a genuine metric interval for the
objective evaluation of E-learning effect. The OLS regression model estimates the magnitude and polarity of E-
learning effect on participants` perception of the training quality. The predicted coefficient of E-learning effect
on the overall quality of CNC training is estimated to be14.88 measurement points with a negative impact on
overall satisfaction. These novel findings shed a new light on the quantitative effect of E-learning on CNC
machine training and contribute to the contemporary scientific literature within the research area.The
developed model is illustrated by real-life data from secondary technological schools from central Serbia.
Keywords: CNC technologies, E-learning effect, E-training process, Evaluation, Fuzzy sets, Production
technologies,
I. Introduction
Education and creation of human resources is an important part of national strategies of developed countries,
especially having in mind the fact that knowledge productivity has a direct influence upon economic growth of a
country. Continual social and economic changes and an increased scientific and technological development,
especially IT development, place a demand for highly educated workforce that would be able to efficiently solve
various problems. In such a way, developed societies become learning societies; apart from a well-structured
formal educational system, they have an opportunity for continual education and training due to a suitable system
of institutions, organizational forms and programs which are not a part of formal education.
The issue of permanent education in the field of modern production technologies, such as CNC technologies
whose application in production processes is immense with an ever growing tendency, is of utmost importance in
developing countries. Venture costs rise dramatically unless there is a suitable number of educated people with
specific knowledge who can promptly enter production processes. Otherwise, the whole venture becomes rather
expensive and risky. Thus, for developing countries, and Serbia as one of these, it becomes of utmost importance
to reform and advance the educational system as far as this field is concerned.
The quality of CNC education has a direct influence upon productivity of workers participating in processes.A
number of papers presented in literature deal with the considered problem. Multimedia technology and Internet
applications are important tools for teaching and learning in engineering education. Nowadays online simulative
environments are indispensable tools for teaching computer numerical control (CNC) course in mechanical
engineering. Previous studies were seldom designed in such a way to consider new emerging innovations and
applications when CNC technologies are concerned.
Some authors [1], [2], [3], are engaged in engineering education and use the concept of Web services and clarify
the advantages of implementing simulation in testing written programs for the CNC machine with the use of
computers which certainly helps trainers/ lecturers to efficiently implement the process of learning to improve
skills of participants.Forinstance, M.Exel et al. [4] write about two web-based training approaches forremote
control.In [5], the author attempts to propose a virtual operating system applied to operation training of
manufacturing facility and manufacturing process simulation.
Virtual reality (VR) technology has many advantages such as excellent interaction, high precision in 3D display
and offers opportunities to manufacturing firms to meet the above challenges. VR technology has been widely
used in manufacture simulation, especially, the web-based virtual machining simulation system that has become
popular due to its cooperation facility and good platform independence [6-10].Discussion regarding training
within the field of mechanical engineering (especially training conducted on CNC machines), with the emphasis
on the application of Internet, E-learning andWeb-based training, and including manners to overcome the
difficulties in implementation of new technologies (CNC, ICT, Internet, E-learning, Web) is presented in 11.

New Fuzzy Model Forqualityevaluationofe-Training of CNC Operators
E-learning education quality is a complex notion which could be estimated according to multiple criteria.The
authors hypothesize that identified criteria do not have relative importance and values. The relative importance
of identified criteria and their values are difficult or rather impropriate to quantify, and as such have to be
thoroughly assessed by decision makers.
A multi-criteria methodology defined from the perspective of learner satisfaction to support evaluation-based
activities taking place at the pre- and post-adoption phases of the web-based e-learning education system
(WELS) life cycle is proposed in 12. Theassessment of relative importance of identified criteria is performed
by investigated learners’ perceptions. This investigation carried out a survey of college students, and the data
obtained is then analyzed by analytic hierarchy process in order to derive an integrated preference structure of
learners as a ground for evaluation. It isfound that learners regarded the learner interface as being the most
important dimension of decision criteria. This allows a higher level of e-learner satisfaction to be reached, and in
the process increases the level of system acceptance and continued use.
Decision makers express their judgments far better by using linguistic expressions than by representing them in
terms of precise numbers.The Fuzzy Set Theory should be employed in the case of scarce evidence data or if the
data values are rapidly modified due to the changes in the environment, according to Zimmermann, 1978. On
the other side, the decision makers define their estimations faster and easier when applying linguistic
expressions compared to exact numbers. Taking this into account it could be said that the Fuzzy Set Theory
represents an adequate tool for handling with uncertainty and imprecision of any kind while simultaneously
offering a possibility to overcome the obstacles regarding imprecision, number of data and the uncertainty
among the decision makers.
In this paper all subjective assessments are described by linguistic expressions which are then modeled by fuzzy
sets, as it has been used in13-14. Fuzzy set theory can provide a valuable framework for handling imprecise
and ambiguous data, and provides reasoning and decision making methods based on such data. It can be said
that by applying the fuzzy approach in the modeling of linguistic expressions, all uncertainties and imprecision
which emerged due to lack of good evidence are eliminated.
In 15 a fuzzy multi-criteria method for e-E-learning quality evaluation has been developed. E-learning training
process is decomposed into sub-processes. The existing uncertainties in the relative importance of sub-processes
and their corresponding values are displayed by discrete fuzzy numbers. However, the values of identified sub-
processes can be crisp and uncertain. The normalization of values of identified crisp sub-processes is given by
using linear normalization procedure 16. The transformation of all the uncertain sub-processes into degrees of
belief is performed by applying a fuzzy set comparison method 17. Determining the weighted normalized
values of sub-processes is represented as a group decision making problem. The aggregation of individual
opinions into group consensus is performed by usingthe averaging method. The overall value of e-learning
process quality is calculated by using the method proposed in 15.
An assessment of E-learning laboratory objectives for control engineering education is considered in 18. The
relative importance of laboratory objectives is described by triangular fuzzy numbers(TFNs) and calculated by
the proposed fuzzy Delphi method. The values of treated objectives are modeled by trapezoidal fuzzynumbers
(TrFNs). The elements of fuzzy decision matrix can be computed as product of the relative importance of
laboratory objectives and their normalized values. The best laboratory with respect to all objectives is found by
applyingthe comparison method, i.e.TFNs19.
In this paper, the relative importance of identified criteria is defined according to AHP framework, as presented
in20-22. It is easier and closer to human reasoning to evaluate the relative importance of each pair of attributes
instead of making a direct evaluation. The assessment of the relative importance is denoted as a fuzzy group
decision making problem. In general, the aggregation of individual opinions into group consensus may be
determined by applying different operators. Another point of view would be that trainers have the equal
relevance, so that consensus may be given by fuzzy logic method (by analogy 23). The weights of the
considered criteria are computed by using the concept of extent analysis 24 which is extensively implemented
approach in the literature. The obtained criteria weights, in this case however, are not the fuzzy numbers.
Within the approach suggested in paper, the estimation of criteria values is based on the subjective assessment
of a trainer who conducts an e-learning training, and of trainees, separately. Furthermore, the isolated effect of
e-learning based CNC machine training is calculated by employing a multivariate regression model.
Contribution of this paper compared with other cited works, mirrors itself in a new method definition that
undoubtedly allows for possibility to determine the values of particular described criteria and sub-criteria. In
addition, it allows for an influence on those (sub)criteria that have the minimal values so that we reach the
improvement of total quality assessment of the CNC training both in classical approach and the e-learning
approach.
A new fuzzy model for assessing CNC e-learningtraining is therefore proposed.All existing uncertainties are
described by pre-defined linguistic expressions which are modeled by triangular fuzzy numbers (TFNs). The
scope of the treated problem is as follows: (1) modeling of the assessments of trainersandparticipants is

performed by using triangular fuzzy numbers (TFNs);(2) the weights vector of criteria are obtained by fuzzy
Analytic Hierarchical Process (FAHP); (3) a new fuzzy model for assessing CNC e-learning training is
proposed;(4) based on the obtained resultstrainers can define priority of management initiativesthat should lead
to the improvement of the e-learning education quality. This may be seen as crucial in a time of economic crisis
because the level of e-learning quality has to absorb the magnitude of the undesirable event and enhance long
term sustainability of the education systems. It can be recognized as the main advantage of the proposed fuzzy
model.
The paper is organized as follows: the basis of the considered problem is given in Section 2, including a new
fuzzy set based approach to model the uncertainty in criteria weights and criteria values according to which the
evaluation of e-learning education is performed. Section 3 describes a new fuzzy multi-criteria method for
calculating e-learning education estimation. The proposed algorithm is illustrated by an example in section 4.
Section 5 provides an interpretation of the generated results followed by an OLS regression model. The
conclusions are presented in Section 6.
II. Modeling Of Uncertainties
In this Section, modeling of uncertainties in the relative importance of criteria is described. Uncertainties are
described by the linguistic expression modeled by fuzzy sets, as discussed in14. A fuzzy set is represented by
its membership function. The parameters of a membership function are the shape, granularity and location in the
universe of discourse. Determining of the membership function can be defined as a task by itself. The
membership function of a fuzzy set can be obtained based on one’sexperience, subjective belief of decision
makers, intuition and contextual knowledge about the concept modeled 25. Pedrycz and Gomide 26 defined
six ways for determining the membership functions. However, subjectivity in determining the membership
function has been considered as the weakest point in the fuzzy sets theory.
The triangular fuzzy numbers (TFNs) are widely used since they offer a good compromise between descriptive
power and computational simplicity. Triangular function is less tolerant because the range of maximum
membership has to be around the crisp point of the triangle. Fuzzy sets of higher types and levels have not as yet
played a significant role in the applications of the fuzzy sets theory 27.
Granularity is defined as the number of fuzzy numbers assigned to the relative importance and values of
identified criteria. Lootsma 28 suggests that human being can have only seven categories at the most. In this
paper, respecting the type and size of the considered problem and Lootsma's suggestion authors used five
linguistic expressions assigned to the existing linguistic variables.
In the related literature there is no strict guideline how to determine domain of any TFN. In general, domains of
TFNs are defined on the real sets into different interval (for an instance:0-1, 1-5, 1-9, etc.). Also, the way
how to determine lower bound, upper bound and modal value of any TFN is not defined in the literature 29-30.
TFNs may be presented symmetrically on the proposed scale or they may be presented in compliance with
decision makers’ experience and knowledge.
This paper offers a problem setup for CNC (Computer Numerical Control) training quality evaluation. A
training conducted by using e-learning approach based on multi-criteria; and a method of CNC training quality
evaluation based on e-learning. The uncertainties detected in the proposed model can be expressed as: 1) the
relative importance of sub-criteria of training evaluation, 2) the values of sub-criteria prescribed by Fuzzy Sets
Theory.
2.1 Basic assumptions
The paper further explains the problem of defining CNC e-learning education quality. Solution of the considered
problem directly influences the effectiveness of management in manufacturing and production system, with
special focus on productivity improvement.
Relative importance of the considered criteria and their values are described by different linguistic variables
which are afterwards modeled by triangular fuzzy numbers.
The following are the assumptions that underline a model of the considered problem:
 Trainers define the criteria that are used for the evaluation of CNC e-learning education. Trainers define the
number and types of criteria based on their knowledge, skills and experience. To each defined criterion an
organized pair (relative importance, value joined) is associated.
 Relative importance of treated criteria is hardly changed over time. Generally, the relative importance of
criteria is different and determined according to knowledge and experience of trainers. It can be stated as
fuzzy group decision making problem.
 Values of defined criteria are determined for each participant, separately, by trainers. In the considered
problem, these values are uncertain.
 All uncertain values are modeled by applying the fuzzy set theory.

2.2Criteria weights modeling
All criteria for evaluating quality of e-learning education are usually not of the same relative importance. Also,
they can be taken as unchangeable during the considered period of time. They involve a high degree of
subjective judgments and individual preferences of trainers. The judgment of each pair of treated criteria suits a
human-decision nature (by analogy with AHP method) the best. In conventional AHP, the pairwise comparison
is established using a standard integer scale. The use of discrete scale of AHP is simple and easy, but it is not
sufficient to take into account the uncertainty associated with the mapping of one’s perception to a number.
Trainers express their judgments far better by using linguistic expressions than by representing them in terms of
precise numbers.
The following notation is used in the paper:
T- number of considered trainers
t- index of trainer
P- number of treated participants
p- index of participant
K number of the identified criteria
k- index of criterion
Wkk′
t
-triangular fuzzy number describing the local importance of criterion k over a criterion k′
, k, k′
=
1, . . , K; kk′
which is given by trainer t, t=1,..,T
Wkk′-triangular fuzzy number describing the aggregated relative importance of criterion k over a criterion k′
,
k, k′
= 1, . . , K; kk′
wk -crisp value describing the criterion weight k, k=1,..,K
~ t
kpv
.
-triangular fuzzy number describing the criterion value k, k=1,..,K at the level trainer t, t=1,..T,
which is estimated by participant p, p=1,..,P
k
~
d -triangular fuzzy number describing the weighted value of criterion k, k=1,..,K respecting the all participants
c-fuzzy number describing the satisfaction level of participants with respect to all criteria and their weights
c-the representative scalar of the fuzzy numberc
In this paper, the trainers' fuzzy rating for each pair of treated criteria is described by linguistic expressions
which can be represented as in Equation (1):
TFN  t
kk
t
kk
t
kk
t
kk
~
'''
'
u,m,l;xW  (1)
with the lower and upper bounds
t
kk
t
kk
'' u,l and modal value
t
kk
'm , respectively. Values in the domain of these
TFNs belong to a real set within the interval from 1 to 5.
If strong relative importance of criterion '
k over criterion k holds, then pair-wise comparison scale can be
represented by Equation (2):
TFN



















t
kk
t
kk
t
kk
t
kk
t
kk
lmu
WW
'''
''
1
,
1
,
1
1
~~
. (2)
If '
kk  ( K,..,1
'
k,k  ), then relative importance of criterion k over criterion '
k is represented by single point
1 which is TFN (1,1,1).
In this paper, the fuzzy rating of each decision maker can be described by using three linguistic expressionsas
following:
low importance-  5,1,1;x1
~
R 
medium importance-  5,3,1;x2
~
R 
most importance-  5,5,1;x3
~
R 
Calculation of the Criteria Priority Weights is presented as follows.
The value of fuzzy synthetic extent with respect to the i-th object is defined as in Equation (3):
1
K
1k
K
1k
K
1k
k
k
k
kk
~
NNS

  
  









(3)

where:








 

K
1k
kk
K
1k
kk
K
1k
kk
K
1k
k
k
'
'
'
'
'
' u,m,lN
(4)
    
    









K
1k
K
1k
K
1k
K
1k
kk
K
1k
K
1k
kk
K
1k
kk
K
1k
k
k
'
'
'
'
'
' u,m,lN
(5)
Nk
kK
k=1
K
k=1
−1
=
1
ukk′
K
k=1
K
k=1
,
1
mkk′
K
k=1
K
k=1
,
1
lkk′
K
k=1
K
k=1
(6)
The value of fuzzy synthetic extent with respect to the k-th criterion is defined by Equation (7):
Sk = lkk′
K
k=1 , mkk′,
K
k=1 ukk′
K
k=1 ∙
1
ukk′
K
k=1
K
k=1
,
1
mkk′
K
k=1
K
k=1
,
1
lkk′
K
k=1
K
k=1
, (7)
The weights vector is represented as in Equation (8):












































 Kkp
SBelSBelSBelW
~~
1
~
,...,,...,
(8)
Where 





k
~
SBel is measure of belief according to which TFB k
~
S is bigger than all other TFBs '
k
~
S ,
)kk;K,..,1k,k(
''
 . This value is obtained by applying the method for fuzzy numbers comparison 31-32.
After normalizing pW , we get the normalized weights vector W as in Equation (9):
 Kk
wwwW ,...,,..,1

(9)
W is a non-fuzzy number and this gives the priority weights of one criterion over the other.
2.3Modeling of uncertain criteria
Values of all defined criteria underlying the estimation of e-learning education quality cannot be stated
quantitatively, as both trainers and trainees most often base their estimates on evidence data, knowledge,
experiences, etc. In such cases, their values are adequately described by five linguistic expressions: verylow
value, low value, moderate value, high value, and very high value. These linguistic expressions are modeled by
TFNswhich domains belong to a real set within the interval from 1 to 9 (by analogy with the Saaty's
measurement scale). These TFNs are given in the following way:
very low value - 1
~
V  2.0,0,0;y
low value- 2
~
V  45.0,3.0,15.0;y
moderate value - 3
~
V  65.0,5.0,35.0;y
high value - 4
~
V  85.0,7.0,55.0.;y
very high value - 5
~
V  1,1,75.0;y
2.4 Modeling of the regions of education satisfaction
Trainer and participant may have different levels of education satisfaction.This paper proposes, three different
state regions of education satisfactionlow education satisfaction region, mid-point education satisfaction
region, and high education satisfaction region. The proposed regions represent the baseline for education
satisfaction assessment. The regions of education satisfactioncan be modeled by one of the three predetermined
linguistic terms. These linguistic expressions are modeled by TFNs.
The TFNs for modeling the regions of education satisfaction are:

 low education satisfaction region -  55.0,0,0;xS1
~

 mid-point education satisfaction region -  85.0,5.0,15.0;xS 2
~

 high education satisfaction region -  1,1,45.0;xS 3
~

The proposed domains of the defined regionsare given with respect to pre-defined values of considered
uncertainties. These values may be changed and adjusted according to the specific needs of the treated education
organizations.
III. The Proposedfuzzy Modelfor Quality Evaluation Of E-Training Of Cnc Operators
In this paper, e-learning education quality is determined in an exact way by applying the proposed fuzzy model
which is realized in several stages.
The proposed procedure can be accomplished through the following steps. A general design of the quality
assessment is given in Fig. 1.
Figure 1-Overall quality assessment of a performed training
Step 1. The fuzzy pair-wise comparison matrix of the relative importance of evaluation criteria is stated:
T,..,1t;kk;K,...,1k,kW
''
KxK
t
kk
~
'









Step 2. Calculate the aggregated assessments of trainers,   ''
kkkkkkkk
~
kk;K,..,1k,k,u,m,lW '''
'
 by
using the procedure proposedin 23, such that we have Equation (10):
t
kk
T,..,1t
kk '' lminl

 , 


T
1t
t
kkkk '
' m
T
1
m , and
t
kk
T,..,1t
kk '' umaxu


(10)

The aggregated fuzzy pair-wise comparison matrix of the relative importance of the treated criteria can be
presented as:
''
KxK
kk
~
kk;K,...,1k,kW '









Step 3. Calculate the criteria priority weights,  Kk1 w,...,w,..,wW  by using the concept of extent analysis
24.
Step 4. Assessment of participant p, p=1,..,P by trainer t, t=1,..,T under criterion k, k=1,..,K is described by
linguistic expressions which are modeled by TFNs,
t
kp
~
v .
Step 5. Construction of weighted aggregated fuzzy decision matrices is shown as:
xK1
k
~
d 





Where in Equation (11):
 
 

P
1p
T
1t
t
kp
~
kk
~
Vw
P
1
T
1
d
(11)
Step 6. Calculate the overall satisfaction of participants,
~
c as in Equation (12):
k
~
K,..,1k
~
dc


(12)
Step 7. Find the representative scalar of the fuzzy numbers
~
c by using the moment method used in14.
Step 8. The region of education satisfaction in the observed trainers and participants can be defined according to
the rule:
IF the value of "overall education satisfaction" equals c, THEN the region of education satisfaction is described
by the linguistic expression where the level of education of participants is given in Equation (13):
 
*
q
~
q
~
SS3,2,1q
cxmax 

.
(13)
IV. Case Study
The participants’ satisfaction with the CNC e-learning education for secondary technical school teachers is
realized as a part of continual education, which is one of requirements of Total Quality Management principles.
A series of the same content trainings were organized during one year, including two major processes: process
of programming and handling process.The considered education process includes 32 participants (High school
lecturers that teach withinthe CNC field and engineers from industrial enterprises). The descriptive statistics of
the aggregate sample – 91 participant is presented in Table 1. Average education level among the participants is
almost 16 years1
, i.e. the majority has a B. Sc. degree. The experience on average is 10.5 years, where some
have only 6 years of professional employment. Itis realized by two trainers. The quality of the treated education
process is evaluated by using satisfaction of the participants. The satisfaction level of the participants is
calculated by using the proposed fuzzy model. Three criteria are considered: quality of the training content
(k=1); quality of training execution (k=2) and equipment quality (k=3).
Table I: Descriptive statistics of the total sample (original sample values given in parentheses)
E-learning Education Experience Grade
Mean 0.13747 (0.13448) 15.84 (15.98) 9.83 (10.27) 4.05 (3. 93103)
Maximum 0.2 18 22 5
Minimum 0.1 14.5 6 1
Range 0.1 3.5 16 4
Stand. 0.02869 (0.02934) 0.60411 3.455314 1.03669
1
Described values in this paragraph relate to the original sample size, excluding those candidates assigned to the
area “unsuccessful”.

Deviation (0.72558) (3.43195) (1.22273)
Variance 0.00082 (0.00086) 0.36495
(0.52647)
11.93919
(11.7783)
1.074725
(1.49507)
The proposed fuzzy model for the quality evaluation is applied in following steps:
Step 1
The relative importance of criteria values estimated by two trainers is presented in the followingfuzzy pair-wise
matrix:












~
1
~
2
~
2
~
3
~
1
~
2
~~
1
~
2
~
2
~
3
~
1
~
2
~~
1/1,,
,/11,
/1,/1/1,/11
RRRR
RRRR
RRRR
Step 2
The aggregated fuzzy pair-wise matrix of the relative importance of treated criteria is:
     
     
      









1,1,15,2,2.05,4,1
5,67.0,2.01,1,15,2,1
1,27.0,2.01,67.0,2.01,1,1
Step 3
The value of fuzzy synthetic extent with respects to the k-th criterion is calculated as follows:
S1 = 1.4, 1.94,3 ∙
1
25
,
1
12.61
,
1
5.8
= 0.056, 0.154, 0.517
S2 = 2.2, 3.67, 11 ∙
1
25
,
1
12.61
,
1
5.8
= 0.088, 0.291, 1.897
S3 = 2.2,7,11 ∙
1
25
,
1
12.61
,
1
5.8
= 0.088, 0.555, 1.897
The weights vector is:
 1,873.0,517.0W p 
The normalized weights vector of criteria is:
w = 0.216, 0.365,0.418
The assessment of participants for each trainer under each treated criterion is presented in Table 2:
Table II: The assessment of the criteria values
k=1 k=2 k=3
t=1 𝑉4, 𝑉5 𝑥31 𝑉4 𝑥4, 𝑉5 𝑥28 𝑉2 𝑥4, 𝑉3 𝑥4 𝑉4 𝑥4, 𝑉5 𝑥20
t=2 𝑉3, 𝑉4 𝑥5, 𝑉5 𝑥26 𝑉3, 𝑉4, 𝑉5 𝑥30 𝑉2 𝑥4, 𝑉3 𝑥4 𝑉4 𝑥4, 𝑉5 𝑥20
Steps 4 and 5
The weighted aggregated values of the participants’satisfaction are calculated as:
 21.0,21.0,16.0;y57xV6xVV216.0
32
1
2
1
d 5
~
4
~
3
~
1
~










 36.0,35.0,27.0;y58xV5xVV365.0
32
1
2
1
d 5
~
4
~
3
~
2
~










 36.0,32.0,25.0;y40xV8xV8xV8xV418.0
32
1
2
1
d 5
~
4
~
3
~
2
~
3
~












Steps 6 and 7: The calculated satisfaction of participants and the representative scalars:
The satisfaction of participants,
~
c as well as the representative scalars are calculated, so that:
c=0.277
Steps 8
The region of education satisfaction for trainers is calculated as follows:

  386.0277.0x
2
~
S
 ,   088.0277.0x
3
~
S

max (0.386, 0.088)=0.386
V. Results And Discussion
Implementation of web-based training, dependent on contemporary Information Technologies (IT) literacy and
actual development of multimedia technology as well as internet applications, could also include the usage of
non-licensed software accompanied with appropriate internet connection stability.
The computed values should be located within the real numbers interval [0-1], where zero represents the
maximal level of satisfaction and one stands for the minimum level of satisfaction (Dubois and Prade, 1980).
However, the boundaries of the interval are supposed not to be reached in reality. The authors` research
experience shows that the good performances of participants within the e-learning framework are to be expected
even though the overall satisfaction degree falls into the third sub-interval with the corresponding value of 0.75.
These arguments of rather positive acceptance of the empirical results are expressed in the end of previous
section. According to previously conducted research experiments, the authors decompose the upper interval into
three sub-intervals: high satisfaction [0-0.3], relative satisfaction [0.31-0.6], low satisfaction [0.61-1]. The
empirical satisfaction of the participants (c=0.277) informs about relatively higher degree of satisfaction among
them, since the value is within the first third of the interval. This could be interpreted in the sense of relatively
higher level of concentration achieved during the training and therefore an increased effort investment and
therefore higher satisfaction with conducted work. We explain this phenomenon as a situation where a
participant is self-content simply due to the effort invested in the actual task.
The mentioned higher concentration level is related to permanent concern over possible damage on the machine
due to exogenous factors, i.e. internet connectivity, lower video coverage. On the other hand, overall satisfaction
among the two trainers is slightly lower, μS=0.386, which informs about the awareness of the risks present
among the lecturers, a priori. This diminishes their degree of satisfaction since they are aware that an absolute
precision will not be reached. Additionally, the relatively insufficient video coverage and ever present
connection buffering increased their skepticism and decreased satisfaction. Therefore the final value is being
located within the second sub-interval.The source of the skepticism among the trainers is depicted on Fig.2.
Figure 2 – Introduction of Safety area during CNC machine manipulation
The genuine contribution of the research article reflects on the introduction of „safety area“ during the direct
training on CNC machines. This area was defined according the authors previous experiences. The proposed
value is 0.1. Unsuccessful is considered to be any value that enters the safety area, i.e. lower than 0.1. Otherwise
the damage would be made on the tool and machine itself. The substantial argument for proposing the safety
area is inability to prevent the accidence beforehand is such an area would not exist. This strongly relates to the
dynamics of the real-time CNC training process. The Basis represents an initial set up for the machine.
Successful is considered to be any value achieved within the suggested sub-zones (grade areas) where each zone
prevents the risk of accidence to occur. This is the first task for the trainer in the process of successful
knowledge transfer. The proposed values of the sub-zones are defined according to previous experience of the
authors. However, of the corresponding grade areas only I-IV are accepted as successful. We have detected 3
(three) candidates that penetrated into the safety area (see Table 3). This validates the initial hypothesis of
correlation between accidence risk and online navigated handling of the machine. The necessity of safety area is
therefore obvious.

Table III: The results of the CNC machine manipulation
By applying the proposed fuzzy model, it could be seen that the equipment quality(k=3) was assigned the
highest weights.This is advantageous for both trainers and trainees, as it defines high quality of the chosen
equipment used in the process of training, which is one of important preconditions for the training quality. In
other hands, it directly connected with the finances supporting such a project. Selection of CNC equipment must
be done in cautious and qualified manner. It should be handed over to a special team of skillful or licensed
instructors.The development and planning of training process is necessary, especially in the case of
CNC technologies and web-based training process. The lowest satisfaction level among participants is
given under the criterion which is denoted as training content (k=1). This refers to alignment of a
designed CNC programming content and the manipulation of CNC equipment in terms of
contemporary scientific and industrial developments; the need of the total economy for a new
technological solutions (i.e. to offer these new solutions to total economy); to offer an approach to the
implementation of these solutions. Additionally, the designed CNC programming training and the
manipulation of CNC machines has to be flexible towards the new market demands and present time.
It can be concluded thatevaluation should be continuously conducted, during and after training
process, with permanent readiness for eventual modifications in the way how training process is being
fulfilled; gained experiences and skills have to be published and used for later trainings.
The main research idea of the article is to precisely define the impact of “e-learning effect” during the training
part related to CNC machine manipulation. The focus is intentionally put on this section due to ever occurring
technicalobstacles. The consequences of this effect are mirrored in the aggregate satisfaction of the participants.
To reach the research goal, we construct an OLS model analogue to the Mincer earning equation2
. The
estimatedmodel is as follows:
𝐺𝑅𝐴𝐷𝐸=𝛽0 +𝛽1𝐸𝐿𝑅𝑁+ 𝛽2𝐸𝐷𝑈+ 𝛽3𝐸𝑋𝑃+ 𝛽4𝐸𝑋𝑃𝑆𝑄+ 𝑢𝑖 (1)
Where,GRADE – is a dependent variable,
βi – estimation coefficient ,
ELRN – E-learning effect,
EDU – Education in years
EXP – Tenure in years,
EXPSQ – Tenure squared (log),
ui – error term.
The sample is augmented by applying the theorem of random sampling, where the main moments are
extrapolated from the original sample (n=32). The number of observations is henceforth 100 instead of 32. The
unsuccessful attempts, see page 12, are excluded from the analysis, which defines the final sample of 91
participants (3 on 32 and 9 on 100). As a proxy for the aggregate satisfaction level of the participants we use the
final marks of each participant during the online CNC manipulation training3
.Here, we deal only with
participants` satisfaction and its causal relation with e-learning effect (see Fig. 3).
Figure 3 – Causal relationship between satisfaction and e-learning effect
2
To capture the effect of e-learning, individual education level and tenure on dependent variable, we find appropriate to
express our model as a modified Mincer equation function.
3
Here we use the final grades obtained during the part of training related to CNC machine manipulation. This however
differs from the final grades obtained after the programming training part conducted on CNC machines.
Area / Values Grade Participants (cum.number)
I 0.10-0.12 5 12
II 0.13-0.15 4 10
III 0.16-0.18 3 4
IV 0.19- 0.20 2 3
V > 0.21 1 0
Unsuccessful < 0.1 3

The authors suppose a high degree of correlation between overall satisfaction of a participant and his grade by
the rule of thumb. Estimation results provide a genuine assessment of how the e-learning affects the
participants` quality assessment of a conducted training. We find a negative correlation between e-learning and
grade with a magnitude of 14.88 (see Table 4).
Consequently, the impact of e-learning is accurately described through uncertainties which it generates. The
lack of preciseness, or put in slightly another way, the variations within the precision levels of participants are
its obvious manifestations. This estimation of e-learning effect during the CNC training represents the main
contribution of this research paper. On the other hand, remaining regressors do have a positive influence on
regressand, with significantly lower estimated coefficients. Furthermore, in proposed model only e-learning
effect regressor is significant. Although the explanatory power of the OLS model is relatively low (R2
= 0.20),
the findings point out to a negative effect of the CNC training conducted online. This finding is a novel
contribution to the mostly scarce scientific research within the field. Additionally, education is positively related
to the grade. Interestingly, not significant when e-learning variable is included in the model. This can be
explained by the existence of the uncertainties during the product processing in a real time on the CNC machine.
The video technology and internet connection buffering hinder the concentration of the participant and
objectively diminish the chances of reaching the absolute precision required in serial production heavily related
to CNC machines.
Table IV: OLS regression model results
Source SS df MS Number of obs. = 91
Model 20.2311645 4 5.05779114 F (4,86) = 5.69
Residual 76.4941102 86 0.889466397 Prob. > F = 0.0004
R-squared = 0.2092
Total 96.7252747 90 1.07472527 Adj. R-squared = 0.1724
Root MSE = 0.94312
Grade Coefficient Stand. Err. t P > | t | [ 95% Conf. Interval ]
E-learning
Education
Experience
Exp_square
_cons
-14.87932
0.2333091
0.0835151
-1.380103
4.257427
3.580344
0.1681665
0.1255672
1.540343
3.495152
-4.16
1.39
0.67
-0.90
1.22
0.000
0.169
0.508
0.373
0.227
-21.99681 -7.761832
-0.1009949 0.5676131
-0.1661043 0.3331345
-4.442203 1.681997
-2.690707 11.20556
Some of the possible improvement strategies that may be employed for improving values of CNC e-learning
process quality area: a)Constant following of IT trends with respect to CNC equipment; b) alignment of training
contents for CNC programming and CNC equipment manipulation. In addition, it is necessary for the
participants who attend the training to possess the fundamental IT knowledge. Furthermore, it is important to
adjust the training materials separately according to each participant group. The training content are defined
with respect to two essential requirements: 1) have to be in accordance with the State Ministry of Education
predefined education programs and study profiles and 2) should allow for skills and knowledge improvement of
the participants in the area of CNC machines (beneficial to participants that reached M. Sc. / Ph. D level and are
employed at the High School with technical orientation). The fulfillment of above mentioned requests bears
certain limitations, for instance: available resources, motivation of teaching stuff towards professional
improvement, an insight that a leading stuff has in IT development and the desire to implement this
development, adequate selection of lecturers according to their affinities and professional goals, etc.
VI. Conclusion
Taking into account the requirement for a continuous improvement of an education process, as well as rapid,
perpetual IT development and the customer demands, it can be said that the lack of constant education,
especially at High Schools within technical area, prevents a fast recovery when the business disturbances occur.
The assessment and management of the considered education process quality may be introduced through
identification of satisfaction level of participants over time. The solution of the considered problem is obtained
in exact way because the solution is less burdened by the subjective judgments of decision makers.
All uncertainties in relative importance of evaluation criteria, and their values are described by predefined
linguistic expressions. Decision makers present their opinion by using linguistic expression in more precise way
than by using precise numbers. Linguistic variables are modeled by TFNs.
The AHP framework is used for the rating of the relative importance of evaluation criteria since it is closer to
the human way of thinking. The determination of elements values of constructed fuzzy pair-wise comparison
matrix is stated as a fuzzy group decision making. The aggregation of individual opinions into a group
consensus is given by the proposed method which can be found in the appropriate literature. By using the
procedure for the handling of fuzzy AHP, the weights of evaluation criteria are given by TFNs. The RF values
are assessed by participants.

The weighted value of each identified criterion is calculated by using fuzzy averaging method and fuzzy algebra
rules. The overall satisfaction level of participants with respect to all criteria is calculated by using union
operation of TFNs.The efficiency and reliability of the proposed methods for evaluating the education
processes’ quality properties are described using the IF-THEN rules.
The proposed procedure is illustrated by real-life data from secondary technical schools in central Serbia. The
outputs of the proposed fuzzy model represent the base for definition of strategy that should be oriented to
enhancement of the education process quality. Furthermore, the quantitative assessment of the e-learning impact
on participants` satisfaction with the conducted training represents the major contribution of the article. We find
that CNC training conducted online will hinder the results of participants per se, and we calculate its magnitude
on 14.88 measurement points. Conversely, other authors account for a positive impact of e-learning, albeit not
precisely targeting the CNC machine training. The formation of “safety area” is another genuine contribution of
the research paper. The comparison with other authors is not possible since the scarcity of quantitative findings
in other articles prevents us to conduct the parallel comparisons. Additionally, we find that the application of
freeware software instead of licenced software did not affect the work confidence of the applicants and the
overall training quality. The arguments are to be found in the questionnaire T4, e.g. quaetion no. 7 and no.8.
Both versions of software were applied in each session, i.e. 2 days training for each version. These all account
for major benefits of conducted research. In this way, the costs and time needed for improvement of the treated
process may be decreased and optimized. The presented approach is novel and presents one of the first steps in
mathematical modeling of CNC e-learning education process quality. Contributions of this paper are the
following: (1) it proposes a new fuzzy model for calculating the quality estimation for each CNC e-learning
education criterion, and (2) it handles uncertainty within considered criteria using fuzzy sets.
Besides the advantages, the proposed model has certain constraints. Having in mind the main constraints of the
proposed fuzzy model are: the number of evaluation criteria, number of trainers and participants. The sample
used for testing the proposed fuzzy model is constrained by: the number of High School with technical
orientation that offer this study programme; the needs of the enterprises that operate in domestic market;
tendency of the top management in industrial enterprises to educate their employees with relevant CNC
programming skills; relatively low financial capabilities of considered enterprises (e.g. generally all enterprises
on domestic market belong to a group of small and medium size enterprises); etc. One of future directions is
set to development of software based on the proposed model. In addition, the software solution could be
expanded with additional functionalities for better education process management.
References
[1]. W. Wu, W. Chen, L. Fang, C. Lu, Development and Evaluation of Web Service-Based Interactive and Simulated Learning
Environment for Computer Numerical Control,Computer Applications in Engineering Education, 18(3), 2010, pp. 407-422.
Availablefrom<http://onlinelibrary.wiley.com/doi/10.1002/cae.20147/abstract;jsessionid=BA704ECF2E7FAC528DAB8BF413086
266.f01t02?userIsAuthenticated=false&deniedAccessCustomisedMessage=> [15.07.2013]
[2]. G. Kearsley, B. Shneiderman, Engagement Theory: A Framework for Technology-Based Teaching and Learning, Educational
Technology, 38(5), 1998, pp. 20-23.
[3]. A. Parush, H. Hamm andA. Shtub, Learning Histories in Simulation-Based Teaching: The Effects on Self-learning and
Transfer,Computers and Education, 39(4), 2002, pp. 319-332. Available from
<http://www.sciencedirect.com/science/article/pii/S036013150200043X> [08.06.2013]
[4]. M.Exel, S.Gentil, F.Michau, D.Rey, Simulation Workshop and Remote Laboratory: Two Web-based Training Approaches for
Control,Proc. of the 2000 American Control Conference, Chicago, USA,28.06.2000,pp.3468-3472.
[5]. H. Hanwu and Wu. Yueming, Web-based Virtual Operating of CNC Milling Machine Tools,Computers in Industry,60(9), 2009, pp.
686-697, Available from <http://www.sciencedirect.com/science/article/pii/S0166361509000955> [23.05.2013]
[6]. S.-H. Suh, Y. Seo, S.-M. Lee, T.-H. Choi, G.-S. Jeong, D.-Y. Kim, Modelling and Implementation of Internet-based Virtual
Machine Tools, The International Journalof Advanced Manufacturing Technology, 21(7), 2003,pp. 516–522. Available from
<http://link.springer.com/article/10.1007%2Fs001700300061> [27.06.2013]
[7]. S.K. Ong, L. Jiang, A.Y.C. Nee, An Internet-based Virtual CNC Milling System, The International Journal of Advanced
Manufacturing Technology, 20(1), 2002, pp. 20–30. Available from <http://link.springer.com/article/10.1007%2Fs001700200119>
[13.05.2013]
[8]. P.-L. Hsu and W. T. Yang, Real-time 3D Simulation of 3-axis Milling Using Isometric Projection, Computer-Aided Design, 25(4),
1993,pp. 215–224. Available from <http://www.sciencedirect.com/science/article/pii/001044859390052P> [19.04.2013]
[9]. W. Lihui, P. Orban, A. Cunningham, S. Lang, Remote Real-time CNC Machining for Web-based Manufacturing, Robotics and
Computer- IntegratedManufacturing,20(6),2004, pp. 563–571. Available from <http:/ /www. sciencedirect. com/ science/
article/pii/S0736584504000481> [24.04.2013]
[10]. C.-H. Yeung, Y. Altintas, K. Erkorkmaz, Virtual CNC System. Part I. System architecture, International Journal of Machine Tools
and Manufacture ,46(10), 2006,pp. 1107–1123. Available from <http://www.s ciencedirect. com /science/ article/pii/
S089069550500218X> [17.03.2013] [11] W.D. Li, S.K. Ong, A.Y.C. Nee, A Web-based Process Planning Optimization System
[11]. for Distributed Design, Computer-Aided Design,37(9), 2005,pp. 921–930. Available from <http:// www. sciencedirect.
com/science/article/pii/S0010448504002593> [22.02.2013]
[12]. D. Y. Shee and Y-S. Wang, Multi-criteria Evaluation of the Web-based E-learning System: A Methodology Based on Learner
Satisfaction and Its Applications, Computers and Education, 50(3), 2008,pp. 894-905. Available from <http:/ /www.
sciencedirect.com/ science/article/pii/S0360131506001424> [14.01.2013]
[13]. G.J. Klir and T. Folger, Fuzzy Sets, Uncertainty, and Information, Prentice Hall, Upper Saddle River, New York, USA, 1988.

[14]. H.J. Zimmermann, Fuzzy Set Theory and Its Applications, Springer Science and Business Media, Boston, USA, 2001.
[15]. M. Stefanović, D. Tadić, S. Arsovski, Z. Arsovski, A. Aleksić, A Fuzzy Multicriteria Method for E-learning Quality Evaluation,
International Journal of Engineering Education,26(5), 2010, pp. 1200-1209. Available from <http://www.researchgate .net /profile
/Miladin_Stefanovic/publication/228409362_A_Fuzzy_Multicriteria_Method_for_E-learning_Quality_Evaluation
/links/0fcfd50fd75c55965f000000.pdf> [19.01.2013]
[16]. J.C. Pomerol and S. Barba-Romeo, Multicriteria Decision Management: Principles and Practice, Springer Publishing, Boston,
USA,2000.
[17]. R. Petrović and D. Petrović, Multicriteria Ranking of Inventory Replenishment Policies in the Presence of Uncertainty in Customer
Demand,International Journal of Production Economics,71(1-3), 2001,pp. 439-446. Available from <http://
www.sciencedirect.com/science/article/pii/S0925527300001390> [30.12.2012]
[18]. M. Stefanović, D. Tadić, S. Nestić, A. Đorđević, An Assessment of Distance Learning Laboratory Objectives for Control
Engineering Education, Computer Applications in Engineering Education, 23(2), 2013, pp. 191-202. Available from
<http://onlinelibrary.wiley.com/doi/10.1002/cae.21589/abstract> [14.02.2013]
[19]. P.P.B. Rao and N.R. Shankar, Ranking Generalized Fuzzy Numbers Using Area, Mode, Spreads and Weights, International
Journal of Applied Science and Engineering, 10(1), 2012,pp. 41-57. Available from <http:/ /ir.lib.cyut. edu.tw:8080/bitstream/
310901800/12642/1/4.pdf> [17.03.2012]
[20]. D. Tadić, T. A. Gumus, S. Arsovski, A. Aleksić, M. Stefanović, An Evaluation of Quality Goals by Using Fuzzy AHP and Fuzzy
TOPSIS Methodology, Journal of Intelligent and Fuzzy Systems,25(3), 2013,pp. 547-556. Available from
<http://www.researchgate.net/profile/Miladin_Stefanovic/publication/236029667_Evaluation_of_Quality_Goals_by_Using_Fuzzy_
AHP_and_Fuzzy_TOPSIS_Methodology/links/0c96052d05a1003fa5000000.pdf> [24.03.2013]
[21]. N.Y. Secme, A. Bayrakdaroglu, C. Kahraman, Fuzzy Performance Evaluation in Turkish Banking Sector Using Analytic Hierarchy
Process and TOPSIS, Expert System with Applications, 36(9), 2009,pp. 11699-11709. Available from <http://www.sciencedirect.
com/science/article/pii/S095741740900253X> [13.04.2012]
[22]. T. Kaya, C. Kahraman, Multicriteria Decision Making in Energy Planning Using a Modified
Fuzzy TOPSIS Methodology, Expert Systems with Applications,38(6), 2011,pp. 6577-6585. Available from
<http://www.sciencedirect.com/science/ article/pii/S0957417410013199> [17.05.2013]
[23]. D. Tadic, D. D. Milanovic, M. Misita, B. Tadic, A New Integrated Approach to the Problem of Ranking and Supplier Selection
under Uncertainties, Proceedings of the Institution of Mechanical Engineers, Part B: Journal of Engineering Manufacture,225(9),
2011,pp. 1713-1724. Available from <http://pib.sagepub.com/content/225/9/1713.abstract> [09.01.2013]
[24]. D. Y. Chang, Applications of the Extent Analysis Method on Fuzzy AHP, European Journal of Operational Research,95(3),
1996,pp. 649-655. Available from <http://www.sciencedirect.com/science/article/pii/0377221795003002> [22.04.2012]
[25]. H.J. Zimmermann, Results of Empirical Studies in Fuzzy Set Theory,Applied General Systems Research, 5, 1978,pp. 303-311.
Available from <http://link.springer.com/chapter/10.1007%2F978-1-4757-0555-3_21> [09.05.2012]
[26]. W. Pedrycz and F. Gomide, F, An Introduction to Fuzzy Sets, Analysis and Design, MIT-Press: Cambridge Massachusetts, 1988.
[27]. J. G. Klir and B. Yuan, Fuzzy Sets and Fuzzy Logic, Theory and Applications, Prentice Hall. New Jersey, 1995. Available from
<http://digilib.uin-suka.ac.id/7049/> [24.06.2013]
[28]. F.A. Lootsma,Fuzzy Logic for Planning and Decision Making, Springer Inc., Boston, USA, 1997.
[29]. M.S. Bass and H. Kwakernaak, Rating and Ranking of Multiple-aspect Alternatives Using Fuzzy Sets,Automatica, 3(1), 1977,pp.
47-58. Available from <http://www.sciencedirect.com/science/article/pii/0005109877900085> [22.02.2013]
[30]. D. Dubois and H. Prade, Decision-making under fuzziness, In M. M., Gupta, R. K., Ragade, R. R.Yager, (Eds.), Advances in Fuzzy
Set Theory and Applications, 1979,pp. 279-302, Amsterdam: North-Holland.

New Fuzzy Model for quality evaluation of E-Training of CNC Operators

More Related Content

What's hot (20)

Similar to New Fuzzy Model for quality evaluation of E-Training of CNC Operators (20)

Recently uploaded (20)

New Fuzzy Model for quality evaluation of E-Training of CNC Operators