You are on page 1of 13

Computer-Aided Design 45 (2013) 937–949

Contents lists available at SciVerse ScienceDirect

Computer-Aided Design
journal homepage: www.elsevier.com/locate/cad

A modeling approach for maintenance safety evaluation in a virtual


maintenance environment
Jie Geng a,b , Dong Zhou a,b,∗ , Chuan Lv a,b , Zili Wang b
a
State Key laboratory of Virtual Reality Technology and System, Beijing, 100191, PR China
b
School of Reliability and Systems Engineering, Beijing University of Aeronautics and Astronautics, Beijing, 100191, PR China

article info abstract


Article history: This paper presents a novel virtual maintenance application for maintenance safety evaluation (MSE)
Received 21 July 2012 to provide recommendations on maintenance safety during the early stages of product design. The
Accepted 27 January 2013 proposed methodology detects the potential defects via virtual maintenance technology, and safety
evaluation could be carried out in a predictive way intuitively and systemically to decrease the possibility
Keywords: of accidents. After reviewing relevant publications, a MSE model with thorough consideration of the
Virtual maintenance
characters in the virtual maintenance environment was addressed. The model consists of evaluation
Maintenance safety evaluation
Virtual design
contributors, evaluation criteria, and evaluation methodology. A case study on APU system maintenance
was used to illustrate the performance of the evaluation model. The result showed the feasibility of virtual
maintenance in improving maintenance safety designs.
© 2013 Elsevier Ltd. All rights reserved.

1. Introduction Bearing the aforementioned observations in mind, this paper


introduces a maintenance safety evaluation model to launch main-
The innovation and development of industrial technology needs tenance safety evaluation (MSE) in the virtual maintenance en-
the constant improvement of newly developed products in vari- vironment (VME). Appropriate maintainability evaluation factors
ous realms. Examples of such innovations in industrial technology were selected according to the VME characteristics. In addition,
include numerous composite materials, high automatization and ascending hierarchical analysis structure for maintenance safety
intellectuality, multiple measurement and control equipments, evaluation was established based on the hierarchic analysis prin-
advanced Built-In Test (BIT) devices, etc. Without a doubt, most ciple. The corresponding evaluation criteria for each factor were
maintenance tasks must be accomplished by man. Moreover, the given by determining the grades. A methodology applied to the
continuous renewal of products results in the difficulty in identi- hierarchic analysis structure was then adopted to quantitatively
fying when and where failure occurs. Furthermore, a larger quan- calculate the qualitative evaluation factors. The authenticity, ac-
tity of maintenance work frequently draws out to a great length, curacy, and integrality of the simulation processes in VM lay the
whereas the high utilization ratio reduces maintenance time. Con- foundation for the later safety evaluation. The whole simulation
sequently, maintenance personnel always bear great psychological process of the virtual human was assumed according to the opera-
and mental pressures under the lack of essential operation training tional guidelines in VME to accomplish this goal. MSE was carried
and manipulation guidance, which naturally influences the occur- out, combining the inspection for virtual prototype and analysis
rence of accidents. tools in VME. Finally, a series of maintenance safety design defects
In actual maintenance safety works, integrated handbook and
were exploded compared with the criteria. This process is essen-
standards are strongly recommended. The lack of maintenance
tial in obtaining more comprehensive suggestions for both main-
safety design guides causes the invalidity of accident prevention
tenance safety literature and maintenance operation guides to
during maintenance processes. As a result, things that need
reduce accidents.
improvement are exposed after use and maintenance.
The rest of this paper is organized as follows: Section 2 reviews
correlative literature in this paper; Section 3 shows the details of
maintenance safety evaluation model architecture; Section 4 gen-
∗ Correspondence to: Room 633, Weimin Building, Beijing University of erally describes the VME and the virtual maintenance simulation
Aeronautics and Astronautics, 37#, Xueyuan Road, Haidian District, Beijing, PR case for APU disassembly and deals with the implementation of the
China. Tel.: +86 010 82338334; fax: +86 010 82313763.
E-mail addresses: ddgj516@126.com (J. Geng), buaazd0926@126.com
evaluation model combining the case study; Section 5 concludes
(D. Zhou), lc@buaa.edu.cn (C. Lv), wzl@buaa.edu.cn (Z. Wang). the paper and presents several discussions.

0010-4485/$ – see front matter © 2013 Elsevier Ltd. All rights reserved.
doi:10.1016/j.cad.2013.01.007
938 J. Geng et al. / Computer-Aided Design 45 (2013) 937–949

2. Literature review to solve complex systems with multiple criteria. Meanwhile, the
consistency check can be unfavorable when a layer contains nu-
Recent publications relevant to this paper were mainly con- merous factors [43–46]. The entropy weight method calculates the
cerned with two research streams: virtual maintenance and main- integrative indicator with a large amount of information from each
tenance safety evaluation. In this section, relevant literature were element, whereas human ignorance will make it aimless [47,48].
summarized. The principle component analysis translates all the indicators into
For virtual maintenance, a considerable variety of studies have several independent integrative indicators; thus, the quantity and
been conducted since the late 1990s. In the recent years, the pre- the increase or decrease of indicators will affect the final result
dominance of virtual maintenance allowed more comprehensive [49,50]. The extension assessment method focuses on matter el-
applications, such as maintenance training [1,2], product design ement construction. However, the completeness of indicators and
[3,4], process simulation, and analysis [5–8]. Consequently, virtual the rationality of weights are the crux of the method [51,52]. The
maintenance can be used as a tool to simulate and analyze the synthetical index method has an uncomplicated principle and cal-
maintenance process to guide product design with virtual proto- culation, but is highly dependent on weight [53,54]. The factor
typing. analysis method creates several specific common factors based on
For maintenance safety evaluation, a great diversity of models the principle component analysis, whereas the common factors
was adopted during the past decades, including a techno-economic have little relevance with each other [55,56]. Table 1 shows the
model applied to the design and maintenance optimization based superior and inferior among the eight methods.
on safety analysis [9], a theoretical model used to carry out a sec-
ondary analysis of available data on maintenance accidents [10], 3. Development of the MSE model
various probabilistic risk analysis models for system safety as-
sessment [11–15], a model that combined a failure model and a
3.1. General framework
maintenance model for analyzing the impact of imperfect main-
tenance on system safety [16], and several other models in dif-
The proposed MSE model consists of three parts, namely, eval-
ferent domains [17–23]. Meanwhile, different approaches and
uation element, evaluation criteria, and evaluation method. Based
methods were proposed for safety management, analysis, and eval-
on existing maintainability design requirements, the evaluation
uation. For example, a systematic approach for the management
element selection should take full accountability of the feasibil-
of safety case change [24], an approach based on the effective
ity of launching evaluation in VME. Therefore, the selected ele-
coupling of genetic algorithms and Monte Carlo simulation for
ments can be analyzed with relating tools in VME. A hierarchical
the multi-objective optimization of nuclear safety systems [25],
structure of evaluation elements is constructed after confirmation.
a safety evaluation tool based on a detailed analysis of the safety
Evaluation criteria are the core of the whole framework; the es-
awarded by different types of measures [26], an evaluation system
tablishment of corresponding criteria to each element greatly in-
that used a combination of Analytic Hierarchy Process, decision
fluences the accuracy of evaluation results. Correlative description
rules and Bayesian tools belonging to operational research [27],
on maintainability design requirement of the selected elements
and a decision-making method for coal mining safety evaluation
is the main reference during evaluation criteria formulation. The
with linguistic values [28].
formulated criteria will grade the evaluation element according
However, few existing models were proposed for maintenance
to appeared maintenance safety condition. The analytic hierarchy
safety evaluation in VME. Moreover, safety evaluation needs to be
process (AHP) calculates the weight of each index due to the hierar-
carried out in a predictive way. If prevention is fully effective only
chical structure of evaluation element. In addition, the principle of
after the error has occurred, it is too late. One must identify risks
selecting snapshots for evaluation and confirming the safety level
to be evaluated and controlled before accidents happen [29,30].
for each element were introduced. The process of generating the
In addition, operational and maintenance activities are closely
MSE model is shown in Fig. 1.
related to each other to contribute to overall safety, which in turn
contributes to production quality and profit [31]. Hence, launching
MSE in VME is harmful to the improvement of product design in 3.2. Detailed description of the threefold in the MSE model
a predictive way during the early stages. Experience is extremely
vital during maintenance process [31]. However, the improvement 3.2.1. Evaluation element
solution from evaluation in VME will commendably guide the Personnel safety is an essential component in general maintain-
maintenance processes. ability design standards. Other aspects relating to personnel safety
Evaluation factor and criteria are highly important for the eval- in maintainability design standards should be considered when se-
uation of results [32,15,33]. Thus, the evaluation factor options lecting evaluation elements.
and criteria should be based on the problem being evaluated and Accessibility design affects the range of maintenance personnel
the scientific and practical relevance under the given circum- vision, the contact area of maintenance personnel, and the space
stances [34,35]. Previous industrial accidents provided predictive of maintenance personnel operation. If accessibility is poorly
information for safety factor confirmation. These information in- designed, maintenance personnel rely heavily on experience and
clude human factors [36–38] and mechanical risk factors [39]. Con- feelings when launching maintenance work, especially when
sequently, both should be concerned about when confirming VME facing the latest products. These scenarios always cause injury to
evaluation factors. the maintenance personnel with no preparation at all. For example,
Among the quantitative evaluation methods, eight common maintenance personnel will try every means to approach the
evaluation methods were compared to confirm a suitable one to maintenance pot with appropriate tools when the target lies on
the following proposed model in ascertaining evaluation index the critical edge of the contact area. In this situation, the target
weight. Gray evaluation method can launch a quantitative evalua- will most likely lose balance and fall from the platform. Therefore,
tion for qualitative factors, but subjectively when grading [40]. The availability is served as the first evaluation element, which consists
Fuzzy comprehension evaluation method qualifies the factors with of three sub-elements: visibility, reachability, and operation space.
a fuzzy relationship. The assurance of membership function is a Error proofing and marking designs influence the accuracy of
crucial part during application [41,42]. Analytical hierarchical pro- maintenance personnel in connecting, disassembling, installing,
cess effectively combines the quantitative and qualitative analysis filling, and so on. Errors objectively and inevitably exist in different
J. Geng et al. / Computer-Aided Design 45 (2013) 937–949 939

Table 1
The comparison among eight methods.
Evaluation method Superior Inferior

Gray evaluation method Analyzes qualitative element quantitatively Grades subjectively


Fuzzy comprehension evaluation Qualifies the factors with a fuzzy relationship Membership function assurance
Analytical hierarchical process Combines the quantitative analysis and qualitative analysis Consistency check might be unfavorable
Entropy weight method Calculates with a large amount of information Ignore human will
Principle component analysis Decrease the number of factors The result relies on the change of factors
Extension assessment method Constructs matter element The result relies on the completeness of indicators and the
rationality of weights
Synthetical index method Uncomplicated principle and calculation Depends on the weight too much
Factor analysis method Creates several specific common factors The common factors have little relevance with each other

Fig. 1. The process of generating the MSE model.

domains. Thus, error proofing design is fundamental to prevent 3.2.2. Evaluation criteria
maintenance personnel accidents. Personnel operation may tread Evaluation criteria are the fundamental basis to launch the
awry if two parts have similar shapes, but entirely distinct whole VME evaluation work. Corresponding contents related to
functions due to poor error proofing design, which may also result maintenance safety in maintainability design standards were con-
in accidents. For example, if the maintenance personnel hand is in sulted to confirm the evaluation criteria. The evaluation criteria
a dangerous area, the product maintenance can be disabled in this will be classified into three grades according to different embod-
condition. Furthermore, the correct making design can be carried iments with the help of relating VME analysis tools. Consequently,
out in striking position. Error proofing and marking designs are pertinent analysis tools in VME were explicated in detail, combined
affirmed as the second element, which naturally covers two child with their application in ranking the criteria. Various types of vir-
elements, namely, error proofing and making designs. tual simulation tolls have been applied in all walks of life and have
Ergonomic design concerns physical and psychological factors their own advantages. A summary of all the corresponding criteria
during the maintenance process. This design investigates the rela- is listed in Table 3.
tionship between body measurements and maintenance objects to
relieve discomfort and increase the efficiency of the maintenance 3.2.2.1. Criteria for accessibility design. (1) Visibility design.
personnel. Faulty ergonomic designs result in incompatible activi- Visibility design should guarantee the maintenance person-
ties and fatigue to the maintenance personnel, which is the hidden nel with a clear view of the target during the maintenance pro-
peril of accidents in several cases. Ergonomic design considers the cess. Hence it plays a central role in a man–machine system while
human body as the core during product design. Rational combina- manipulation [57,58]. Most frequently used displays should be
tion between product design and ergonomic design can effectively grouped together and placed in the optimum visual zone and ar-
minimize maintenance accidents; thus, ergonomic design is taken ranged in sequence within functional groups to provide a view-
as the third element. The Rapid Upper Limb Assessment (RULA) is ing flow from left-to-right or top-to-bottom, and vision models
an ergonomic technique that evaluates human exposures to pos- are obtained with a elliptical appearance in a computational way
tures, forces, and muscle activities. This technique shows the risk recently [59–63]. If the maintenance target is located in the best
level from continuous work. The ergonomic evaluation approach vision area, the maintenance personnel can clearly see the target
results in a risk score between one and seven, where higher scores and easily launch the work, as shown in Fig. 3. This result indicates
signify greater levels of apparent risks. RULA is adopted as the child a favorable visibility design; this situation is classified as rank A.
element of the third element ergonomic design. If the maintenance target is located in the widest vision area, al-
Physical harm occurs a lot during practical maintenance work. though the maintenance personnel have adjusted the posture, the
Heat, electrical, and mechanical injuries are examples of physical eyes may feel tired in focusing the obscene scene for a long time.
harm. Heat injury can decrease efficiency which leads to human This result illustrates a poor visibility design; this situation is clas-
error, electrical injury occurs from incorrect operation, whereas sified as rank B. If the maintenance target is located in the invisible
mechanical injury springs from shape corner and edge on equip- vision area even if the maintenance personnel adjusted the pos-
ments. Consequently, physical harm preventing design is consid- ture, they will not see the target at all and attempt to run the work
ered as the last element comprising the aforementioned injuries. A in terms of past experience. This situation shows a bad visibility
hierarchical structure of evaluation element is shown in Fig. 2. design and is classified as rank C.
940 J. Geng et al. / Computer-Aided Design 45 (2013) 937–949

Fig. 2. The hierarchical structure of evaluation element.

(2) Reachability design.


Reachability design should make it feasible to reach the mainte-
nance target along predefined maintenance path. The spatial reach
area for workplace design purposes is determined by not only the
length of forearm, but also the shoulder breadth, shoulder and
torso rotation, spine bending, and hand operation characters. Var-
ious methods, such as photo printing, device measuring, and so
on, have been presented to dynamically measure the reach area
and generate a reach envelope to show the most possible range
of the hands or fingers according to arm movement limitation
[64–66]. Considering finger and wrist bending in operation, we
take the hand length 18 cm (50th percentile male data) as the
quantitative consideration for reachability criteria. As shown in
Fig. 4, maintenance target locating in space 1 (the inner space
18 cm away from the boundary of the envelope) indicates a fine
reachability design, which is classified as rank A. Furthermore,
maintenance target locating in space 2 (the exterior space 18 cm
out of the boundary of the envelope) illustrates a poor reachability
Fig. 3. Vision area of static eyes.
design, which is classified as rank B. If the maintenance target is
located just around the boundary (the nearby space 18 cm around
the boundary of the envelope) of the envelope, the maintenance
personnel will try to instinctively approach, which can cause acci-
dents. Therefore, it is classified as rank C.
(3) Operation space design.
Operation space design should provide enough space for human
movement and operation. Most VMEs launch collision detections
between two manikins or between a manikin and device based
on two major categories: the discrete collision detection method
and the continuous collision detection method [67–70]. Collision
detection can intuitively analyze the collision condition between
the human body and surroundings in real-time. Therefore, this
method is adapted to check the operation space for new products.
As shown in Fig. 5, collision part will become red if collision occurs,
Fig. 4. Reach envelope of an arm.
which is beneficial for operation space design analysis.
The total operation time at each maintenance spot is certain ac-
cording to the maintenance schedule, so we take the percentage of steps in the whole maintenance procedure, (the percentage of col-
collision state duration as the quantitative consideration for oper- lision duration is 10%–90% of the total operation time) operation
ation space criteria. If collision hardly occurs between the human space design should be improved, and thus be classified as rank B.
body and ambient equipments during the whole maintenance pro- If the human body always collides with its surroundings (the per-
cess (the percentage of collision duration is less than 10% of the to- centage of collision duration is more than 90% of the total operation
tal operation time), the maintenance personnel can freely operate time), the operation space design requires immediate change, and
with hands or tools. This result expresses adequate space for oper- otherwise, maintenance work cannot be carried out. This situation
ation and is classified as rank A. If collision occurs during several is classified as rank C.
J. Geng et al. / Computer-Aided Design 45 (2013) 937–949 941

(LUBA) [72], and rapid upper limb assessment (RULA) [73] are com-
monly used in ergonomic analysis. In addition, RULA is widely used
because of the simple and feasible procedure without interruption.
RULA launched the arm, wrist, neck trunk, and leg analysis, respec-
tively. RULA followed the predefined step-by-step procedure and
dealt with two analysis results to obtain a final score. The inter-
mediate scores, represented by a number and a color, were used
to calculate the final RULA score [73]. Table 2 indicates the score
range for each segment and the associated color. If the final score
is 1 or 2, the corresponding human limb will be green. This result
means an acceptable ergonomic design and the situation is clas-
sified as rank A. A final score of 3 or 4 and a yellow colored limb
means that the situation needs further investigation. A final score
of 5 or 6 and an orange colored limb means that the situation needs
further investigation and has to be changed soon. Both situations
Fig. 5. Collision detection between human and device. are classified as rank B. If the human limb turns red with a highest
final score of 7, the situation must be investigated and immediately
3.2.2.2. Criteria for error proofing design. Error proofing design changed for human safety. This situation is classified as rank C.
ensures the accomplishment of maintenance safety with an eye
on improving designs. Changing the sharp surface of connectors 3.2.2.4. Criteria for physical hurt prevention design. Prototype check
connected with cables that have the same radius but completely is employed to confirm the criteria for physical harm prevention
different functions, or selecting cables with distinguished colors which is the same criteria for error proofing design.
are examples of error proofing designs. However, the complexity of (1) Heat injury prevention design.
the product leads to the accuracy of dangerous position. Therefore, Essential heat protecting devices should be established to avoid
marking design is essential in this situation, such as direction contact between the skin and extremely hot or cold surfaces. Thus,
arrow, caution alert, tool warning, and so on. Prototype check is checking the position of heat protecting devices such as metal
effective especially for complex systems in the VME and discovers mesh and insulating panels and examining the contact between
an extraordinary number of defects. Thus, prototype check is the human limbs and product surfaces can be used to confirm the
main approach to confirm the criteria in this section. criteria for heat injury prevention designs. The total amount of heat
(1) Error proofing design. injury prevention is certain according to the given maintenance
The total amount of error proofing design is certain according process, so we take the percentage of heat injury prevention
to the given maintenance process, so we take the percentage of characters amount as the quantitative consideration for heat injury
error proofing design characters amount as the quantitative con- prevention design criteria. If necessary protective devices are
sideration for error proofing design criteria. After checking the appropriately set (the percentage of heat injury prevention design
corresponding position of the virtual prototype according main- characters amount is more than 90%) and no contact happens in the
tenance procedures, corresponding characters (the percentage of whole simulation, the condition is classified as rank A. Meanwhile,
error proofing design characters amount is more than 90%) in the if the protective devices are deficient (the percentage of heat
prototype can reasonably reflect the error proofing design and ef- injury prevention design characters amount is 10%–90%) or contact
fectively reduce accidents. Thus, this situation is classified as rank between human limbs and hazardous surfaces sometimes occur in
A. Error proofing designs in products are sometimes (the percent- the maintenance simulation, the situation is classified as rank B. If
age of error proofing design characters amount is 10%–90%) incom- few protective devices are present (the percentage of heat injury
pletely and irrationally devised; thus, this situation is classified as prevention design characters amount is less than 10%) or human
rank B. If few error proofing designs can be noted in the deserved limbs are always in contact with dangerous surfaces during the
prototype position (the percentage of error proofing design charac- simulation process, the situation is classified as rank C.
ters amount is less than 10%), accidents may happen if the mainte-
(2) Electrical injury prevention design.
nance personnel operate without too much experience. Thus, this
Most electrical equipments can cause electrical injuries because
situation is classified as rank C.
of the voltage, electric, and frequency together. Live working is
(2) Marking design. forbidden during maintenance processes to avoid electric shocks.
Similarly, prototype check is adapted to establish marking Electric cables should be distant from sharp-edged products. Both
design criteria, and we also take the percentage of marking design aspects should be concerned about confirming the electrical injury
characters amount as the quantitative consideration for marking prevention design criteria. If electricity is shut down by virtual
design criteria. A distinct and logical marking in the prototype humans in the simulation process and electricity cables during
can be noticed for accident prevention (the percentage of marking maintenance procedures probably wire away from sharp-edged
design characters amount is more than 90%). Thus, this situation objects, the situation is classified as rank A. If live working is
is classified as rank A. Meanwhile, confused and fragmentary
avoided, but electricity cables are in contact with sharp-edged
markings in the prototype present a marking design that requires
objects in several conditions, the situation is classified as rank
proof (the percentage of marking design characters amount is
B. Meanwhile, if live working is not avoided in the simulation
10%–90%); thus, is classified as rank B. A prototype with few
process whether or not electricity cables wire away, the situation
marking designs is classified as rank C (the percentage of marking
is classified as rank C because live working is extremely perilous.
design characters amount is less than 10%).
(3) Mechanical injury prevention design.
3.2.2.3. Criteria for ergonomic design. Ergonomic designs focus Mechanical injury covers several situations, including crush,
on the man–machine interface including physiological and psy- friction, stabbing, and so on. In this section, we focus on sharp-
chological factors to identify different defects in the design and edged objects exposed in the maintenance path that can affect
provide suggestions to ensure safe operation during the mainte- maintenance personnel safety. The total amount of mechanical in-
nance process. In general, ovako working-posture analyzing sys- jury prevention is certain according to the given maintenance pro-
tem (OWAS) [71], posture loading on the upper body assessment cess, so we take the percentage of mechanical injury prevention
942 J. Geng et al. / Computer-Aided Design 45 (2013) 937–949

Table 2
2RULA score.

Table 3
Summary of criteria to corresponding evaluation element.
Evaluation element Rank A Rank B Rank C
Criteria Corresponding Criteria Corresponding Criteria Corresponding
illustration embodiment in illustration embodiment in illustration embodiment in
VME VME VME

Visibility design Target could be Target locates in Target could be Target locates in Target could not Target locates
Accessibility
seen directly the best vision seen partly the widest vision be seen at all in the invisible
design
area because of area after vision area after
interruption adjusting adjusting
Reachability Maintenance Target locates in Maintenance Target locates Maintenance Target locates
design could reach the the envelope could not reach out of the personnel may just around the
target easily the target envelope try to approach boundary of
instinctively envelope
Operation space Maintenance Collision warning Maintenance Collision warning Maintenance Collision
design personnel could occurs hardly personnel occurs personnel always warning always
operate freely collides with sometimes collides with occurs
surroundings surroundings
sometimes
Error proofing Error proofing Safety accident Error proofing Safety accident Error proofing Safety accident Few error
design design could be reduced design is could be reduced design may happen proofing design
effectively reasonably in to a certain incompletely and probably if can be noted in
prototype extent irrationally in maintenance deserved
prototype personnel position of
operates without prototype
too much
experience
Marking design Distinct and Confused and Few marking
logical marking fragmentary design could be
marking found
Ergonomic RULA Acceptable Final score is 1 or Ergonomic Final score is 3–6 Ergonomic Final score is 7
design ergonomic 2 with green design need to be with yellow or design need to be or higher with
design color investigated brown investigated red color
further and further and
change soon change
immediately

Physical hurt Heat injury Necessary protecting devise is set Protecting devise is deficient or human Few protecting devise could be
preventing preventing appropriately and no contact happens limb contact with the dangerous discovered or human limb always
design design between human limb and dangerous surface sometimes contact with dangerous surface
surface
Electrical injury Electric power is shut down in Live working is avoided but electricity Live working is not avoided no
preventing simulation and corresponding cables contact with sharp edge in matter electricity cables wire
design electricity cables wire away from several conditions probably or not
sharp edge probably
Mechanical Sharp edge is chamfered Chamfer is missing in some necessary Few chamfer occurred
injury preventing comprehensively and precisely position
design

characters amount as the quantitative consideration for mechan- design characters amount is less than 10%), this situation is classi-
ical injury prevention design criteria. If sharp-edged objects are fied as rank C.
comprehensively and precisely chamfered (the percentage of me-
chanical injury prevention design characters amount is more than 3.2.3. Evaluation method
90%), the situation is classified as rank A. Moreover, if chamfer is Evaluation method focuses on three relevant issues: selecting
missing in several necessary positions (the percentage of mechan- snapshots for evaluation; judging the safety level for each element;
ical injury prevention design characters amount is 10%–90%), the and confirming factor weights for calculation.
situation is classified as rank B. If few chamfer in the product are The snapshots used for evaluation are determined by main-
potential hazards (the percentage of mechanical injury prevention tenance spots during a maintenance process. The maintenance
J. Geng et al. / Computer-Aided Design 45 (2013) 937–949 943

process is continuous, whereas the operation at each spot can be Table 4


regarded as a discrete point. Operation at each spot shows high Scale of relative importance.

similarity in that it focuses on hand or arm activities from main- Intensity of importance Verbal scale
tenance personnel. Other body parts are relatively fixed, and all 1 Equal importance
the maintenance operation spots occupy most of the whole main- 3 Weak importance of one over another
tenance time. Consequently, maintenance spots are selected as the 5 Essential or strong importance
snapshots are used for evaluation to comprehensively and system- 7 Demonstrated importance
9 Absolute importance
atically detect safety design flaws. Furthermore, a designer needs 2, 4, 6, 8 Intermediate value between the two adjacent
to confirm several main spots according to maintenance schedules judgment
or maintenance simulations, such as disassembling bolts, lifting
hatch covers, disconnecting cables, removing device, and so on.
Table 5
After selecting snapshots for evaluation, the safety level of each
RI value.
element is judged according to human operation analysis or pro-
n 1 2 3 4 5 6 7 8 9
totype checks of all the snapshots from the maintenance process.
RI 0 0 0.58 0.90 1.12 1.24 1.32 1.41 1.45
For each element, the selected snapshots provide different results
to reflect the corresponding product design characters. The worse
result will be treated as the safety level of each element because
The corresponding RI value is listed in Table 5.
safety accident can occur from any micro manipulation. None of
Calculate CR with Eq. (5).
the design limitation should be ignored. In this way, designers can
focus on the design flaws that reduce product safety level directly CI
and rapidly. CR = . (5)
RI
According to the aforementioned established evaluation model,
If CR ≤ 0.10, the consistency is considered acceptable [81].
the analytical hierarchical process (AHP) is adapted to confirm
(4) Suppose a goal has m criteria A1 , A2 , . . . , Am with their weights
factor weight with information from actual experience. AHP has
a1 , a2 , . . . , am , and Aj (0 < j < m) has n subcriteria
various applications since its introduction by Saaty [43–46]. Nu-
B1 , B2 , . . . , Bn with the weights b1j , b2j , . . . , bnj , Eq. (6) is in-
merous applications of this method can be found in maintenance,
troduced to obtain the weight of level B to the goal.
such as confirming the rational weight of importance of mainte-
nance priority ranking factors [74], selecting corresponding main- m

tenance policy [75], and prioritizing maintenance activities [76]. bi = bij aj , i = 1 , 2 , . . . , n. (6)
The AHP is illustrated briefly as follows. j =1

(1) Construct a hierarchy after analyzing the relationship between Calculate the total CR with Eq. (7).
the goal, criteria, and subcriteria. Thus, a flexible hierarchy can (5)
be constructed reasonably according to the problem that needs
m
to be solved. 
CI(j)aj
(2) Determine the relative importance of the alternatives with re- j =1
gard to each of the criteria or between two criteria, and gen- CR = m
. (7)
erate an n-by-n matrix A = (aij )n×n to represent the qualified RI(j)aj

judgment on pairs [77], where aij is the relative importance of j =1

criterion Ci and criterion Cj . If CR ≤ 0.10, the consistency is considered acceptable [81].


a
11 a12 ··· a1n

.. ..
a
 . ··· .
 4. Case study
A =  21  aij > 0,

 .. .. ..
. . . an−1,n Based on the established MSE model and corresponding virtual

an1 an2 ··· ann maintenance simulation for an APU disassembling process, main-
tenance safety is evaluated to determine the design flaws and op-
aji = 1/aij (i, j = 1, 2, . . . , n). (1)
timize maintenance safety design as early as possible. DELMIA was
In this process, linguistic terms generally associated to scalar selected as the virtual maintenance environment.
numerical values are adopted to express the judgment of the
decision maker [75]. The scale [78] proposed by Saaty [79] is 4.1. APU disassembling process simulation
shown in Table 4.
(3) Measure the consistency through the Consistency Ratio (CR) Maintenance personnel adjusts the supportability device to
value obtained by dividing the Consistency Index (CI) by the assure that all brackets remain properly in position, and then
corresponding Random Index (RI) evaluated by Saaty through connect the APU and the supportability device to disassemble the
the generation of a random matrix with different dimensions. connecting components between the APU and the plane. Finally,
Eq. (2) is applied to obtain the weight of each alternative, separate the supportability device from APU. The simulation
where w is an eigenvector of A with the maximum eigen- process is shown in Fig. 6 and following the principle of selecting
value n, and then calculate the maximum eigenvalue λmax from snapshot, five maintenance spots are selected as the snapshot for
Eq. (3), where w̄ is the normalized eigenvector of A and analysis.
λmax ≥ n
Aw = nw (2) 4.2. Maintenance safety analysis
Aw̄ = λmax w̄. (3)
Detailed analysis for each evaluation element is illustrated with
Calculate CI [80] with Eq. (4). figures and corresponding descriptions. The word description in
λmax − n each figure with green, yellow, and red background indicates rank
CI = . (4) A, B, and C, respectively.
n−1
944 J. Geng et al. / Computer-Aided Design 45 (2013) 937–949

Fig. 6. APU disassembling simulation process.

4.2.1. Accessibility analysis 4.3. The AHP process


Visibility analysis: All the connecting components and brackets
1 2 3 3
 
have favorable visibility design and belong to rank A of the visibility
1/2 1 4 2
criteria. The detailed visibility analysis is shown in Fig. 7, and The incidence matrix of four criteria is A = 1/3 1/4 1 2
,
visibility analysis is evaluated as rank A. 1/3 1/2 1/2 1
Reachability analysis: All the connecting components and
0.7275
brackets have favorable visibility design, except the front connect- and λmax = 4.2596, corresponding eigenvector ω = 00..5886
2845
,
ing component, as shown in Fig. 8. The target is located around 0.2083
the boundary of the envelope and belongs to rank C, whereas the 0.4022
others belong to rank A. Thus, reachability analysis is evaluated as 0.3254
and normalized eigenvector ϖ = 0.1573
.
rank C.
0.1151
Operation space analysis: As shown in Fig. 9, collision occurs The incidence matrix of the elements under each criterion is as
hardly when operating on the bracket and two connecting compo- follows.
nents outboard belongs to rank A. Collision occurring sometimes
  1 2 1
when operating on the two interior brackets belonging to rank B, The incidence matrix of accessibility criteria is B1 = 1/2 1 3 ,
1 1/3 1
whereas collision occurring always when operating on the three 0.8257 
interior connecting components and front connecting component λ1 = 3.0536, corresponding eigenvector ω1 = 0.5201 , and
belongs to rank C. 0.2184
0.5279
Thus, operation space analysis is evaluated as rank C because of
the frequent collision. normalized eigenvector ϖ1 = 0.3325 .
0.1396
 
1 3
4.2.2. Error proofing analysis The incidence matrix of error proofing criteria is B2 = 1/3 1
,
0.9487
 
The two terminal sockets in the pump under APU need an error λ2 = 2, corresponding eigenvector ω2 = , and normalized
0.3162
proofing design because of the same radius, as shown in Fig. 10.
0.7441
 
Error proofing designs irrational and no marking design can be eigenvector ϖ2 = 0.2559 .
found in the prototype, so error proofing analysis is evaluated as
The incidence matrix of ergonomic criteria is B3 = [1] , λ3 =
rank B and the marking analysis is evaluated as rank C.
1, ω3 = [1], and normalized eigenvector ϖ3 = [1].
The incidence matrix
 of physical harm prevention criteria is
4.2.3. Ergonomic analysis 1 6 1/3
All the ergonomic analysis have an unfavorable final score B4 = 1/6 1 3 λ4 = 3.0183, corresponding eigenvector
of 7 because of the arm supporting and checking balance when
3
0.6323 1/3 1 0.3874
maintenance personnel stand on the support device, as shown in ω4 = 0.7238 , and normalized eigenvector ϖ4 = 0.4434 .
Fig. 11. However, the outboard primary connecting component has 0.2762 0.1692
a final score of 5. Thus, ergonomic analysis is evaluated as rank C. Consistency calculation.
For matrix A, CI = λmax −n
n −1
= 4.25964−1
−4
= 0.0865, and RI = 0.90;
0.0865
4.2.4. Physical hurt analysis thus, CR = RI = 0.9 = 0.096 < 0.1.
CI

No heat source exists in APU and no electronic operation Consequently, the consistency is considered acceptable.
occurs in the simulation. Thus, heat injury prevention design and Similarly, for matrix B1 , B2 , B3 , B4
electronic injury prevention design are configured as rank B to CR1 = 0.0462 < 0.1
represent an average design. All the sharp edges on the equipment
and frame around the operation spot and all the contact on the CR2 = 0 < 0.1
support device are designed without chamber, as shown in Fig. 12. CR3 = 0 < 0.1
Thus, mechanical harm analysis is evaluated as rank C. CR4 = 0.0158 < 0.1.
J. Geng et al. / Computer-Aided Design 45 (2013) 937–949 945

Fig. 7. Visibility analysis.

Fig. 8. Reachability analysis.


946 J. Geng et al. / Computer-Aided Design 45 (2013) 937–949

Fig. 9. Operation space analysis.

Table 6 Ranks A, B, and C are assigned 1, 2, and 3 scores, respectively. A


Maintenance safety analysis of the case study. lower score indicates a more favorabledesign.
0.5279

Criteria Element Classification
= 1.9442, and the
 
For accessibility, R1 = 1 3 3 0.3325
Accessibility Visibility A 0.1396
Reachability C analysis result is between A and B. 
Operation space C  0.7441
For error proofing, R2 = 2 3 0.2559 = 2.2559; thus, the

Error proofing Error proofing B
Marking C
analysis result is between B and C.
Ergonomic RULA C
Physical hurt prevention Heat injury B For ergonomic, R3 = (3)(1) = 3; thus, the analysisresult 
is C.
Electrical injury B   0.3874
For physical hurt prevention, R4 = 2 2 3 0.4434 =
Mechanical injury C
0.1692
2.1624, and the ergonomic analysis result is between B and C.
All the consistency is considered acceptable. Consequently, for maintenance safety,
Construct a matrix ω′ with ω1 , ω2 , ω3 , ω4 R4 ω
 
R = R1 R2 R3
0.5279 0.7441 1 0.3874 0.4022
   
w =

0.3325 0.2559 0 0.4434  0.3254
= 1.9442 2.2559 2.1624 

0.1396 0.1692 3
0 0 0.1573
0.1151
and
= 2.2368.
0.5279 0.7441 1 0.3874
 
w ω = 0.3325 0.2559 0 0.4434
′ The maintenance safety analysis result is between B and C,
0.1396 0 0 0.1692 and close to rank B. Appropriate improvement can be carried
0.4022 out pertinently according to the four criteria analysis results and
 
0.6563
 
0.3254 maintenance safety analysis result, respectively, to enhance the
× = 0.2680
0.1573 maintenance safety design.
0.0756
0.1151
CR = 0.0462 + 0 + 0 + 0.0158 = 0.062 < 0.1. 5. Conclusion and discussion

Consequently, the consistency is considered acceptable. In this paper, a maintenance safety evaluation model applicable
The maintenance safety analysis is summarized in Table 6. to the virtual maintenance environment was proposed to evaluate
J. Geng et al. / Computer-Aided Design 45 (2013) 937–949 947

Fig. 10. Error proofing analysis.

Fig. 11. Ergonomic analysis.

maintenance safety design with the help of maintenance simula- the problem. Second, safety analysis at each snapshot aims at
tion and prototype check. The presented evaluation model includes the relatively static virtual human, whereas the postural variation
evaluation element, evaluation criteria, and evaluation methodol- happens obviously during maintenance process. Thus, the amount
ogy after fully considering the characters and framework of the vir- of slight variation affecting the final result of each snapshot should
tual maintenance environment. The advantages of the proposed be considered. Third, the dummy in VME is nude and does not
system are summarized as follows: (1) Launch the maintenance reflect the restrictions clothing impose, such as reachability and
safety evaluation in the product design stage to improve the main- ergonomics. If clothing is considered, partial analysis will slightly
tenance safety design and decrease the possibility of accidents. vibrate. Recent research has focused on novel product prototypes
(2) Maintenance safety evaluation work performed intuitively via from different domains in the laboratory. Parts of those users have
virtual maintenance technology. (3) Combine static check and dy- provided positive feedback for improving the maintenance safety
namic simulation from maximizing prototype to determine the design against similar products. Safety maintenance requirement
design flaws. (4) Evaluate the qualitative maintenance element is essential, and more feedback information including past accident
quantitatively with AHP to provide more purposeful advice for de- records is needed to validate the proposed method in this
sign improvement. paper. Ongoing and future work will focus on improvements
However, several limitations exist in the proposed evaluation including virtual manikin anthropometry, slight variation at
model. First, only the 50th percentile anthropometric data was each snapshot, and clothing consideration. Furthermore, more
used with human models. Thus, the diversity of virtual dummy evaluation elements in the virtual maintenance environment and
anthropometry was ignored. Several amendment methods or corresponding criteria will be added to the existing model if
appropriate multivariate data and techniques are needed to solve possible.
948 J. Geng et al. / Computer-Aided Design 45 (2013) 937–949

Fig. 12. Physical hurt analysis.

Acknowledgments [12] Abdul-Nour Georges, Demers Michel, Vaillancourt Raynald. Probabilistic


safety assessment and reliability based maintenance policies: application to
the emergency diesel generators of a nuclear power plant. Comput Ind Eng
The authors thank the reviewers for their careful evaluation and 2002;42:433–8.
thoughtful comments. The authors also thank the National Natural [13] Sachon M, Paté-Cornell E. Delays and safety in airline maintenance. Reliab Eng
Science Foundation of China (No. SKVR-10-17) for financially Syst Saf 2000;67:301–9.
[14] Beugin J, Renaux D, Cauffriez L. A SIL quantification approach based on
supporting this research. an operating situation model for safety evaluation in complex guided
transportation systems. Reliab Eng Syst Saf 2007;92:1686–700.
[15] Marono M, Pena JA, Santamaria J. The ‘PROCESO’ index: a new methodology
References for the evaluation of operational safety in the chemical industry. Reliab Eng
Syst Saf 2006;91:349–61.
[1] Li JR, Khoo LP, Tor SB. Desktop virtual reality for maintenance training: an [16] Flammini Francesco, Marrone Stefano, Mazzocca Nicola, Vittorini Valeria. A
object oriented prototype system (V-REALISM). Comput Ind 2003;52:109–25. new modeling approach to the safety evaluation of N-modular redundant
[2] Kara Ali, Ozbek MehmetEfe, Cagiltay NergizErcil, Aydin Elif. Maintenance, computer systems in presence of imperfect maintenance. Reliab Eng Syst Saf
sustainability and extendibility in virtual and remote laboratories. Procedia 2009;94:1422–32.
Soc Behav Sci 2011;28:722–8. [17] Haber S, O’Brien JN, Methlay DS, Crouch DA. Influence of organizational factors
[3] Pouliquen Mamy, Bernard Alain, Marsot Jacques, Chodorge Laurent. Virtual on performance reliability. In: NUREG/CR-5538. 1991.
hands and virtual reality multimodal platform to design safer industrial [18] Apostolakis GE, Davoudian K, Jya-Syn W. The work process analysis model
systems. Comput Ind 2007;58:46–56. (WPAM). Reliab Eng Syst Saf 1994;45:107–25.
[4] Esque S, Mattila J, Saarinen H, Siuko M, Virvalo T, Muhammad A, et al. The use of [19] Baumond G, Menage F, Scheiter JR, Spurgin A, Vogel A. Quantifying human
virtual prototyping and simulation in ITER maintenance device development. and organizational factors in accident management using decision trees: the
Fusion Eng Des 2007;82:2073–80. HORAAM method. Reliab Eng Syst Saf 2001;70:113–24.
[5] Ng AmosHC, Adolfsson Josef, Sundberg Martin, De Vin LeoJ. Virtual manufac- [20] Van Steen JFJ, Brascamp MH. On the measurement of safety performance.
turing for press line monitoring and diagnostics. Int J Mach Tools Manuf 2008; In: Proceedings of the Eighth International Symposium on Loss Prevention and
48:565–75. Safety Promotion in the Process Industries. 1995.
[6] Elzendoorn Ben, de Baar Marco, Chavan Rene, Goodman Timothy, Heemskerk [21] Papazoglou IA. Management factors in process plant safety. Technical note
Cock. Analysis of the ITER ECH upper port launcher remote maintenance using I.94.102, ISEI/IE/2660/94. 1994.
virtual reality. Fusion Eng Des 2009;84:733–5. [22] Pate Cornell ME, Murphy DM. Human and management factors in probabilistic
[7] Liu Xinhua, Peng Gaoliang, Liu Xiumei, Hou Youfu. Development of a risk assessment: the SAM approach and observations from recent applications.
collaborative virtual maintenance environment with agent technology. Reliab Eng Syst Saf 1996;53.
J Manuf Syst 2010;29:173–81. [23] Oien K. A framework for the establishment of organizational risk indicators.
[8] Ieronutti Lucio, Chittaro Luca. Employing virtual humans for education and Reliab Eng Syst Saf 2001;74:147–67.
training in X3D/VRML worlds. Comput Educ 2007;49:93–109. [24] Kelly TP, McDermid JA. A systematic approach to safety case maintenance.
[9] Wang J, Yang JB, Sen P, Ruxton T. Safety based design and maintenance Reliab Eng Syst Saf 2001;71:271–84.
optimization of large marine engineering systems. Appl Ocean Res 1996;18: [25] Marseguerra Marzio, Zio Enrico, Podofillini Luca. A multiobjective genetic
13–27. algorithm approach to the optimization of the technical specifications of a
[10] Hale AR, Heming BHJ, Smit K, Rodenburg FGTh, van leeuwen ND. Evaluating nuclear safety system. Reliab Eng Syst Saf 2004;84:87–99.
safety in the management of maintenance activities in the chemical process [26] Gauthier François, Giraud Laurent, Bournival Sylvain, Bourbonnière Réal,
industry. Saf Sci 1998;28(1):21–4. Richard Jean-Guy, Daigle Renaud, et al. Development of a loading dock safety
[11] Duncan H, Ken Brewer, Canady S. Probabilistic safety assessment support for evaluation tool. J Saf Res 2007;38:35–51.
the maintenance rule at Duke Power Company. Reliab Eng Syst Saf 1999;63: [27] Carnero Ma Carmen. An evaluation system of the setting up of predictive
243–9. maintenance programs. Reliab Eng Syst Saf 2006;91:945–63.
J. Geng et al. / Computer-Aided Design 45 (2013) 937–949 949

[28] Wei Chunfu, Pei Zheng, Li Huamin. An induced OWA operator in coal mine [54] Zhang Peihe, Jin Xiuliang, Liu Yuhui, Wang Zhengxi, Liu Nana. Synthetical
safety evaluation. J Comput System Sci 2011; analysis on geological factors controlling coalbed methane. Procedia Earth
http://dx.doi.org/10.1016/j.jcss.2011.11.003. Planet Sci 2011;3:144–53.
[29] Lu Lixuan, Jiang Jin. Analysis of on-line maintenance strategies for k-out-of-n [55] Skerman HelenM, Yates PatsyM, Battistutta Diana. Identification of cancer-
standby safety systems. Reliab Eng Syst Saf 2007;92:144–55. related symptom clusters: an empirical comparison of exploratory factor
[30] Bamber L. Principles of accident prevention. In: Ridley J, editor. Sufify at work. analysis methods. J Pain Symptom Manage 2012;44(1):10–22.
London: Butterworths; 1986. p. 131–43. [56] Afkhami Abbas, Khajavi Farzad, Khanmohammadi Hamid. Investigation of
[31] Zhao Yuyang, Li Ming, Yuan Yi. Operation and maintenance integration to oxidation and tautomerization of a recently synthesized Schiff base in
improve safety. Comput Chem Eng 2000;24:401–7. micellar media using multivariate curve resolution alternative least squares
[32] Elvik Rune. Economic deregulation and transport safety: a synthesis of and rank annihilation factor analysis methods. Anal Chim Acta 2009;647:
evidence from evaluation studies. Accid Anal Prev 2006;38:678–86. 189–194.
[33] Senge PM. La quintadisciplina, el arte y la practica de la organizacion abierta [57] Vinayak DibakarSen. A vision modeling framework for DHM using geometri-
alaprendizaje. Buenos Aires: Ediciones Juan Granica SA; 1992. cally estimated FoV. Comput Aided Des 2012;44:15–28.
[34] Benedyk Rachel, Minister Sarah. Applying the BeSafe method to product safety [58] Anderson JR, Matessa M, Lebiere C. Act-r: a theory of higher level cognition
evaluation. Appl Ergon 1998;29:5–13. and its relation to visual attention. Human Comput Interact 1997;12(4):
[35] Pedersen LM. Reply to letter regarding realistic evaluation as a new way to 439–462.
design and evaluate occupational safety interventions. Saf Sci 2011; [59] Itti L, Koch C. Computational modelling of visual attention. Nat Rev Neurosci
http://dx.doi.org/10.1016/j.ssci.2011.11.016. 2001;2:194–203.
[36] Sachon M, Paté-Cornell E. Delays and safety in airline maintenance. Reliab Eng [60] Liu Hantao, Heynderickx Ingrid. A simplified human vision model applied to a
Syst Saf 2000;67:301–9. blocking artifact metric. In: Computer analysis of images and patterns. 2007.
[37] Neitzel RichardL, Seixas NoahS, Harris MichaelJ, Camp Janice. Exposure to fall p. 334–41.
hazards and safety climate in the aircraft maintenance industry. J Saf Res 2008; [61] Neokleous KleanthisC, Schizas ChristosN. Computational modeling of visual
39:391–402. selective attention. Procedia Comput Sci 2011;72:44–245.
[38] Ural Suphi, Demirkol Sitki. Evaluation of occupational safety and health in [62] Georgeson MarkA. Edge-finding in human vision: a multi-stage model based
surface mines. Saf Sci 2008;46:1016–24. on the perceived structure of plaids. Image Vis Comput 1998;16:389–405.
[39] Reiman Teemu, Oedewald Pia. Assessing the maintenance unit of a nuclear [63] Bijaoui Albert, Rue Frkdkic. A multiscale vision model adapted to the
power plant-identifying the culture conceptions concerning the maintenance astronomical images. Signal Process 1995;46:345–62.
work and the maintenance organization. Saf Sci 2006;44:821–50. [64] Bullock MargaretI. The determination of functional arm reach boundaries for
[40] Cheng Dong-quan, Gu Feng. Application of improved grey correlative method operation of manual controls. Ergonomics 1974;17(3).
in safety evaluation on fully mechanized mining faces. Procedia Earth Planet [65] Woodson WesleyE. Human factors design handbook. McGram-Hill; 1981.
Sci 2011;52:58–63. [66] Nowak Ewa. Determination of the spatial reach area of the arms for workplace
[41] Liu Shuang, Hurley Michael, Lowell KimE, Siddique Abu-BakerM, Diggle Art, design purposes. Ergonomics 1987;21(7):493–507.
Cook DavidC. Using an integrated fuzzy set and deliberative multi-criteria [67] Bez HE, Bricis AM, Ascough J. A collision detection method with applications in
evaluation approach to facilitate decision-making in invasive species manage-
CAD systems for the apparel industry. Comput Aided Des 1996;28(1):27–32.
ment. Ecol Econ 2010;69:2374–82.
[68] Chang Jung-Woo, Wang Wenping, Kim Myung-Soo. Efficient collision
[42] Wen Kun-Li. A Matlab toolbox for grey clustering and fuzzy comprehensive
detection using a dual OBB-sphere bounding volume hierarchy. Comput Aided
evaluation. Adv Eng Softw 2008;39:137–45.
Des 2010;42:50–7.
[43] Ossadnik W, Lange O. AHP-based evaluation of AHP-software. European J Oper
[69] Jia Xiaohong, Choi Yi-King, Mourrain Bernard, Wang Wenping. An algebraic
Res 1999;118:578–88.
approach to continuous collision detection for ellipsoids. Comput Aided Geom
[44] Triantaphyllou E, Mann SH. Using the analytic hierarchy process for decision
Design 2011;28:164–76.
making in engineering applications: some challenges. Int J Ind Eng Appl Pract
[70] Chen Ying, Shen Li-Yong, Yuan Chun-Ming. Collision and intersection
1995;2(1):35–44.
[45] Mohanty RP, Deshmukh SG. Advanced manufacturing technology selection: detection of two ruled surfaces using bracket method. Comput Aided Geom
strategic model for learning and evaluation. Int J Prod Econ 1998;55:295–307. Design 2011;28:114–26.
[46] Bhutta KS, Huq F. Supplier selection process: a comparison of the total cost of [71] Karhu O, Kansi P, Kuofina I. Correcting working postures in industry: a practical
ownership and the analytic hierarchy process approaches. Int J Supply Chain method for analysis. Appl Ergon 1977;8(4):199–201.
Manag 2002;3:126–35. [72] Kee D, Karwowski W. LUBA: an assessment technique for postural loading on
[47] Yari G, Chaji AR. Maximum Bayesian entropy method for determining ordered the upper body based on joint motion discomfort and maximum holding time.
weighted averaging operator weights. Comput Ind Eng 2012;63:338–42. Appl Ergon 2001;32(4):357–66.
[48] Zou Zhi-Hong, Yun Yi, Sun Jing-Nan. Entropy method for determination [73] McAtamney L, Corlett EN. RULA: a survey method for the investigation of
of weight of evaluating in fuzzy synthetic evaluation for water quality work-related upper limb disorders. Appl Ergon 1993;24(2):91–9.
assessment. J Environ Sci 2006;18(5):1020–3. [74] Ramadhan RH, Wahhab HIA, Duffuaa SO. The use of an analytical hierarchy
[49] Elangovan M, Babu Devasenapati S, Sakthivel NR, Ramachandran KI. Evalua- process in pavement maintenance priority ranking. J Qual Maint Eng 1999;
tion of expert system for condition monitoring of a single point cutting tool 5(1):25–39.
using principle component analysis and decision tree algorithm. Expert Syst [75] Emblemsvag J, Tonning L. Decision support in selecting maintenance
Appl 2011;38:4450–9. organization. J Qual Maint Eng 2003;9(1):11–24.
[50] Schweizer Katrin, Cattin PhilippeC, Brunner Reinald, Müller Bert, Huber Cora, [76] Shen Q, Lo K, Wang Q. Priority setting in maintenance management: a modified
Romkes Jacqueline. Automatic selection of a representative trial from multiple multi-attribute approach using analytic hierarchy process. Constr Manag Econ
measurements using Principle Component Analysis. J Biomech 2012;45: 1998;16:693–702.
2306–9. [77] Ramadhan RH, Wahhab HIA, Duffuaa SO. The use of an analytical hierarchy
[51] Ren Song, Bai Yue-Ming, Jiang De-Yi, Li An Yu. An extension evaluation model process in pavement maintenance priority ranking. J Qual Maint Eng 1999;
for building safety under the influence of blasting seismic. Procedia Eng 2012; 5(1):25–39.
43:168–73. [78] Triantaphyllou E, Lootsma FA, Pardalos PM, Mann SH. On the evaluation and
[52] Hwang CG, Ingraffea AR. Virtual crack extension method for calculating the application of different scales for quantifying pairwise comparisons in fuzzy
second order derivatives of energy release rates for multiply cracked systems. sets. J Multi Crit Decis Anal 1994;3(3):133–55.
Eng Fract Mech 2007;74:1468–87. [79] Saaty TL. The analytic hierarchy process. New York: McGraw Hill; 1980.
[53] Lei Zhang. Grey synthetical evaluation of university’s engineering innovation [80] Saaty TL. The analytic network process. Pittsburg: RWS Publications; 2001.
ability. Syst Eng Procedia 2012;3:319–25. [81] Forman E, Selly MA. Decision by objectives. London: World Scientific; 2001.

You might also like