Professional Documents
Culture Documents
Physical and Cognitive Effects of Virtual Reality Integrated Training
Physical and Cognitive Effects of Virtual Reality Integrated Training
Integrated Training
Richard T. Stone, Kristopher P. Watts, Peihan Zhong, and
Chen-Shuang Wei, Iowa State University, Ames
assess the impact of VR integrated training in & Chryssolouris, 2006). There are many studies
the context of absolute performance, training that have compared VR training with traditional
time, and physical and cognitive development. training in fields other than welding (e.g., Ye,
The VR system used in this study provides the Banerjee, Banerjee, & Dech, 1999), particularly
user with a fully immersive environment that in the field of laparoscopic surgery. Physicians
included manipulation of physical weld tools. have been trained in laparoscopic procedures
The use of feed-forward visual overlays and with the use of the VR laparoscopic trainer
postweld feedback in the VR system allowed MIST VR (Wilson, Middlebrook, Sutton, Stone,
the user to improve specific aspects of their & McCloy, 1997). Several laparoscopic VR
welds during training. This level of oversight studies have failed to find a significant differ-
and guidance is simply not possible during nor- ence between VR training and traditional train-
mal weld training because of environmental ing (e.g., Munz, Kumar, Moorthy, Bann, &
factors and time constraints. We believe that Darzi, 2004; Torkington, Smith, Rees, & Darzi,
such a tool has great potential to revolutionize 2000), whereas others have found that VR train-
the learning process in such a way as to exceed ing is superior to traditional training in many
traditional training paradigms. respects (Gurusamy, Aggarwal, Palanivelu, &
We hypothesize that the use of VR integrated Davidson, 2008).
training will result in student performance (cer- Laparoscopic surgery and welding tasks are
tifications and quality of weld) that is equal to dissimilar with regard to the overall procedural
or better than that of students trained by tradi- goals; however, we believe that they are similar
tional methods. Additionally, we hypothesize in the complex nature of physical movement
that the use of VR integrated training will and kinesthetic memory that are essential for
greatly reduce overall student training times execution of both tasks. Jobs requiring regular
when compared with traditional training meth- performance of highly nuanced actions, such as
ods. It is expected that the physical develop- suturing or welding, necessitate the honing of
ment of students trained with the use of VR highly specific physical movements. Welding
integrated training will be similar to that of stu- also requires that the welder have a sufficient
dents trained by traditional methods and that knowledge base to be able to judge variables
both groups will differ similarly from domain related to creating a structurally sound weld.
experts. It is further hypothesized that the cog- To gauge the effectiveness of VR integrated
nitive development of students trained with VR training with respect to traditional training sys-
integrated training will be similar to that of stu- tems, the cognitive, physical, and performance
dents trained by traditional methods. Last, we impacts should all be addressed. The primary
hypothesize that mental workload (MWL) will performance measure of interest in welding
not be significantly different between students training is the certification of the welders. Past
trained with VR integrated training and tradi- studies have measured welds on the basis of
tionally trained students. We believe the find- specific criteria (Casalino et al., 2004; Kannan
ings of this study will be directly applicable to & Murugan, 2006; Reyes-Valdés, Torres-
most, if not all, similar integrated VR training Treviño, Lopez-Cortéz, de la Cruz-Márquez,
regimens, particularly if the task has both physi- & Briceño-Ramirez, 2006), which are part of
cal and cognitive components. the American Welding Society (AWS) certifi-
cation requirements. For this reason, all stu-
Related Work dent welders in this study were trained by an
Very few studies have compared the effec- AWS certified welding instructor (CWI) and
tiveness of VR training in the context of a com- were given the opportunity to certify through
plex skill domain, such as welding. Those studies the AWS. Establishing the effectiveness of a
that have been conducted are limited in scope training tool in the context of its primary goal
and focus primarily on the VR technology as (weld certifications) is a valid metric for estab-
opposed to the human components (Fast, Gifford, lishing training impact. Furthermore, the CWI
& Yancey, 2004; Mavrikios, Karabatsou, Fragos, judged welds produced in this study on the
basis of the quality of the weld and graded each attended a formal weld school and earned several
individual weld on the basis of AWS criteria welding certifications and qualifications. In
for weld quality. addition, several other investigators were trained
However, certifications themselves do not and have achieved welding certifications and
explain how human development is affected by qualifications. Investigators conducted 16 inter-
the use of VR systems. For this reason, it is crit- views with experienced welders, welding
ically important to evaluate the cognitive impact instructors, and welding engineers. A detailed
associated with VR training environments biomechanical evaluation of eight expert weld-
(Wickens & Baker, 1995; Zanbaka, Lok, Babu, ers performing the weld types discussed in this
Ulinski, & Hodges, 2005). Most studies com- article was also conducted. The eight experts
paring VR and traditional training have looked were fitted with electromyograph (EMG) elec-
only at the performance and cognitive aspects trodes and were asked to perform all of the weld
associated with training (e.g., Munz et al., 2004; tests examined in this article. These pilot data
Seymour et al., 2002). These prior studies did were used to refine the study protocol to ensure
not address the fact that a human’s physical or technical accuracy and applicability to the field
kinesthetic memory development can be an of welding.
important developmental factor, depending on
the type of training being conveyed. Kinesthetic Experimental Materials and
memory refers to the extent that the human Setting
body can recall its movements and postures. A traditional welding facility was con-
Through the use of sensory-motor learning and structed on the Iowa State University campus
development of kinesthetic memory, a person (see Figure 1). Specifically, a 740-square-foot
can accomplish specific physical movements room housed six welding booths. Each booth
without thinking about how his or her body’s was equipped with the following: a new Lincoln
parts should move (Ebert, Deller, Steffen, & Electric Power MIG 350MP welder with stick
Heintz, 2009). Interviews with 16 professional metal arc welding (SMAW) attachments, two
welders and welding engineers revealed that auto-adjusting welding helmets, welding jack-
they felt that sensory-motor learning was a very ets and gloves, power grinders, slag hammer,
important component of successful welding. wire brushes, welding table, quenching buck-
Studies have demonstrated differences in ets, and other miscellaneous welding equip-
muscle activity between expert and novice ment. Consumables, welding plates, and
welders (Keir & MacDonell, 2004). According electrodes were also made available to the
to previous research (Beauchamp, Marchand, trainees. The welding facility was outfitted with
& Galopin, 1997; Kadefors, 1976) and feed- 1,200 flat stock plates, 500 groove plates, and
back and observations gathered from experts, eight thousand 7018 electrodes.
muscles that are of significant importance to One floor below the traditional welding
welding performance include the deltoid, trape- facility, a VR weld training facility was con-
zius, extensor digitorum, and flexor carpi ulna- structed. A 500-square-foot room was converted
ris. The activation and interactions of these to house weld booths of the same size and
muscles serve to define expert welder control, dimensions as their traditional counterparts (see
ability, and stability during the commission of a Figure 1). Each booth contained a new VRTEX
weld. 360 VR Welding Trainer with SMAW attach-
ments and welding jackets and gloves. The
Method
VRTEX 360 trainer was chosen because it is the
Prestudy highest-fidelity VR simulator currently avail-
Prior to the experiment, ethnographic evalua- able and allows users to be fully immersed in a
tion and expert interviews as well as observa- 3-D VR environment while conducting welds.
tional studies and limited controlled experiments The VR helmet on the VRTEX 360 features ste-
were performed to gather information on weld- reo sound, with stereovision super video graph-
ing and weld training. The principal investigator ics array (800 × 600 pixels at 24-bit depth with
Figure 1. Iowa State University welding school: (a) traditional welding booth; (b) virtual welding booth.
200:1 high contrast) goggles that provide a 40° that 50% VR integration represented the higher
diagonal field of view, allowing for very realis- end of implementation that they would be com-
tic image sizes (105 inches at 12 feet). fortable accepting for use in near-term welding
Although the system does not simulate heat schools.
or friction forces associated with welding, it
does require the user to use and manipulate Participants
weld tools in the same way he or she would in There were 22 participants in total (21 males
real-world welding. The VR tables were set at and 1 female). None of the participants had any
the same height as the traditional tables. Hence, prior practical welding exposure or experience
the virtual training systems created a very high- in SMAW welding prior to the beginning of the
fidelity situation, in which the user wore a weld study. All participants committed to 80 training
helmet, used a SMAW weld attachment (of the hours in 2 weeks. The participants were ran-
same size and dimensions as a real weld attach- domly assigned to either the TW or the VR50
ment), and could hold placed weld stock (the group. The TW group had an average age of 44
plates) as would a real-world welder. In addi- (SD = 13) years, average height of 70.2 (SD =
tion, the VR system has an advanced physics 3.9) inches, and average weight of 215.3 (SD =
engine that allowed for the execution of real- 26.3) pounds. The VR50 group had an average
time welds that were virtually indistinguishable age of 41 (SD = 13.6) years, average height of
from welds made in the real world. 70.2 (SD = 2.4) inches, and average weight of
In this study, two groups were assigned. One 228.6 (SD = 46.7) pounds.
was trained with 100% traditional training
(TW), and the other spent 50% of its time in Independent and Dependent
traditional training and 50% of its time in the Variables
virtual environment (VR50). A 50:50 ratio was The primary independent variable in this
chosen because it gave experimenters ample experiment was training type at two levels, rep-
time to collect detailed information and because resenting the type of interface being tested: TW
interviews with welding instructors indicated and VR50. There were five dependent measures
Figure 2. The four certifiable weld positions in this study, depicted in order of increasing difficulty.
to ensure that there was no contact between between traditional training and VR integrated
VR50 and TW participants. Before participants training was in the training system itself.
arrived at the test site, they were randomly Participants in the VR50 group spent only 50%
assigned to either the VR50 or the TW group. of their time training (lectures and practical lab
Prior to experimentation, all participants were training) with the direction of a CWI for each
given informed consent forms, followed by weld type. The remaining 50% of their time was
individual screening tests to ensure that they spent training on the VRTEX 360 system. During
possessed normal visual acuity, depth percep- this time, the VR system itself served as the
tion, and hearing. instructor by providing feedback after every
In the traditional welding school (TW group), weld and by providing optional use of visual
participants were trained in the application of overlay that would guide the user to improve key
welding techniques starting with the simplest aspects of his or her weld, such as travel speed,
weld (2F) and proceeding through the most dif- work and travel angle, and arc length.
ficult weld (3G). The time allotted to teaching During VR training time, the participants (in
each weld was fixed and included formal lec- pairs) used the VRTEX 360 to conduct virtual
tures and practical lab training conducted by a welds of each of the four weld types. If a par-
fully certified AWS CWI. Following the train- ticipant was able to earn a machine-generated
ing for each weld type, participants were given quality score of 85% at least twice for a weld,
a single weld certification test piece. All weld he or she was permitted to discontinue VR train-
tests (2F, 3F, 1G, and 3G) were performed in the ing time early. If the VR50 participant reached
presence of the CWI. When completed, test the total time allotted for VR training (equal to
pieces underwent an on-site visual inspection 50% of the TW group’s total allotted training
by the CWI. If the test piece passed visual time), he or she would be moved to the real-
inspection, it was then sent to an independent world training regardless of whether a VR train-
laboratory for structural testing. Certification or ing score of 85% or better had been achieved. In
failure for the participant was based on the this study, it was rare for a participant to fail to
results of this structural testing. achieve a VR training score of ≥85%; this situ-
During practical lab training, participants ation occurred in only eight instances. After
were fitted with electrodes, enabling experi- participants had moved on from VR to real-
menters to record EMG data while participants world training, they were not allowed to return
conducted welds. The EMG recorded activity of to the virtual training. Furthermore, they were
the previously discussed muscles of interest. allowed to spend only as much time in real-
During practical lab training periods, the CWI world training as they had used in the VR train-
evaluated practice welds to determine whether a ing; this rule was instituted to ensure the 50:50
participant was ready to be tested prior to the training time ratio.
end of his or her total allotted training time. If
Results
the total allotted time had been used, the partici-
pants were required to test regardless of whether Certification and Quality
the CWI had determined the participant ready The number of certifications achieved by
to test. Immediately following the certification participants for each of the four weld types was
test for each of the four weld types, participants tallied for each of the two training conditions.
were given a written cognitive test related to the Figure 3 shows the total number of certifica-
welding unit used and the weld type they had tions or each weld type by training condition (a
just performed. Further, a NASA-TLX test was maximum of 11 certifications was possible for
also administered to measure MWL. any given weld type). Because of the categori-
In the VR integrated welding training (VR50 cal nature of the data, chi-square analysis was
group), the experiment was conducted in the used to look for significant differences between
same basic manner as in the TW group. Both groups for each of the four weld types. The chi-
groups were given the same total allotted training square results were 2.2, 3.667, 1.692, and 0.188
time for each weld type. The major difference for 2F, 1G, 3F, and 3G, respectively, all of which
Cerficaon Rate
12 with the TW group across all weld types: 2F
11 Group 1 (TW)
10 Group 2 (VR50) (T0.05, 1, 20 = 6.367, p ≤ .001), 1G (T0.05, 1, 20 =
10
3.468, p = .002), 3F (T0.05, 1, 20 = 3.340, p = .003),
# Parcipants Cerfied
9
8
8 and 3G (T0.05, 1, 20 = 3.015, p = .007).
6
6
5
Cognitive Development
5
4
4 Cognitive development was assessed across
four categories (knowledge, comprehension,
2 application, and analysis), each representing a
0 different aspect of cognitive capability. Figure 5
2F 1G 3F 3G shows the mean scores within each of the four
Weld Type categories of cognitive development by weld
type for both groups. To determine whether the
Figure 3. Number of certifications awarded by weld experimental groups were significantly differ-
type (in order of increasing difficulty). ent with regard to cognitive development, a T test
was conducted for each question type within
each weld type (2F, 3F, 1G, and 3G). The T test
were less than 3.841, χ2(1, 0.05).The number of compared the knowledge scores, comprehen-
certifications earned for the VR50 and TW sion scores, application scores, and analysis
groups were not found to be significantly dif- scores for the groups within each weld type.
ferent given α = .05 for the 2F, 1G, 3F, or 3G The results of this analysis can be seen in Table
weld types. Although no significance was found 2. Each of the tests was conducted with an α =
via the chi-square test, the result for 1G weld .05, and degrees of freedom of 20 (with the
outcomes was very close to 3.841, which indi- exception of the 2F test, which has only 19
cated that the VR50 group had a practically degrees of freedom since 1 participant was
higher certification rate than did the TW group. given a test in the wrong order).
It should be noted that in all cases, the VR50 Our results indicate four instances of signifi-
group showed a descriptively higher rate of cance within the scores of the Crooks’s taxon-
certifications than the TW group did. omy. Of these results, two are higher for the
In addition, each of the weld tests was accom- VR50 group, and two are higher for the TW
panied by an overall quality score ranging from 0 group. The cognitive development areas with
to 100. These scores were based on structural higher VR50 scores include 2F comprehension
variables (bend tests) as well as visual parame- and 3F analysis. The cognitive development
ters (dimensions or measurements of the weld). areas with a higher TW scores are 3F applica-
The mean quality score for both groups are tion and 3G comprehension.
shown in Table 1. Inferential analysis did not
show a significant difference in quality between Physical Development
the two groups in terms of the 2F, 3F, and 3G Physical development was assessed with
weld types. However, the VR50 group was found respect to the average muscle activity expressed
to have significantly higher quality scores for the as a percentage of MVC for the interaction of
1G weld type (T0.05, 1, 20 = −2.237, p = .037). the four muscles of interest (deltoid, trapezius,
extensor digitorum, and flexor carpi ulnaris
Training Time muscles) within each of the weld types for
The total amount of time used by participants expert welders and both of the experimental
to complete their training and certification test groups (VR50 and TW). Figure 6 shows the
for each weld type was averaged for both exper- muscle activity interaction profiles for each of
imental groups. Figure 4 shows the mean task the four weld types.
times for both training conditions by weld type. MANOVA was used to account for the mul-
Analysis revealed that the VR50 group had sig- tiple dependent variables (four muscles) that
nificantly shorter training times as compared define kinesthetic memory development in this
Weld Type
2F 1G 3F 3G
Group M SD M SD M SD M SD
VR50 92 9 88 10 81 16 61 24
TW 86 20 70 26 73 23 62 23
Note. Values based on Bloom, Englehart, Furst, Hill, and Krathwohl (1956).
VR50 = virtual reality integrated training; TW = traditional welding training.
12
10 The results of the 3F weld type showed no
8 significant difference between the three condi-
6 tions, F(8, 46) =1.1931, p = .3240. The out-
4 comes from the 3G analysis also showed no
2
0
difference between any of the three conditions,
2F 1G 3F 3G F(8, 44) = 0.5230, p = .8327.
Weld Type
Group 1 (TW) Group 2 (VR50) MWL
Figure 4. Mean training times by weld type (in order Overall MWL ratings were averaged from rat-
of increasing difficulty). ings of four welding types for the VR50 and TW
groups. A series of T tests showed no significant
differences between training conditions in terms
study. The experts’ physical development was of MWL. The mean MWL rating was descrip-
taken as the standard or optimal approach; tively lower for the VR50 group when compared
therefore, a result showing no significant differ- with the TW group; however, there was no sig-
ence between each of the training groups and nificant difference found between the two groups
the expert group would indicate desirable phys- when we considered the four welding types.
ical development in the trainees.
The results of the analysis for the 2F weld
Discussion
type revealed a significant difference between
three conditions (expert, VR50, and TW), F(8, The main issues addressed in this study were
34) =2.4057, p = .0356. However, further analy- (a) evaluation of the ways in which VR inte-
sis shows there was no significant difference grated training (VR50) affected the overall
between any two conditions: VR50 and TW, training performance compared to traditional
F(4, 10) = 2.9188, p = .0772; expert and VR50, training (TW) and (b) comparison of the cogni-
F(4, 13) = 2.3230, p = .1114; expert and TW, tive and physical developments in VR inte-
F(4, 8) = 2.1801, p = .1617. grated weld training with those in traditional
The outcomes for the 1G weld type also weld training. These issues now are discussed
revealed a significant difference between the by addressing the hypotheses posed at the
three conditions, F(8, 42) = 3.3556, p = .0046. beginning of this article.
Further analysis revealed that the expert and We first hypothesized that the use of VR inte-
TW groups, F(4, 13) = 1.2238, p = .3480, did grated welding training would result in student
not differ from one another, whereas the VR50 performance (number of certifications) that was
2F
4.5
3.5
Average Score
2.5
2 Group1 (TW)
1.5 Group2 (VR50)
0.5
0
Knowledge Score Comprehension Score Application Score Analysis Score
Cognitive Development Categories
1G
4.5
3.5
3 Group1 (TW)
Average Score
1.5
0.5
0
Knowledge Score Comprehension Score Application Score Analysis Score
Cognitive Development Categories
3F
3
2.5
2
Average Score
Group1 (TW)
Group2 (VR50)
1.5
0.5
0
Knowledge Score Comprehension Score Application Score Analysis Score
Cognitive Development Categories
3G
5
4.5
4 Group1 (TW)
3.5 Group2 (VR50)
Average Score
3
2.5
2
1.5
1
0.5
0
Knowledge Score Comprehension Score Application Score Analysis Score
Cognitive Development Categories
Figure 5. Cognitive scores by category across weld types for the 50% virtual reality–trained (VR50) and
traditionally trained (TW) groups.
2F
Knowledge 4 3.18 1.15 0.87 −1.84197 0.0811
Comprehension 0.6 1.18 0.52 0.6 2.362453 0.029*
Application 1.4 1.18 0.97 0.75 1.113528 0.2794
Analysis 2 2.636 1.33 0.81 1.337024 0.197
1G
Knowledge 3.64 3.45 1.03 0.82 −0.45883 0.6513
Comprehension 2.18 2.27 0.75 1.01 0.239732 0.813
Application 1.82 2.18 0.87 0.6 1.135924 0.2694
Analysis 0.27 0.55 0.47 0.69 1.088214 0.2894
3F
Knowledge 1.45 1.64 0.69 0.5 0.707107 0.4877
Comprehension 1.64 1.81 0.92 0.4 0.597614 0.5568
Application 2.64 2.1 0.5 0.7 −2.09529 0.0491*
Analysis 0.81 1.9 0.75 1.14 2.656845 0.0151*
3G
Knowledge 3.45 4.36 1.04 1.12 1.976424 0.0621
Comprehension 1.18 0.55 0.87 0.52 −2.07322 0.0513
Application 2 2.36 0.63 0.67 1.304656 0.2068
Analysis 2 2.36 0.45 0.81 1.304656 0.2068
Note. Values based on Crooks (1988) and Bloom, Englehart, Furst, Hill, and Krathwohl (1956). TW = traditional
welding training group; VR50 = virtual reality integrated training group.
equal to or better than that of students trained by indicating that students using VR integrated
traditional methods. Results of this study training have certifications equal to (2F, 3F, and
revealed that the overall performance for three of 3G welds) or better than (1G weld) those of stu-
the four weld types (2F, 3F, and 3G) was not dis- dents trained by traditional methods.
tinguishable between the training groups, given a The second hypothesis stated that the use of VR
confidence interval of 95%. Hence, for three integrated training would greatly reduce student
weld types, overall performance is similar for training times when compared with traditional
both VR50 and TW groups, with some descrip- training methods. The training times for partici-
tive benefits favoring VR training to traditional pants in this study showed significant reductions in
training. These findings are in agreement with favor of the VR50 group across all weld types.
those of several previous VR studies (e.g., Munz Participants in the VR50 group acquired task-
et al., 2004; Torkington et al., 2000). Of particu- critical skills faster than their TW group counter-
lar interest was the finding that the certifications parts, thus reducing the time investment needed to
for the 1G weld type shows that the VR50 group achieve competency levels necessary for testing.
outperformed the TW group to the point of prac- These time-based findings are highly consistent
tical significance. As mentioned earlier, the abso- with previous studies in VR training (e.g., Ganai,
lute number of 1G certifications earned was Donroe, St. Louis, Lewis, & Seymour, 2007; Jordan,
nearly significant, in addition to the fact that the Gallagher, McGuigan, McGlade, & McClure, 2000;
quality scores for this weld type were found to be Lapointe & Robert, 2000).
significantly higher for the VR50 group. The The reduction in training time requirements
results of this study support the original hypothesis, in VR50 students is believed to be attributable
2F 1G
25.0% 18.0%
16.0%
20.0% 14.0%
12.0%
% of MVC
% of MVC
15.0% 10.0%
10.0% 8.0%
6.0%
5.0% 4.0%
2.0%
0.0% 0.0%
Muscle Muscle
3F 3G
12.0% 12.0%
10.0% 10.0%
8.0%
% of MVC
% of MVC
8.0%
6.0% 6.0%
4.0% 4.0%
2.0% 2.0%
0.0% 0.0%
Muscle Muscle
Figure 6. Muscle activity interaction profile by weld type for expert, traditionally trained (TW), and 50%
virtual reality–trained (VR50) welders. MVC = maximum voluntary contraction.
in part to the fact that the VRTEX 360 system our hypothesis, VR integrated weld training
gives constant feedback to users as they train, greatly reduced student training times when
informing them which specific aspects of their compared with traditional weld training meth-
welds require attention. In traditional training, ods in this study.
students must take time to seek out instruction The third hypothesis theorized that the phys-
from the CWI; further time is expended as the ical development of welders trained with the
CWI performs visual inspection and analysis of use of VR integrated weld training would be
a weld and then provides the student with feed- similar to that of welders trained by traditional
back. Postexperiment interviews confirmed that methods; both groups of newly trained novice
participants in the VR50 group were much more welders were expected to differ similarly from
likely to seek feedback from the VR system expert welders. Given the relatively short train-
than from the instructor because they felt the ing period (2 weeks), the difficulty of the task,
information given by the system was “delivered and the precision required to perform a passing
in a more timely manner” than could be gain weld, we did not expect to find a statistically
from a shared instructor (the CWI). In addition, significant difference for any of the weld types
the VR environment allowed users to perform in terms of kinesthetic memory or physical
welds in less time by reducing the setup time development.
(time spent transporting and placing test plates In the case of the 2F weld type, the physical
in the work area), thus allowing the welder in development of the TW and VR50 participants
the VR environment to complete more welds in were not distinguishable from one another, nor
the allotted training time. In accordance with were they distinguishable from expert welders.
Downloaded from hfs.sagepub.com at University of York on June 5, 2016
Physical and Cognitive Effects of VR Integrated Training 569
This finding was understandable, given the rela- master and refine their welding technique, might
tively simple nature of the 2F weld. In the case of be discarded.
the 1G weld type, the physical development of The significant difference of muscle activity
participants in the VR50 group was significantly for 3F between the VR50 group and the other two
different from that of expert welders, whereas groups was also attributable to different postures
there was no significant difference between the adopted by welders. In the VR50 group, 4 of 11
TW group and the expert group. The VR50 par- participants used a physical prop (either a can or
ticipants used a modified approach to performing even the grinder) to support their elbow, which
the 1G weld that led to increased shoulder abduc- resulted in lower muscle activity level than in the
tion and consequently significant higher deltoid other two groups. However, most often in real
activity. This distinctive approach did not appear welding, welders will not be provided with any
to be used by the expert welders, nor was it taught such physical supports because of the complexity
by CWIs. The reason for this difference in of typical welding environments. It is difficult for
approach is likely multifactorial in nature. a VR system to mimic every condition or environ-
The VR machine encourages exploration and ment. However, the VR system provides trainees
experimentation in learning the welding proce- with general instruction on performing the task
dure. It allows the user to actively evaluate his or and is useful in forming a basis of understanding
her progress in real time and adjust his or her and skill that can then be applied and modified as
techniques accordingly. VR training also allows needed in the real-world environment.
the user to start over and abandon poor welds In the case of the 3G weld type (most diffi-
without concern about wasted consumable cult weld assessed in this study), the physical
materials. These attributes embolden users and development of participants in the TW and
encourage exploration into new welding proce- VR50 groups was not distinguishable from one
dures and techniques without fear of accruing another, nor was either group found to be distin-
excessive expenses. This freedom to experiment guishable from the expert welders. This out-
combined with real-time system feedback led come is better understood when the difficulty of
users of the VR system in this study to adopt a this weld is considered.
previously untaught approach to welding (spe- In the case of the 3F and 3G weld types (both
cifically, a minor increase in shoulder abduction vertical welds, the two more difficult welds
appeared to be advantageous to optimal travel assessed in this study), the physical develop-
angle and arc control). Postexperiment video ment of participants in the TW and VR50 groups
review and interview sessions regarding the use was not distinguishable from one another, nor
of the exaggerated should abduction revealed was either group found to be distinguishable
that nearly all the VR50 participants were using from the expert welders. This outcome is better
visual overlays to perfect the first weld pass in understood when the nature of vertical welds is
the 1G position. This level of detail encouraged examined. Vertical welds tend to leave less
them to adopt an increased angle (resulting in room for error and, as such, are less forgiving of
increased shoulder abduction) beyond that sug- experimental techniques and deviation of physi-
gested by the CWI. cal approach. As a result, novice welders in both
The specific approach taken by the VR50 par- training groups adopted a more limited physical
ticipants allowed for greater overall control in strategy; this strategy was highly similar to that
flat position welds, thus increasing the students’ used by expert welders.
chances of making a structurally sound weld (as When considering the 2F, 3F, and 3G weld
demonstrated with the higher 1G quality and cer- positions, no significant difference between the
tifications) by way of more consistent travel VR50 and TW groups was apparent in terms of
angle and arc length. However, considering the physical response. In addition, both groups of
higher muscle activity level than that in the novice welders differed from, or were equally
expert group, it can be conclude that the VR50 similar to, expert welders in a statically similar
participants compromised their physical effi- manner. These findings were in accordance with
ciency to achieve a sound weld, which, were they our third hypothesis; however, a discordant find-
to practice more real welding and gradually ing arose in evaluating the physical development
Downloaded from hfs.sagepub.com at University of York on June 5, 2016
570 October 2011 - Human Factors
outcomes of the 1G weld type. Results clearly not possible to conclude that a greater overall
indicate that the VR50 students adopted an alter- cognitive development existed for either group.
native and favorable strategy when compared In addition, there appears to be no correlation
with the traditional welders when performing the between cognitive development and certifica-
1G weld. Furthermore, within the 1G weld type, tion, indicating that students trained with the vir-
the VR50 group was significantly distinctive tual integrated system had effectively the same
from the expert welder group, whereas the TW level of cognitive development as did students
group was not. Hence the third hypothesis was trained by traditional methods. Results of this
not robust across all weld types. This finding indi- study revealed that the overall scores for cognitive
cates that VR integrated weld training systems development did not appear to be significantly
give trainees more freedom to develop their own better for either training group. This finding is in
approach of welding, which might not be practi- agreement with the original hypothesis.
cal in real welding. Providing trainees with proper The fifth and final hypothesis proposed that
movement restriction will be useful to help them MWL would not be significantly different
learn a standard posture more quickly and easily. between welders trained with VR integrated
The fourth hypothesis presented in this arti- versus traditional systems. The results of the
cle stated that the cognitive development of stu- MWL analysis in this study clearly showed no
dents trained with the use of VR integrated weld difference between participants in the VR50
training would be similar to that of students and TW groups.
trained by traditional methods. In most cases,
cognitive development was not found to be sig- Study Limitations and Implications
nificantly different between the VR50 and TW for Future Work
groups. However, it should be noted that two The primary limitation of this study lies in
cognitive development areas favored the VR50 the fact that the optimal ratio of VR to real-
group, and similarly, two cognitive develop- world (traditional) training is unknown. The
ment areas favored the TW group. findings of this study show that 50% VR inte-
Two of the differences in cognitive develop- gration shows some benefit; however, this
ment were related to the 3F weld type. The cog- result does not necessarily indicate that this
nitive evaluation of the 3F weld type indicated ratio is optimal. Another limitation of this study
that the VR50 group had significantly higher involves the length of training time allotted.
analysis development, but the TW group had Although 2 weeks is a realistic amount of time
significantly higher application development. for a condensed weld school attempting to
The knowledge development and comprehen- teach the 2F, 1G, 3F, and 3G weld types, an
sion development were not distinctly different investigation of welders in training across a
for either group. This finding indicates a differ- longer period could provide additional informa-
ence in the focus of learning between groups for tion to be used in defining how weld complex-
this weld type; however, one cannot conclude ity influences student attitudes and interactions
that this result constitutes a greater overall cog- with VR versus real-world welds.
nitive development for either group. In our future work, a 100% VR weld training
The remaining two differences in cognitive school will be examined and the progress of all
development were found in the 2F and 3G weld students tracked and tested with an AWS CWI for
types. Specifically, the VR50 group was found all the weld types discussed in this article. This
to have significantly higher comprehension proposed study will help further define the bound-
development for the 2F weld type, and the TW aries and potentials of modern VR technology.
group was found to have significantly higher
comprehension development for the 3G weld Conclusion
type. These differences in development did not The results of this study strongly support the
appear to cause a significant downstream differ- use of VR integrated training. VR integrated train-
entiation in analysis development or application ing was shown to lead to comparable cognitive
development for either group. Once again, it is development when compared with traditional
alternating imaging, normal laparoscopic imaging, and virtual Wormell, D., & Foxlin, E. (2003). Advancements in 3D interac-
reality training in laparoscopic psychomotor skill acquisition. tive devices for virtual environments. In ACM International
American Journal of Surgery, 180, 208–211. Conference Proceeding Series; Proceedings of the Eurograph-
Kadefors R., Petersen I., Herberts P. (1976). Muscular reaction to ics Workshop on Virtual Environments (Vol. 39, pp. 47–56).
welding work: an electromyographic investigation. Ergonomics New York, NY: Association for Computing Machinery.
19, 543 -558. Ye, N., Banerjee, P., Banerjee, A., & Dech, F. (1999). A com-
Kannan, T., & Murugan, N. (2006). Effect of flux cored arc weld- parative study of assembly planning in traditional and virtual
ing process parameters on duplex stainless steel clad quality. environments. IEEE Transactions on Systems Man and Cyber-
Journal of Materials Processing Technology, 176, 230–239. netics, Part C—Applications and Reviews, 29, 546–555.
Keir, P. J., & MacDonell, C. W. (2004). Muscle activity during Zanbaka, C. A., Lok, B. C., Babu, S. V., Ulinski, A. C., & Hodges,
patient transfers: A preliminary study on the influence of lift L. F. (2005). Comparison of path visualizations and cognitive
assists and experience. Ergonomics, 47, 296–306. measures relative to travel techniques in a virtual environment.
Lapointe, J. F., & Robert, J. M. (2000). Using VR for efficient IEEE Transactions on Visualization and Computer Graphics,
training of forestry machine operators. Education and Infor- 11, 694–705.
mation Technologies, 5, 237–250. Zipp, P. (1982). Recommendations for the standardization of
Mavrikios, F., Karabatsou, V., Fragos, D., & Chryssolouris, G. (2006). lead positions in surface electromyography. European Journal
A prototype virtual reality-based demonstration for immersive of Applied Physiology and Occupational Physiology, 50, 41–54.
and interactive simulation of welding processes. International
Journal of Computer Integrated Manufacturing, 19, 294–300.
Mellet-d’Huart, D. (2006). A model of (en)action to approach Richard T. Stone, PhD, AEP, is an assistant professor
embodiment: A cornerstone for the design of virtual environ-
at Iowa State University in the Departments of
ments for learning virtual reality. Journal of Virtual Reality,
10, 253–269. Industrial and Manufacturing Systems Engineering
Munz, Y., Kumar, B. D., Moorthy, K., Bann, S., & Darzi, A. (2004). and Mechanical Engineering. He received his PhD
Laparoscopic virtual reality and box trainers: is one superior to in industrial engineering from the State University of
the other. Surgical Endoscopy, 18, 485–494.
Park, M., Schmidt, L., Schlick, C., & Luczak, H. (2007). Design
New York at Buffalo in 2008. His current research
and evaluation of an augmented reality welding helmet. Human focuses on enhancing human performance with con-
Factors and Ergonomics in Manufacturing, 17, 317–330. sideration to cognitive, biomechanical, physical,
Perotto, A. (1996). Anatomic guide for the electromyographer: The environmental, and technological factors. Current
limbs and trunk. Springfield, IL: Charles C. Thomas.
Reyes-Valdés, F. A., Torres-Treviño, L., Lopez-Cortéz, V. H., de la
domain areas include augmented reality and telero-
Cruz-Márquez, F., & Briceño-Ramirez, F. (2006, March) Cor- botic systems and human power generation design.
relation between GMAW process and weld quality parameters:
A neural network approach applied in the automotive industry. Kristopher Watts, BS, is a 2nd-year PhD student
Paper presented at FABTECH and AWS 2006, Atlanta, GA.
Seymour, N. E., Gallagher, A. G., Roman, S. A., O’Brien, M. K., studying under Dr. Stone and Dr. Peters in the
Bansal, V. K., Andersen, D. K., & Satava, R. M. (2002). Virtual Department of Industrial and Manufacturing Systems
reality training improves operating room performance. Annals Engineering at Iowa State University. His research
of Surgery, 236, 458–464.
interest includes cognitive engineering, applied bio-
Sheik-Nainar, M. A., Kaber, D. B., & Chow, M. Y. (2005). Con-
trol gain adaptation in virtual reality mediated human-telerobot mechanical engineering, and manufacturing systems
interaction. Human Factors and Ergonomics in Manufacturing engineering.
& Service Industries, 15, 259–274.
Stefanidis, D., Korndorffer, J. R., Jr., Markley, S., Sierra, R., Heni-
Peihan Zhong, BS, is a 2nd-year PhD student studying
ford, B. T., & Scott, D. J. (2007). Closing the gap in opera-
tive performance between novices and experts: Does harder under Dr. Stone in the Department of Industrial and
mean better for laparoscopic simulator training? Journal of the Manufacturing Systems Engineering at Iowa State
American College of Surgeons, 205, 307–313. University. Her research interest includes human per-
Torkington, J., Smith, S. G., Rees, B. I., & Darzi, A. (2001). Skill
transfer from virtual reality to a real laparoscopic task. Surgical
formance engineering and human-robot interaction.
Endoscopy, 15, 1076–1079.
White, S. A., Prachyabrued, M., Chambers, T. L., Borst, C. W., Chen-Shuang Wei, MS, is a 3rd-year PhD student
& Reiners, D. (2010). Low-cost simulated MIG welding for studying under Dr. Stone in the Department of Industrial
advancement in technical training. Virtual Reality, 15(1), 1-13. and Manufacturing Systems Engineering at Iowa State
Wickens, C. D., & Baker, P. (1995). Cognitive issues in virtual
University. Her research interests include human
reality. In T. A. F. Woodrow Barfield III (Ed.), Virtual envi-
ronments and advanced interface design (pp. 514–541). New machine training systems and visualization.
York, NY: Oxford University Press.
Wilson, M. S., Middlebrook, A., Sutton, C., Stone, R., & McCloy,
R. F. (1997). MIST VR: A virtual reality trainer for laparo-
scopic surgery assesses performance. Annals of the Royal Col- Date received: September 20, 2010
lege of Surgeons of England, 79, 403–404. Date accepted: May 8, 2011