![Operations and interventions in the analyzed process [own work].](https://www.researchgate.net/profile/Agnieszka-Kujawinska/publication/279248156/figure/fig1/AS:294403336884226@1447202633701/Operations-and-interventions-in-the-analyzed-process-own-work_Q320.jpg)
Content uploaded by Agnieszka Kujawińska
Author content
All content in this area was uploaded by Agnieszka Kujawińska on Jul 23, 2015
Content may be subject to copyright.
Management and Production Engineering Review
Volume 6 • Number 2 • June 2015 • pp. 25–31
DOI: 10.1515/mper-2015-0013
HUMAN FACTORS IN VISUAL QUALITY CONTROL
Agnieszka Kujawińska, Katarzyna Vogt
Poznan University of Technology, Faculty of Mechanical Engineering and Management, Poland
Corresponding author:
Agnieszka Kujawińska
Poznań University of Technology
Faculty of Mechanical Engineering and Management
Piotrowo 3, 61-138 Poznań, Poland
phone: (+48) 61 665-27-38
e-mail: agnieszka.kujawinska@put.poznan.pl
Received: 30 April 2015 Abstract
Accepted: 8 May 2015 Visual control (visual inspection) is often used in production because – in comparison to
other kinds of control – it is relatively easy to conduct. It does not require any specialized
technical equipment. Human senses, usually sight, are the measurement tool. Unfortunately,
visual control does not guarantee a fully correct assessment. The reason is the limited human
reliability. There are plenty of factors which influence ability of a human to assess the process
or product quality properly. An important group of them are ergonomic factors. The goal
of the paper is to identify and discuss their influence on the efficiency of the visual quality
control in manufacturing processes.
The research was carried out in manufacturing company from automotive industry. The
paper presents the investigation of work organizational factors influence on visual control
effectiveness. A controller can make two types of errors in the process of visual inspection: to
assess a conforming product as “defective” or to assess a non-conforming product as “good”.
Effectiveness of sequential visual controls in s elected pro cess was examined. As a measure of
visual control effectiveness Control First Pass Y ield index was defended. Three operations
were analyzed: assembly of components, m elting components and applying a protective
coating.
Keywords
production, process efficiency, visual inspection.
Quality control
Controlling is one of the four major functions
of management, together with planning, organizing,
and leading. Among different types of c ontrolling in
production companies quality control is of particular
impo rtance, as it aims at checking process or prod-
uct compliance with the requirements of an internal
or external customer [1, 2].
Quality control in production processes consists
in the evaluation of one or more features of the prod-
uct, and comparing the result with the expectations.
Quality control may be divided into different types,
one of them being the division into the control of
measurable (quantitative) and immeasurable (qual-
itative) feature s. The other is frequently referred to
as alternative or attributable control [3–5].
Alternative product assessment is used when a
direct or indirect measurement of a given product
feature, expr e ssed with a numerical value, is either
impo ssible, difficult or cost ineffective. The outcome
of the alternative assessment does not provide infor-
mation on the extent to which the examined feature
complies with requirements; it is the bas is for a de-
cision w hether a given product may be considered
“good” or if it should be rejected and regarded as
poorly made, “bad” (defective). Therefore, the de-
cision whether a product meets or fails to meet the
sp e c ific requirements most often leads to the prod-
uct being classified to one of the two (rarely more)
conditions.
Alternative contro l may be perfor med using spe-
cialized equipment, which classifies products auto-
matically (e.g. pattern-recognizing machines which
verify PCB, devices evaluating the colour of print
etc.), or exclusively with the use of human s e ns es.
The other method is usually called organoleptic con-
trol. Visual inspection is a particular example of such
control.
25
Management and Production Engineering Review
Visual inspection
At the beginning of the 20th century visual in-
sp e c tion conducted by a man was considered as one
of the most reliable among the alternative quality
control methods (see Table 1). However, the view
was challenged in 1950s – 1970s, when it was under-
stood that the man is the weakest link in the quality
control process. Thus, one of the major trends in the
research on visual quality control at the time focused
on complete automatio n of the control process by re-
placing the man with a machine [6].
Table 1
Research tr ends in visual inspection [6–9].
Years Selected research trends
1950s Visual inspection analysis
1960s Development of techniques for the assessment
of visual inspection perf ormed by inspectors
1970s Development of signal detection theory and
mathematical models of visual inspection
1980s Continuation of research on the automation of
visual inspection
Development of inspector sel ection techniques
1990s Computer-aided visual inspection (instruc-
tions)
Investigation of visual inspection reliability in
aviation industry
2000 Continuation of research on the automation of
visual inspection
2010 The use of virtual reality in the training of
quality inspectors
This was conducive to the development of auto-
mated vision systems. It turned out however, that
successful implementation of such systems into in-
dustrial practice is not always possible (1980s and
1990s). One of the many reasons for that was and
still is the limited flexibility of automated vision sys-
tems, which becomes particularly evident in new,
non-standard situations, which were not taken into
account at the system design stage. Today at incom-
ing quality control, inter-operational in many pro-
duction proc e sses, visual inspection requires human
assessment. In the last decade research in this area
has focused mainly on the development of techniques
supporting human visual inspe ction (Table 1) [6–9].
Today it is r e c ognized that visual ins pection is
economically viable: it does not require the use
of expensive equipment, and it’s a no n-destructive
method, which means that it does not lead to the
wear and tear of the inspected pro duct. Still, using
visual insp e ction and keeping in mind its str e ngths,
one may not disregard its weakness: it is falli-
ble, and do e s not guarantee 100% correct assess-
ment [10].
The thesis is corroborated by numerous studies
conducted at different times by different researchers
(Table 2). Drury stated that inspection error is a
fact of life but it can be reduced with appropriate
interventions, such as employee training, better in-
structions, improvement of work conditions etc. [10].
For example, Swain and Guttman [6] estimated that
the visual inspection error rate in simple control
tasks appears to be on the order of 3-10%. Drury,
Karawanu and Vanderwarker suggest that the error
rate may in fact be much higher and range from 20%
to 30 % of incorrect classification [6].
Table 2
The effectiveness of vis ual inspection in actual processes –
examples [10].
Researcher Process
Effectiveness
of inspection
[%]
Jacobson
(1952)
Inspection of solderi ng
defects
45–100
Heida
(1989)
Aircraft landing gear
inspection
57–98
Drury
et al. (1997)
Aircraft visual inspec-
tion
68
Leach
and Morri s
(1998)
Visual inspection of
subsea structures and
pipelines
53
Graybeal
et al. (2002)
Routine inspection of
highway bridges
52
Visual inspection may be conducted by a man,
a machine or a hybrid: man-machine. Studies con-
ducted by Jiang demonstrated that a man-machine
combination is the best solution for such control in
terms of cost effectiveness. He also stated that hu-
man participation is crucial at the stage of decision
making, when the product is classified to the appro-
priate group, e.g. to “good” or “defective” products.
Visual inspection process
– structure and effectiveness
The structure of the visual inspec tion proc ess is
one of the mo st important features that influences
its effectiveness [10]. According to Fox [11], from the
work process perspective visual inspection consists
of several stages:
• visual “screening”/search for po tential defects,
• finding a defect (“detection”),
• defect classifica tion,
• decision that classifies a component, product or
service.
Each of the stages has an impact on the effec-
tiveness of inspection. The first stage, when an ob-
ject is visually examined by a man, requir e s vigi-
lance, heightened sensitivity of sight to detect poten-
26 Volume 6 • Number 2 • June 2015
Management and Production Engineering Review
tial err ors. In the first and second stage of inspection,
when the level of inspector’s perception is of particu-
lar significance, appro priate working conditions and
inspec tor’s knowledge about potential defects are ab-
solutely required.
In the third stage, based on his k nowledge about
the defects and classification criteria, the inspector
makes the decision on the type of defect detected in
the product.
In the final part of the inspection process the in-
sp e c tor dec ides if the product may be forwarded to
further steps of the process, or if it should be sepa-
rated from good quality products.
Two of the four s tages mentioned above (search-
ing for defects and decision-making) seem to be of
particular importance from the point of view of vi-
sual control. It turns out that they are most exposed
to decision variability of the ope rators. In the in-
sp e c tion process they may make two types of erro rs
(Table 3): classify a good quality product as defec-
tive (FALS) and classify a defective product as good
(MISS).
Table 3
Four possible human decisions in visual inspection of
products [4].
Actual condition of product
Decision
Defective product
(NOK)
Good quality
product
(OK)
Rejection
Correct decision:
Product rejection
Incorrect decision:
product rejection
(FALSE ALARM)
Approval
Incorrect decision:
product approval
(MISS)
Correct decision:
product approval
The likelihood of c ommitting these two types of
errors and the fraction of products that do not con-
form with requir e ments after the inspection process
are the key indicators of inspection e fficiency [12, 13].
In the literature ther e are many measures for the
effectiveness of visual ins pection [12], among oth-
ers the ratio of Freeman-McCornack, Youden-Nelson,
Wallack-Adams.
In the second part of this paper we use the CE
– Control Efficiency index, which is a percentage of
detected defects and an original index control First
Pass Yield (cFPY), which defines the share of defects
found during the first inspec tion in the total number
of defects at the given stage of the process, expressed
as a percentage.
cFPY =
a
∗
b
∗
, (1)
where a
∗
– number of defects found during the first
inspec tion∗100%, b
∗
– total number of defects at the
given stage of the process.
The maximum value of cFPY (10 0%) for an in-
sp e c tion means that every defect is detected at the
stage when it should be detected. That means that
no defect enters further stag e s of the process. Low
cFPY indicates that a given ins pection is ineffective.
The majority of defects which should be detected by
a given inspection a re only detected in further stages
of the process, by secondary inspections (occurring
later in the process sequence) o r are not detected at
all.
Factors affecting the efficiency
of visual inspection
There are many factor s that affect the efficiency
of visual inspection. Ma king the decision concerning
the quality of inspected products requires not only
a specific knowledge of the industry, but often also
individual approach to every inspected product and
high sensitivity to defects.
Relevant research shows that the efficiency of vi-
sual inspection is affected by independent factors and
factors related to and dependent on man (Fig. 1) [6,
7, 9, 14]. These two main groups of factors can b e
divided into five categories, as shown in Table 4 .
Fig. 1. Operations and interventions in the analyzed
process [own work].
Table 4
Factors affecting visual inspection efficiency [6, 7, 9, 14].
Factors Examples
Technical Type of defects; Defect visibility; Qual-
ity level; Standards (tests); Control au-
tomation; Other
Psychophysical Age; Sex; Observation skills; Experi-
ence; Temperament; Creativity; Other
Organizational Training; Scope of decision making;
Feedback; Precise instructions; Other
Workplace
environment
Light; Noise; Temperature; Work time;
Workstation organization; Other
Social Team communication; Pressure; Isola-
tion; Other
Technical factors are associated with the physi-
cal e xecution of visual inspectio n in the production
process. They include, for exa mple, factors rela ted to
the actual quality level, product fea tur e s subject to
inspec tion (their accessibility for visual inspection),
Volume 6 • Number 2 • June 2015 27
Management and Production Engineering Review
to the standards, based o n which the product is con-
trolled, the availability of tools used during the in-
sp e c tion, etc.
Psychophysical factors are ass ociated with men-
tal and physical conditions of inspectors. These in-
clude age , sex, intelligence, temperament, health con-
dition etc. Research in this area aims at identifying
the characteristics compr ising the profile of the ideal
inspec tor.
The nex t group of factors affecting the effective-
ness o f visual inspection are organizational factors.
These include support in decision-making during the
inspec tion, acquiring inspector skills, number and
type of inspections, information on efficiency and
accuracy of conducted inspections, as well as stress
factors influencing the inspector, such as time, c on-
sequences of incorrect assessment (no bonus, loss of
company image, etc.).
Wor kplace environment conditions are associa t-
ed with the workplace, wher e the inspection takes
place. These include physical factors, such as light,
noise, temperature, as well as the organizatio n of the
workstation itself.
The last group comprises factors related to the so-
cial environment, where ins pecto rs work. Their work
often involves pressure from people, whose interest is
contrary to the inspector’s work. For example, pr o-
duction staff (often colleagues) exert pressure expect-
ing approval o f their work (which is related to the
payment of salaries, bonuses). In turn, employees of
the management board may exert pressure to mini-
mize reinspections of products with unambiguous as-
sessment.
Case study
Research methodology
To determine the influence of selected abovemen-
tioned factors on inspection efficiency, a study was
carried out for thr e e operations in the PCB produc-
tion process.
The study fo c used on the visual inspection
processes during the assembly and so ldering of c om-
ponents, and the application of protective coating.
CE – Control Efficiency was chosen a s the effective-
ness index, expre ssed as a fraction of detected defects
and the cFPY measure. The study was conducted
over 32 weeks.
Three interventions took place in the examined
production process (Fig. 1):
• operation 1 : assembly and soldering of through-
hole components,
• operation 2: application of protective coating on
the PCB,
• operation 3: functional test.
For a given process, five control ope rations have
been distinguished (Fig. 1):
• inspectio n 1: performed by the operator during
the assembly and soldering of components – on-
line;
• inspectio n 2: performed in an separate location
by the inspector – interoperational off-line;
• inspectio n 3: performed by the operator during
the application of the protective coating – on-line;
• inspectio n 4 : performed in the functional test
operation – on-line;
• inspectio n 5: performed in a separate location
by the inspector – off-line.
The sources of possible defects in the process in-
clude ope ration 1, opera tion 2, and component sup-
plier (Table 5 ). Defects that originate in the assembly
and so ldering of components (operation 1) may in-
clude: lack of component, incorrect ass embly of the
component, component is not lead thro ugh and o ut-
side the asse mbly hole, excess of solder, lack of solder,
and other. We assumed that the defects should be
first detected at inspection 1 and inspe c tion 2. They
are also detected by inspection 4 and 5 (secondary
detection).
Table 5
The matrix of oper ations and visual inspections i n the
process and locations where the categorized defects originate
and are detected (Symbols: I – defects in component
assembly; II – defects in comp onent soldering pro cess; II I –
defects in protective coating; IV – impurities; V – defects
related to component quality. P – place where defect
originated; PW – location where the defect should have been
detected during the first inspection; WW – location of
detection during re-insp ection, when a defect was moved to
subsequent stages of the process).
Defects in the operation application of protective
coating may include co ating in prohibited a rea of the
circuit and lack of coating. It was assumed that the
28 Volume 6 • Number 2 • June 2015
Management and Production Engineering Review
defects should be detected during inspe c tio n 3, i.e. by
the operator applying the coa ting . Defects originat-
ing in this operation are also detected in inspection 4
and final (5) inspection.
Defects related to impurities and inappropriate
quality of components originate at the supplier. We
assumed that the defects should be first detected at
inspec tion 1 and inspection 2.
In the experiment we identified 4165 defects, in-
cluding 400 defects associated with assembly errors,
1053 defects related to soldering, 1207 defects result-
ing from the application of coating, 1395 defects as-
sociated with impurities and 110 defects of unaccept-
able component quality.
One hundred percent of PCBs were inspected.
The operator verified objects at 4x magnification,
and in uncer tain cases, at 10x magnification. An
identified defect was qualified and directed to de-
struction (unrepairable defective products) or to re-
pair (repairable defects).
Analysis of each inspection effectiveness was con-
ducted for the five categories of defects given in
Table 5 according to the orga nization and type
of inspection (o n-line and off-line), defect location
(known and unknown defect location) and work shift
(1, 2, a nd 3 shift). As a measure of effectiveness we
adopted the percentage of defects detected in the g iv -
en inspection (CE ), and the cFPY index.
The cFPY is deter mined only twice: for inspec-
tion 1 a nd 2, and inspection 3. (Fig. 2). For the first
two inspections we assumed that four categories of
defects should be detected: defects of assembly, sol-
dering, impurities, a nd quality of components. They
appear at this inspection point for the first time.
Fig. 2. Flow chart of quality control in the analyzed
process (Symbols: p% – percentage of defects detected
at a given stage (CE), cFPY – control First Pass Yield,
q% – percentage of undetected defects).
In the case of inspection 3 the cFPY index is c al-
culated only for the defects related to the protective
coating – their first occurrence.
For all control operations we also ca lculated the
percentage of undetected defects (the so-called con-
trol efficiency, CE) in relation to all defects which
occurred at the entry to the inspection process . It
should be noted that fo r inspection 1 and 2 the ba-
sis for determining the fraction of defects detected
was the number of defects related to four categories,
while for inspection 3,4 and 5 it was the total of de-
fects that were not detected in previous inspections
and defects concerning the c oating which appeared
in operatio n 3.
The organization and location of inspection
and its e ffectiveness
The organization and location of inspection is un-
derstood as its setting in the production process se-
quence. Two basic types of inspe c tion have been dis-
tinguished: on- line andoff-line. On-line inspection is
performed on an ongoing basis and it is one of the
stages of the flow of semi-finished/finished products
through the production line. Each time it ac com-
panies a technological operation (e.g. assembly and
soldering of components). Off-line inspection consti-
tutes a se parate stag e of a process, it does not ac-
company any technological oper ation. As shown in
Fig. 1, the process is divided into three on-line and
two off-line inspec tions.
Comparing the value of the cFPY index for in-
sp e c tion 1 and 2, and for inspection 3, we may say
they are relatively low (Fig. 3). For inspection 3, the
value of this index was o nly 11%. Such low level of
effectiveness definitely results from the relation be-
tween the on-line inspection and the performed op-
eration, which increases the risk of miss ing a defect.
In the case of inspection 1 and 2 the cFPY value
was higher: the operators found 28% and 40%, re-
sp e c tively, of all defects which occurred for the first
time. The value of the index was higher in inspection
2, because the inspection was conducted off-line.
Fig. 3. The value of cFPY index and CE – fractions of
detected defects for each inspection.
The percentage of detected defects expressed
with the CE index in the on-line inspection was
very low: 15% and 17%. Off-line inspections proved
much more effective, particularly the final inspection,
when all defects which “entered” the control opera-
tion were detected.
Volume 6 • Number 2 • June 2015 29
Management and Production Engineering Review
Defect location and its influence
on inspection efficiency
We distinguished two types of defect detection
complexity level in printed circuits. As regards the
4A. Defects of known location: assembly and soldering
defects
4B. Defects of unknown location: coating and impurity
defects
Fig. 4. Value of the cFPY index and fraction of errors for
defects of known and un known location in the process of
assembly, soldering, and coating (CE).
location of the defect in the product, there are de-
fects of known, rep eatable location and defects of
unknown, random location in the product. The first
group includes defects resulting from the assembly
and so ldering, and other – defects rela ted to impuri-
ties and quality o f protective coa ting (Fig. 4).
Comparing the efficiency of each inspection based
on the cFPY index value (Fig. 4) it should be noted
that defects with constant location k nown to inspec-
tors are detected most successfully. For comparison,
at inspection stations no. 1 and 2 operators detect
as much as 81% of all assembly defects, 80% of all
soldering defects, and only 31% of all defects r e lated
to impurities. In the c ase of defects of the protective
coating, fo r inspection 3 the index is very low at only
11%.
Defects that are difficult to predict, with random
location on the circuit, are a much greater challenge
for operators in the product evaluation process.
Work shift and its impact on inspection effec-
tivenes
The company works a thr e e -shift schedule. An-
alyzing the influence of the shift system on inspec-
tion effectiveness, we only focused on defects related
to soldering and coating. Res ults of the analysis are
shown in Fig. 5.
Fig. 5. Value of the FPY index for the soldering and
coating process for different work shifts.
30 Volume 6 • Number 2 • June 2015
Management and Production Engineering Review
Assessing the impact o f the work shift on the ef-
fectiveness of inspection at workstations, one may
note a certain trend: the effectiveness (CE) increases
with the progress of the production shift. For ex-
ample, for defects in component soldering, the CE
index in inspection 2 is 41% for the first shift, 49%
for the second shift, and as much as 62% for the
third (night) shift. Similarly, in inspection 4 the CE
index for the first shift was 16% and for the third
shift – 2 8%. Only in inspection 3 (related to the
detection of solder ing defects on-line), the value of
the CE index decreases. T he night shift is not con-
ducive to detecting such defects. The value of the
CE index for inspection 3 is 13% for the first s hift,
15% for second shift, and only 3% for the third
shift. It turns out that the night shift is not con-
ducive to the detection of defects of unknown loca-
tion.
Conclusions
The effectiveness of visual quality control per-
formed by a man is a complex issue, as it is influenced
by many fa c tors: both organizatio nal a nd those re-
lated directly to the man.
The ab ove was c onfirmed by the 32-week long
observation of visual inspections in the pr ocess of
PCB production. Such fac tors as location of the in-
sp e c tion, defect type, and work shift were taken into
account in the study.
When assessing the efficiency of inspection se-
quence with the C E and cFPY indexes fo r differe nt
types of defects it was noted that already at inspec-
tion 1 defects with k nown, repeatable loc ation in the
product are detected much more successfully. Defects
with random location are us ually detected as a result
of a final control and r e quire a comprehensive prod-
uct assessment.
Additionally, we should no te that the off-line con-
trol (e.g. no. 2 or 5) is more efficient than on- line, par-
ticularly for defects with a co nstant, repeatable lo-
cation. Combining quality control interventio ns with
the technological work disrupts the work o f the op-
erator. The risk of error in on-line control is greater
than in the off-line control.
Moreover, differences in inspection efficiency dur-
ing different work shifts result not only from the
organization of work itself, but also from the psy-
chophysical abilities of the man. This influence be-
comes apparent in the differences between the effi-
ciency of control ope rations performed during the
subsequent work shifts.
References
[1] Hamrol A., Quality Management with examples,
PWN, Warsaw, 2008.
[2] Price P.S., Right First Time, Gower, London, 1985.
[3] Bauer J.E., Duffy G.Z., Westcott R.T., T he Quality
Improvement Handbook, Quality Press, Milwaukee,
2002.
[4] Oakland J.S., Statistical Process Control: A Prac-
tical Guide (5th Edition), Butterworth-Heinemann,
Oxford, 2003.
[5] Wheeler D.J., Understanding Statistical Process
Control, SPC Press, Knoxwille, 1992.
[6] Drury C.G., Sinclair M.A., Human and machine
performance in an inspection task. Human Factors,
25, 391–399, 1983.
[7] Hamrol A., Kowalik D., Kujawińska A., Impact of
chosen work condition factors on quality of manual
assembly process, Human Factors and Ergonomics
in Manufacturing, 21, 2, March/April 2011, item:
3331, pp. 156–163, 2011.
[8] Jo Y-D ., Park K-S., Dynamic management of hu-
man error to reduce total risk, Journal of Loss Pre-
vention, 16, 313–321, 2003.
[9] Kalaykov I, Wide P., Intelligent Feature Estimation,
Based on Human‘s Perceptual Features, SCIMA,
International Workshop on Soft Computing Tech-
niques, Measurements and Related Applications,
2003.
[10] Drury C.G., Karwan M.H., Vanderwarker D.R., The
two-inspector problem, IIE Transactions, 18, 174–
181, 1986.
[11] Fox J.G., Ergonomics contribution to subjective in-
spection in industrial quality control, The Produc-
tion Engineer, 1975.
[12] Raz T., Littchwager J.M., Ten measures of Inspec-
tion Performance, Jurnal of Quality Technology, 21,
2, 1989.
[13] Więcek-Janka E., Decisions and games, Publishing
House of Poznan University of Techn ology, 2011.
[14] Więcek-Janka E., Psycho-profile of Polish Microen-
trepreneur [in Polish: Sylwetka psychofizyczna pol-
skiego M ikroprzedsiębiorcy], Ekonomika i Organiza-
cja Przedsiębiorstwa, p. 116–120, 2007.
Volume 6 • Number 2 • June 2015 31
