Automation in Construction 85 (2018) 209–219

Contents lists available at ScienceDirect

Automation in Construction
journal homepage: www.elsevier.com/locate/autcon

Origins and probabilities of MEP and structural design clashes within a T

federated BIM model
E.A. Pärna,⁎, D.J. Edwardsa, Michael C.P. Singb
a
Faculty of Computing, Engineering and the Built Environment, City Centre Campus, Millennium Point, Birmingham B4 7XG, United Kingdom
b
Department of Building and Real Estate, 7/F, Block Z, The Hong Kong Polytechnic University, Hong Kong

A R T I C L E I N F O A B S T R A C T

Keywords: ‘Design clashes’ encountered during the development of a large multi-storey educational building, awarded
Building information modelling under a Joint Contracts Tribunal (JCT) Design and Build contract, are reported upon. The building was devel-
Clash detection oped in Birmingham, UK and the contract value was circa £36 million (UK Sterling, 2015). Members of the
Probability density function project management team (PMT) produced designs that were subsequently integrated by the main contractor
Cumulative distribution function
into a federated building information modelling (BIM) model; at this stage 404 error clashes were evident
Generalized Gamma distribution
between the positions of the mechanical, electrical and plumbing (MEP) designer's and structural designer's
Log Logistic (3P) distribution
building compartments. The contractor deemed that these particular clashes were ‘mission critical’ as previous
experience suggested that project costs could spiral uncontrollably if left unabated. Participatory action research
was employed to acquire a deeper understanding and knowledge of the clash incidents. Clash data accrued (in
mm) was subsequently quantitatively modelled using the probability density function (PDF) and the cumulative
distribution function (CDF). Two models produced were the Log Logistic Three Parameter (3P) (using all data
including outliers) and Generalized Gamma distribution (excluding outliers). Both models satisfied Anderson-
Darling and Kolmogorov-Smirnov goodness of fit tests at α 0.01 and 0.02 levels of significance. Model para-
meters could be used to forecast similar clashes occurring on future projects and will prove invaluable to PMT
members when accurately estimating the time and resource needed to integrate BIM designs. The predictive
modelling revealed that 92.98% of clashes reside within the 30–299 mm range while the most probable oc-
currence of a clash overlap resides in a discrete category of 100–199 mm. Further qualitative investigation is also
conducted to understand why these clashes occurred and propagate ideas about how such may be mitigated. The
research concludes on two important points, namely: i) BIM is not a panacea to design related construction
project rework and that innovative 21st century digital technologies are hampered by 20th century management
practices; and ii) improvements in clash and error mitigation reside in a better understanding of tolerances
specified to alleviate the erroneous task of resolving unnecessary clashes. Future research is proposed that seeks
to: automate the clash detection management, analysis and resolution process; conduct further investigative
analysis of the organizational and human resource management influences impacting upon design clash pro-
pagation; and devise and validate new procedural methods to mitigate clash occurrence using a real-life project.

1. Introduction digital devices employed within the workplace generates vast quantities
of data, information and knowledge that can be further exploited via
The digital jacquerie transcends the narrow confines of the in- automated and intelligent analytics [4]. Business intelligence and
formation and communication technology sector and is ubiquitous concomitant data analysis have the inherent potential to uncover pat-
throughout all industry [1]. This paradigm shift in business and com- terns, trends and associations related to design data, human behavior,
merce has been enabled through the application of cloud computing and the interactions between the two, for improved decision making
[2]. Cloud computing is advantageous to all organizations (large and [5,6]. Indeed, the extant literature postulates (cf. [7,8]) that business
small) because utilizing internet-based services can reduce start-up intelligence enables organizations to gain value from business analytics.
costs, lower capital expenditures and increase computational power to Multitudinous benefits of digitization have similarly been promulgated
augment business/market intelligence [3]. A menagerie of ‘networked’ within the architecture, engineering, construction and owner-operated

⁎
Corresponding author.
E-mail address: [email protected] (E.A. Pärn).

http://dx.doi.org/10.1016/j.autcon.2017.09.010
Received 12 December 2016; Received in revised form 19 March 2017; Accepted 13 September 2017
Available online 05 November 2017
0926-5805/ © 2017 Elsevier B.V. All rights reserved.
E.A. Pärn et al. Automation in Construction 85 (2018) 209–219

(AECO) sector [9]. Prominent digital technologies include: sensors individual, organizational, and interorganisational error prevention
[10]; laser scanners [11]; machine vision [12]; and building informa- practices.”
tion modelling (BIM) [13]. Amalgamated, these technologies have
Successful error mitigation should therefore nurture learning from
spearheaded the advancement of the digital construction modus oper-
within individual design disciplines to encapsulate the entire project
andi [71]. BIM is ostensibly the most prevalent of these advanced
team (ibid). BIM inherently offers this potential but as the first stage of
technologies within extant literature and is gradually becoming con-
design error mitigation, clash detection and consequential resolution
ventional in both design and construction practice globally [14]. BIM
between design team members has received scant academic attention.
provides a digital portal through which an integrated project manage-
Amongst the various structural elements, MEP design errors have tra-
ment team (PMT) can collaboratively work upon, and share knowledge
ditionally dogged the design process, arguably due to the confined
of, a construction or infrastructure development pre-, during and post-
spaces left for MEP systems [74]. Recent research conducted by Pean-
construction [15,16]. This innovative approach enables PMT members
supap and Ly [33] examined five categories of structural and MEP re-
to enhance their inter-disciplinary interactions in order to optimize
lated design errors, but the study was confined to schedule delays and
resultant decisions and afford greater whole life value for the asset [17].
omitted any discussion on how BIM can facilitate error mitigation at the
During the design stages of pre-construction, BIM drawings and
detailed design stages. Research that has examined design clashes in a
plans produced by individual designers (e.g. the architect, structural
BIM environment remains anecdotal or based upon a limited scope of
engineer and mechanical, electrical and plumbing (MEP) designer) are
analysis [19,34,35].
integrated into a federated model and tested to identify design clashes
[18]. Design clashes consist of ‘positioning errors’ where building
components overlap each other when the original individual designer 2.1. Clash reports and nomenclature
models are merged. Resolving these design clashes is imperative to
project performance, particularly if costly rework is to be circumvented When reporting upon design clashes, the main contractor produces
during the construction phase. However, design clash mitigation and periodic clash detection reports that contain information including: i)
the utilization of deterministic modelling to enhance decision making thematic groupings of clashes that report upon individual clashes
are two areas that have been grossly overlooked within the literature within each compartment category (for example, and in this research
([19,72]). Given scant research within this important area and the ‘MEP vs building column’ and ‘MEP vs building frame’); ii) snapshots of
opportunity to improve construction business performance, this work every clash identified to aid communication with all designers
reports upon the findings of participatory action research (PAR) which throughout the PMT; iii) clash point co-ordinates (as x, y and z co-
sought to examine design error clashes that occurred during the com- ordinates) to determine the exact pin-point location of the clash within
pilation of a federated BIM model for a multi-storey educational the federated BIM model; iv) the date that the clash was found; v) clash
building development. Such work provides invaluable insight into a status (active and unresolved or resolved); vi) a written description of
previously unexplored area of digital built environment research. The the clash; and vii) a numerical value in metres (m) or millimetres (mm)
research objectives are to: better understand why clashes occur and that specifies the linear magnitude of the positional (clash) error.
engender wider academic debate; demonstrate how the probability Manual data cleansing is then undertaken by the contractor's BIM
density function (PDF) and cumulative distribution function (CDF) can manager using industry nomenclature to define four key clash cate-
accurately predict the probability of future occurrence for a specific gories, namely: i) clash errors –fault clashes that must be identified and
project; formulate innovative ideas for reducing their occurrence and resolved within the federated model; ii) pseudo clashes – permissible
mitigating their impact upon construction business processes and per- fault clashes that can be tolerated within the design and do not require
formance; and suggest future work that seeks to maximize business resolution; iii) deliberate clashes – intentional clashes, for example,
intelligence through automation and apply the deterministic techniques ducting through a floor or web of a structural steel component; iv)
adopted to a larger number of project developments as a means of duplicate clashes – multiple versions of the same ‘singular clash’ that are
generalizing the findings. repeated throughout a building (e.g. an MEP pipe that travels along the
entire length of a structural column will be observed and recorded
2. Design errors within digital construction numerous times even though it actually represents one error). Duplicate
clashes often originate from one of the three other variants of clash.
Design errors are a prominent root cause of diminished construction
project performance and manifest themselves as adverse symptoms 3. Research approach
such as: rework [20–22]; cost overruns [23–25]; schedule delays (ibid);
and unsafe working environments [73]. Literature proffers that the The research design employed participatory action research (PAR)
main sources of design error are inextricably linked to iterative and (cf. [36,37]) where the lead researcher was embedded within, and
recurrent design cycles that result from: unanticipated changes [26]; worked closely with, the PMT to develop various aspects of the BIM
poor management and communication [27]; realignment of traditional/ model. The PMT included the client's representatives (i.e. the building's
institutionalized organizational and human resource practices [28]; and estates department) and design related disciplines (including the BIM
interoperability between various software platforms [29]. These chal- process manager, the lead architect, contractor's construction manager,
lenges have engendered frenzied research activity and resulted in the: the contractor's BIM manager, principle designer for mechanical en-
development of system dynamics models for planning and control [26]; gineering and plumbing and the lead structural engineer). Note that the
identification of critical design management factors [30]; and ex- estate's department held four fundamental roles, namely that of: client's
amination of causal factors [31]. Despite this herculean effort, anec- representative; BIM process manager; project manager; and estates
dotal evidence from industry reveals that design errors remain a per- department and consequently, covered all three major phases of the
sistent problem. building's life cycle. PAR was adopted because it offers pluralistic or-
BIM offers a potential digital solution space for design error man- ientation to knowledge creation and change thus affording greater
agement as a collaborative and inclusive platform [32]. Yet to date, flexibility to excoriate beneath the corporate façade that can obscure
limited research has investigated whether BIM in the AECO sector is truth in the interests of preserving reputation and consequential prof-
effectively mitigating digital design errors. Love et al. [73] further itability. This approach to self-experimentation grounded in experience
proffer that the process of design error mitigation implies that: was augmented by: fact-finding, to acquire a deeper knowledge and
understanding [38,39]; learning, through a recurrent process of re-
“…learning from errors is a collective capacity that can produce
flection [40]; and evidential reasoning to interpret information and

210
E.A. Pärn et al. Automation in Construction 85 (2018) 209–219

knowledge characterized by varying degrees of uncertainty, ignorance Fig. 2). Spatial coordination between the various design discipline
and correctness [41]. Participatory action research is particularly ben- models was carried out at regular fortnightly intervals (every ten
eficial because research implementation which embodies collective working days) throughout the design and construction stages. The BIM
enquiry and experimentation [42], occurs within the PMT rather than manager was integral within this process and facilitated regular co-
‘for it’. Consequently, PMT stakeholders are more likely to adopt ordination of team meetings, model updates, clash revisions and con-
emergent findings, recommendations and modify their future practices. trol. Clash detection in BIM is a global phenomenon; unlike other
Data collated was analyzed using a mixed methods approach that countries worldwide, it has been forcefully mandated in the United
combined predominantly quantitative probability modelling of clash Kingdom (UK) [44,45]. According to the UK Government mandated
data with qualitative investigation and delineation of the model fed- BIM Level 2 requirements, design teams must undertake weekly or
eration and clash management process. Once this aforementioned fortnightly task information and clash rendition tasks to ensure designs
process was succinctly documented in illustrative format, unstructured are fully coordinated and clash free, ensuring that requests for in-
interviews were then conducted with members of the PMT to identify formation are minimised during construction stages [44,45]. This
challenges that exacerbate the problem of clash propagation. The government intervention seeks to mitigate design error prominence
contractor was particularly insistent that error clashes between the within BIM implementation. The client also required that the main
positions of the MEP designer's and structural designer's building contractor employed a clash detection management process on a fort-
compartments were analyzed in greater detail. Such clashes were nightly basis. Clash detection resolution was implemented via Virtual
deemed to be ‘mission critical’ as previous anecdotal experience (ac- Design and Construction (VDC) coordination meetings with the re-
crued from past projects completed) suggested that project costs could spective design teams. The BIM execution plan (as outlined by the main
spiral uncontrollably if these were left unabated. contractor during tender) stated that:
The construction of a large multi-storey educational building lo-
“The aspiration is that beyond Stage 4, the model will be managed
cated in Birmingham UK (entitled the ‘Mary Seacole Building’ – refer to
by the principal contractor and modifications to the model be made
Fig. 1 for external visualization) provided the contextual setting for the
in house or by the design team. Throughout the project the BIM lead
research. The contract value exceeded £36 million UK Sterling and
from each company and the soft landings champion will attend
created 10,000 sq. m of new teaching space. The project commenced
regular VDC coordination meetings. Efforts will be made to co-
with a client sign off on March 2015 and is currently ongoing with an
ordinate the VDC meetings with design team meetings. During
expected handover to client by September 2017. A Joint Contracts
construction it will be led by the main contractor.”
Tribunal (JCT) Design and Build was employed and procurement was
implemented via the Official Journal of the European Union (OJEU) The main contractor and its team members adopted cloud based
tender submissions. OJEU is used for all tenders from the public sector platforms to alleviate the number of discrepancies between the ‘as-
which are valued above a certain financial threshold according to constructed’ and the ‘as-built’ BIM model. Specifically, Autodesk's®
European Union legislation [43]. BIM360 platform for design coordination and as-constructed validation
was chosen as the cloud-based BIM tool for this task. Clash detection
4. Analysis was also conducted via this cloud based platform enabling stakeholders
to link discipline specific design models (obtained from the MEP de-
A federated BIM model was used to identify clash detections. signer, structural engineer and architect) into the main contractor's
Federated models are deployed using various BIM-related platforms federated model (i.e. Autodesk® Glue). Although open architecture was
including: Bentley Navigator®, Autodesk Navisworks® and Autodesk used within the federated model to reduce errors, 404 design clashes
Glue®. For this research, Autodesk Glue® was used to facilitate cloud were identified between the MEP designer's model and the structural
based model federation. The project employer information require- designer's model (refer to Fig. 3a and b).
ments stipulate that for the contractor:
4.1. Data mining
“Glue Coordination models will be created at different stages. They
will be used for a number of reasons, some of these are, clash de-
Within this data sub-set of design clashes, 150 observations related
tection, MDM creation, 4D and 5D modeling, and used as the base
to MEP vs building column clashes and 254 related to MEP vs building
model for the ‘BIM 360 Field’ database – these are but some of the
frame clashes. Summary statistical data analysis in Table 1a presents
uses.”
parametric and non-parametric descriptive measures of central ten-
The main contractor employed a permanent BIM manager to dency and measures of variation or dispersion within the sample data
manage clash detection of the federated model in Navisworks® (refer to [46]. Evidence of skewness was apparent given the distance between
the arithmetic mean and median values (namely 212.82 mm and
166.78 mm respectively). Skewness measures the asymmetry of the
probability distribution of a real-valued random variable about its mean
[47]. It was observed that the clash detection data was positively
skewed; the majority of data fell within the 41.09 mm to circa 250 mm
measurement range but a long tail extending to 550.03 mm was re-
corded. Because the presence of outliers was suspected an established
outlier detection test was used to confirm this and subsequently remove
them prior to conducting the analysis for a second time. The outlier test
used was:
Outlier = ((Q3 − Q1) × 1.5) + Q3) (1)
where: Q1 = is the first quartile value; Q3 is the third quartile value;
and 1.5 is a constant.
The outlier limit value was noted as 440.74 mm but further data
analysis revealed that two observations extended beyond this and were
predominantly responsible for the long tail observed. These two values
Fig. 1. Proposed extension of Mary Seacole Building (Sheppard Robson architects).
were duplicate clashes (457.534 mm (frequency = 24) and 550.031 mm

211
E.A. Pärn et al. Automation in Construction 85 (2018) 209–219

Fig. 2. Client requirement processes adopted for fortnightly clash detections.

Table 1a
Summary statistical analysis of error clashes (structural vs MEP - all data).

Statistic Value Percentile Value

Sample size 404 Min 41.09

Range 508.94 5% 54.95
Mean 212.82 10% 74.528
Variance 19,197 25% (Q1) 122.89
Std. deviation 138.55 50% (median) 166.78
Coef. of variation 0.65102 75% (Q3) 250.03
Std. error 6.8933 90% 457.53
Skewness 1.1496 95% 550.03
Excess kurtosis 0.30751 Max 550.03

Table 1b
Summary statistical analysis of error clashes (structural vs MEP - outliers excluded).

Statistic Value Percentile Value

Sample size 344 Min 41.09

Range 329.06 5% 53.811
Mean 163.69 10% 66.37
Variance 5892.2 25% (Q1) 116.77
Std. deviation 76.761 50% (median) 148.64
Coef. of variation 0.46895 75% (Q3) 222.65
Std. error 4.1387 90% 250.03
Skewness 0.75898 95% 350.11
Excess kurtosis 0.35379 Max 370.15

statistical analysis is therefore presented in Table 1b that excludes

outliers and illustrates that the arithmetic mean and median are much
closer together (153.69 mm and 148.64 mm) and that skewness has
been reduced (although not eliminated).
The two pools of design clash data (with and without outliers) were
Fig. 3. a - Structural vs. MEP clashes in Autodesk Navisworks (MEP service in column).
then modelled using empirical PDF and CDF for a continuous dis-
b - Structural vs. MEP clashes in Autodesk Navisworks (MEP service in beam). tribution; these models were used to improve knowledge of clashes that
propagate during design works. A comparative analysis between the
goodness of fit tests generated for both types of probability modelling
(frequency = 36)) and accounted for 60 outliers in total. The treatment
was undertaken to measure any observable differences.
of outliers is a contentious issue within extant literature and could
broadly involve either removing or transforming them using for ex-
4.2. Probability modelling
ample, square root, log10 or box-cox transformations [48]. It can be
argued that removing outliers squanders important data (and hence
The PDF for a continuous distribution can be expressed in terms of
knowledge) in the subsequent analysis but keeping them produces an
an integral between two points:
uncharacteristic pattern in the trend. Given the contentious nature of
outlier treatment, subsequent analysis examined both data sets – un- b

transformed original data with and without outliers. A revised summary P ∫ f (x ) dx = P (α ≤ X ≤ b)

α (2)

212
E.A. Pärn et al. Automation in Construction 85 (2018) 209–219

A CDF is the probability that a variate takes on a value less than or Table 2a illustrate that the best fit probability distribution for the size of
equal to x. For continuous distributions, the CDF is expressed as a curve clash detections was the Log Logistic Three Parameter (3P) at α = 0.01
and denoted by: and 0.02 confidence intervals; notably, the fit was not achieved at
x α = 0.05. The three parameters are:
F (x ) = ∫ f (t ) dt α = 2.2943; β = 147.33; and γ = 23.249
−∞ (3)
The PDF (Fig. 4) and CDF (Fig. 5) for the Log Logistic 3P distribu-
The empirical CDF is displayed as a stepped discontinuous line de-
tion fitting are defined in Eqs. (7) and (8) respectively as:
pending upon the number of bins and is denoted by:
α−1 α −2
1 α ⎛x − γ⎞ ⎛ ⎛ x − γ ⎟⎞ ⎞
Fn (x ) = . [Number of observations ≤ x ] f (x ) = ⎜ ⎟ ⎜⎜ ⎟
n (4) β⎝ β ⎠ ⎝⎝ β ⎠ ⎠ (7)
where bins are the number of equal vertical bars contained within a α −1
β ⎞ ⎞
CDF histogram, each representing the number of sample data values F (x ) = ⎜⎛1 + ⎜⎛ ⎟ ⎟

(that are contained within each corresponding interval), divided by the ⎝ ⎝x − γ⎠ ⎠ (8)
total number of data points. where: α is a continuous shape parameter with α > 0 ; β is a continuous
The PDF, CDF and distribution parameters (e.g. α, β, γ, μ, k, m, σ, ξ) scale parameter with β > 0; and γ is a continuous location parameter
for 36 different continuous distributions, including Beta, Exponential, where γ ≡ 0 yields the two parameter-Log Logistic distribution. The
Frechet, Gumbel Max/Min and Wakeby, were examined using the esti- domain for this distribution is γ < x < + ∞.
mation method Maximum Likelihood Estimates. The best fit distribu-
tion was then determined using two goodness of fit tests, namely the: 4.4. Distribution fitting: probability of the size of clash – model two (outliers
Anderson-Darling statistic (A2); and Kolmogorov-Smirnov statistic (D). excluded)
Combined, these goodness of fit tests measure the compatibility of a
random sample with a theoretical probability distribution function – or For the second model, 344 observations were analyzed (excluding
put simply, how well the distribution fits the data. duplicate clash outliers). Results reported in Table 2b illustrate that the
The Anderson-Darling statistic (A2) is a general test to compare the fit best fit probability distribution fitting for the size of clash detections
of an observed CDF to an expected CDF. The test provides more weight was the three parameter Generalized Gamma at α = 0.01, 0.02 and
to a distribution's tails than the Kolmogorov-Smirnov test. The Anderson- 0.05 confidence intervals – this represented a minor improvement upon
Darling statistic is defined as: model one. The three parameters are:
n
1 k = 0.99505; α = 4.5101; and β = 35.997
A2 = −n −
n
∑ (2i − 1)⋅[InF (xi) + In (1 − F (xn−i +1))]
i=1 (5) The PDF (Fig. 6) and CDF (Fig. 7) for the three parameter Gen-
The Kolmogorov-Smirnov statistic (D) is based on the largest vertical eralized Gamma distribution fitting are defined in Eqs. (9) and (10)
difference between the theoretical and empirical CDF. It is defined as: respectively as:
k
i−1 i kx kα − 1 ⎛ x ⎞
D= max
1 < i < n(F (x i ) − , − f (x i ) ⎞⎟ f (x ) = exp ⎜−⎛⎜ ⎞⎟ ⎟
n n β kαΓ (α ) β
⎠ (6) ⎝ ⎝ ⎠ ⎠ (9)

These goodness of fit tests were used to test the null (Ho) and al- Γ (x β )k (α )
ternative hypotheses (H1) of the datasets: H0 - follow the specified F (x ) =
Γ (α ) (10)
distribution; and H1 - do not follow the specified distribution. The hy-
pothesis regarding the distributional form is rejected at the chosen where: k is a continuous shape parameter k > 0 ; α is a continuous
significance level (α) if the statistic D and, A2 are greater than the shape parameter α > 0; β is a continuous scale parameter β > 0; and γ
critical value. For the purposes of this research, 0.01, 0.02 and 0.05 is a continuous location parameter (γ ≡ 0 yields the three-parameter
significance levels were used to evaluate the null hypothesis. Generalized Gamma distribution).
The p-value, in contrast to fixed α values, is calculated based on the Both distribution fitting models illustrate a good fit at the 0.01 and
test statistic and denotes the threshold value of significance level, in the 0.02 confidence intervals and therefore the removal of outliers was not
sense that Ho will be accepted for all values of α less than the p-value. a prerequisite requirement to obtaining a valid result. Using the para-
Once the ‘best fit’ distribution was identified, the probabilities for a meters contained within model two, delimiters (X1 and X2) were used
design clashes were calculated using the CDF. to calculate the probabilities of obtaining a discrete category of clash
ranging from 30 to 99 mm, 100–199 mm, 200–299 mm, 300–399 mm
4.3. Distribution fitting: probability of the size of clash – model one (all and 400–470 mm (refer to Table 3). These tolerance categories were
data) defined and delineated by the contractor for the purposes of clash de-
tection. The research team felt that such: i) was an arbitrary decision
All 404 data points were analyzed for model one. Results reported in inordinately influenced by a hired BIM consultant; and ii) lacked logic
and a meaningful basis for this decision. From this discrete analysis, it
Table 2a
was apparent that 92.98% of clashes reside within the 30–299 mm
Goodness of fit (all data) - Log Logistic (3P). range; where this range consists of the 30–99 mm = 19.85%;
100–199 mm = 51.05%; and 200–299 mm = 22.08% discrete cate-
Kolmogorov-Smirnov Sample size 404 gories.
Statistic 0.07126
p-Value 0.03144
α 0.05 0.02 0.01 5. Clash management challenges and considerations
Critical value 0.06756 0.07552 0.08105
Anderson-Darling Sample Size 404 The quantitative analysis conducted within this research illustrates
Statistic 2.7754
that PDF and CDF can successfully model the probability of design
α 0.05 0.02 0.01
Critical value 2.5018 3.2892 3.9074 clashes that occur during the development of a federated BIM model.
Such modelling will prove useful to the client and members of the

213
E.A. Pärn et al. Automation in Construction 85 (2018) 209–219

Fig. 4. Probability density function – Log Logistic (3P) all

data.

Fig. 5. Cumulative distribution function – Log Logistic (3P)

all data.

Table 2b structural engineers and other design consultants populated BIM se-
Goodness of fit (outliers excluded) – Generalized Gamma. mantic data within a discipline specific BIM model in an iterative manner.
These discipline specific models were then integrated into an initial
Kolmogorov-Smirnov Sample size 344
Statistic 0.05869
federated model. Tier two involved the implementation of the con-
p-Value 0.1797 tractor's cloud computing solution that provided a two-way communica-
α 0.05 0.02 0.01 tion portal between the designers and contractor. Within the cloud,
Critical value 0.07322 0.07322 0.07322 Autodesk Glue® was used to federate the model; BIM 360 Field was
Anderson-Darling Sample size 344
used to store and upload site photographs and facilitate communication
Statistic 1.8396
α 0.05 0.02 0.01 between individual PMT members; and BIM 360 Layout was used as a
Critical value 2.5018 2.5018 2.5018 tool to input Cartesian coordinates (of the building and site) using a
total station. In tier three, the contractor, contractor's BIM Manager and
designers implemented a recurrent process of clash detection and re-
design team who seek to better understand and mitigate future clash solution. The designers identified model clashes as a first step towards
occurrence. However, the origins of clashes cannot be explained by developing resolved model clashes that were uploaded into an initial
quantitative analysis alone, hence further qualitative investigation of clash report. The contractor's BIM Manager then used this clash report to
the model federation and clash management process was conducted iteratively work with designers to resolve clashes within a final feder-
(refer to Fig. 8). A three tier process was implemented that consisted of: ated model that was uploaded into the cloud for all members of the PMT
tier one – the design stage; tier two – cloud computing; and tier three – to access. This clash management process was further explored using
clash detection. During tier one, the architects, MEP designers, unstructured interviews with members of the PMT and highlighted

214
E.A. Pärn et al. Automation in Construction 85 (2018) 209–219

Fig. 6. Probability density function – Generalized Gamma

outliers excluded.

several important challenges facing practitioners working within a di- were identified and resolved afforded limited opportunity for members
gital construction environment. These challenges can be conveniently of the PMT to learn from mistakes made by maximizing upon readily
grouped into the following thematic groupings, namely: organizational available business intelligence. This problem is further exacerbated by
influences; manpower and training; automation of analysis (machine software and model exchange issues when different members of the
learning); and cross industry knowledge transfer. PMT work on design work sets in isolation; a member of the PMT said:
“For example, the structural engineer could do a lot of work and not tell
5.1. Organizational influences the architects about it. This might happen, then both could upload their
model into a centralised location and now we have multiple clashes be-
BIM has been heralded as a 21st century innovation that will not cause the architects did not update their model and the structural en-
only improve the efficiency of geometric modelling of a building's gineer has now done some changes to the steel frame.”
performance but also the management of construction projects [49].
Other researchers eulogize over BIM virtues pertaining to: energy sav- This finding concurs with earlier research conducted by Porwal and
ings and concomitant cost reductions [50]; greater control of the de- Hewage [28] who reported that organizational and people centered
sign, construction and operation of an asset throughout its whole life issues pose the greatest challenge for BIM implementation. Other or-
cycle [51,52]; and significant time savings in the production process ganizational issues relate to intellectual property (IP) rights particularly
and consistency of the product [53,54]. However, the research pre- for architectural designs; a member of the PMT said:
sented here observed that a singular PMT is neither cohesive nor unified “They [architects] are still failing to produce a coordinated design even
and consists of disparate teams working together to populate the fed- though they are sitting next to each other [with other design members in
erated BIM model. Moreover, the mechanistic manner via which clashes

Fig. 7. Cumulative distribution function – Generalized

Gamma outliers excluded.

215
E.A. Pärn et al. Automation in Construction 85 (2018) 209–219

Table 3
Probabilities of incurring a clash magnitude (range in mm).

Probability of incurring a clash magnitude (range in mm) P(X < X1) P(X > X1) P(X1 < X < X2) P(X < X2) P(X > X2)

30–99 mm 1.4919E-5 0.99999 0.19852 0.19853 0.80147

100–199 mm 0.20364 0.79636 0.51057 0.71421 0.28579
200–299 mm 0.71779 0.28221 0.22085 0.93864 0.06136
300–399 mm 0.9398 0.0602 0.05611 0.99591 0.00409
400–470 mm 0.99608 0.00392 0.00385 0.99993 7.0710E − 5

Fig. 8. Model federation and clash management.

the PMT]. This is all about intellectual property [IP] rights. Because of research supports this assertion and suggests that some members of the
the IP, the architects that own the model don't want you to easily edit it, PMT have deliberately created a pretense of full BIM compliance, when
so for example when you ask them for the Revit file they will refuse to in fact their approach is compromised by ad hoc arrangements. A
share it. This is because models are easily editable in Revit (you can member of the PMT said:
design in Revit) and once they give you a Revit model you can copy it and
“It's all about knowledge, how the software is used. At the moment a lot
paste it somewhere else. And they [architects] can charge you for it…”
of the consultancies are running away with BIM, where they are just
Cumulatively, these improvised communication, organizational and modelling using the CAD drawings. Rather than using a proper BIM
administrative arrangements make clash eradication per se difficult draughtsman, they employ a Revit technician. The Revit technician re-
within a BIM environment particularly when a silo mentality prevails. ceives CAD drawings and redraws these into Revit, which is not a col-
laborative way of working. The structural engineer is doing all the cal-
culations and measurements in the CAD drawings in 2D and then this is
5.2. Manpower, training and competence development being transferred into 3D with errors!”

Prior research [55,56] advocates that professionals within the PMT

must develop core BIM competencies in order to secure performance
improvement. Such improvement could be achieved via organizational Evidence suggests that a huge BIM knowledge gap has developed
learning that seeks to create, retain and transfer knowledge within an between senior professionals (architects, MEP designers, etc.) and small
organization [57]. The research presented, provides an opportunity for to medium enterprises (SMEs) that is compounded by innate skill lim-
sharing knowledge through the exploitation of business intelligence and itations [60]. SMEs are quintessentially important as their services are
experiential learning amongst members of the PMT [58]. However, often used in the design, construction and/or maintenance of buildings
organizational learning is hampered within industry by the exponential [61]. Higher education institutes (and other education providers) must
rate of software-hardware technology development and the con- collaborate more closely with these practitioners to fully embrace the
comitant need to continually retrain personnel to remain at the fore- concept of a ‘life-long learner for digital construction’ in order to avoid
front of knowledge and developments [59]. Evidence accrued from this tacit knowledge redundancy within SMEs.

216
E.A. Pärn et al. Automation in Construction 85 (2018) 209–219

5.3. Automation of analysis (machine learning) consultants. So for instance, I am getting uploads of the latest models to
the single cloud storage to check them. But I am also coordinating them,
Machine learning (ML) has its entomological roots grounded in ar- so all the clashes which should not be there, are there to be checked by
tificial intelligence (AI) and embraces computer learning without ex- myself and my colleagues. Because the designers have been working
plicit programming [62]. ML focuses on the development of computer within their own silos and then just upload the models into the cloud
programs that can teach themselves to grow and change when exposed based platform for a clash detection.”
to new data [63]. Within the AECO sector, ML is already being used to:
Working from a cloud would alleviate many of the problems and
monitor construction progress using 4D BIM [54,64]; automate rule
issues faced when working in a multi-disciplinary team where software
checking within BIM models [65]; automate as-built 3D reconstruction
and hardware requirements fail to synergize and often require frequent
using computer vision [66]; and monitor construction performance
annual updates. Annual updates in a cloud would ensure that all team
using still images [67,70]. However, despite these significant advances,
members are using the most up to date version. One common cloud-
clash detection remains a laborious, mechanistic, time consuming and
based modelling platform would provide an ideal solution but agree-
costly exercise. Each and every clash must be manually integrated,
ment between five or more software providers of alternative platforms
analyzed and accessed by the BIM manager to first determine the type
could be problematic particularly on commercial grounds. A potential
of clash (i.e. clash errors, pseudo clash, deliberate clash or duplicate clash)
solution would be to eliminate errors within IFCs and ensure ever-
before taking suitable action and monitoring progress where a resolu-
greater interoperability between software vendors – transference of best
tion is required. Automated methods are urgently required to: rapidly
practice from more technologically advanced sectors could present an
assimilate vast quantities of geometric data accessed from a larger
ideal solution to this conundrum. A member of the PMT said:
range of construction and civil engineering projects to build accurate
benchmark clash detection profiles that could inform future decision “The guys [contractors] internally have got their heads around it [cloud
making; define and delineate between the various clash types to pro- based app] because there are a lot of changes. So over the course of the
vide greater business intelligence regards which clashes require re- year the site team has changed slightly. Traditionally, there would be a
solution thus eliminating the need for manual intervention; and elim- lot of information that is stored on emails, although they were sitting next
inate the need for manual intervention and the introduction of human to each other and talking with one another… Because all of the issues
errors or omissions. have been raised on the iPads [on cloud] they are already there for the
next site manager to find. So at least they're not completely blind when
5.4. Cross industry software-knowledge transfer they have to come in to resolve the issues.”

In other more technologically advanced industries (e.g. automotive

and aerospace), software exchange file formats have been standardized 6. Conclusions
to aid communication between various designers and manufacturing
production processes [68]. Within the AECO sector the BIM authoring Despite the euphoria that often surrounds digital construction
platforms adopted lack standardized user interfaces and file formats in within extant literature, this research has shown that BIM is not yet a
an open architecture environment. Although the Industry Foundation panacea to mitigating design errors. Rather the nature of design error
Classes (IFCs) specification sought to alleviate these issues, anecdotal propagation has changed and evolved in parallel with ‘new technolo-
evidence from practitioners suggests that IFCs are not error free. For gies’ applied that are being managed by ‘traditional management’
example, geometry and semantic information can disappear when file processes and procedures. In addition, a distinct lack of organizational
formats are exported from the original BIM authoring platform. A learning within the PMT was evident and so the opportunity to secure
member of the PMT said: experiential learning is often lost. Rather than learn from clash occur-
rences and proactively work to mitigate them, members of the PMT
“… many companies and consultancies are reluctant to give us the Revit
take a short-term reactive approach to identifying and resolving them.
files. That is why the IFC was invented and generated, to allow for the
Part of the problem is that clash detection software for example, cur-
export from any piece of software. This was the holy grail of the BIM
rently lacks automation and requires a labour intensive (and costly)
model, that you can export into a single format which can be opened by
analysis and post-investigation of clash data by the BIM manager/co-
any company or any BIM software vendor and federated in IFC's. But
ordinator. For an entire project (dependent upon scope), design clashes
obviously software vendors [vendor name removed] are failing to pro-
alone could equate to several tens of thousands of observations and in
duce usable IFC's, so it's very hard to export correct IFC from Revit. For
the longer term, such an approach is untenable. Members of a frag-
example, today I received some export IFC's from a vendor [vendor name
mented design team were also observed to be working in isolation and
removed] and they are coming out with strange geometries that are not
with bespoke BIM authoring platforms. Although IFCs were meant to
meant to be in the model.”
overcome this issue, errors with IFCs still doggedly persist.
Currently, there is no commercially available cloud-based BIM au- PDF and CDF probability distribution models developed within this
thoring platform that allows designers to work collaboratively. As an research offer invaluable insight into the size and frequency of clash
exemplar of contemporary industry practice, members of the project occurrence – such could be used to develop probability profiles that
design team worked within separate BIM authoring platforms – for enable BIM managers to better define and delineate tolerances prior to
instance, the architect used REVIT, the structural engineer used Tekla conducting clash detection. Such work should be extended to other
and MEP used REVIT MEP. These various software packages, processes building compartments (for example, architecture) and for other
and procedures have been developed organically and iteratively to meet buildings so that a comprehensive knowledge bank of benchmark in-
industry needs but as yet, a single system that encapsulates holistic dicators can be established and used to monitor clash errors, resolution
coverage has eluded the sector. This is most likely because platform and mitigation.
design specifications are often ill-defined, frequently complex and in- In many instances reported upon in this research, a 21st century
volve iterative processes, and user needs and specifications evolve as technological innovation and collaborative means of working is being
the temporal and recurrent ‘design to user-experience’ process con- managed by a 20th century management and individualistic mentality.
solidates into an optimal product solution [69]. A member of the PMT Future work is therefore required in several key areas, namely to: i)
said: extend the models developed to other building compartments to cover a
wider range of clash detection across the entire building and multiple
“BIM 360 Glue allows you to view and federate the models from different
buildings throughout industry. Such work could form the basis of

217
E.A. Pärn et al. Automation in Construction 85 (2018) 209–219

invaluable business intelligence that would inform and optimize deci- schedule overruns in construction projects, J. Oper. Res. Soc. 64 (8) (2013)
1231–1247, http://dx.doi.org/10.1057/jors.2013.29.
sion making for future design projects; ii) develop machine learning [24] P.E.D. Love, X. Wang, Sing, C.-p., and Tiong, R. L. K., Determining the probability of
processes and procedures to automate clash analysis and prognosis; iii) project cost overruns, J. Constr. Eng. Manag. 139 (3) (2013) 321–330, http://dx.
transfer knowledge of successful digital modelling technologies from doi.org/10.1061/(asce)co.1943-7862.0000575.
[25] P.E.D. Love, R. Lopez, J.T. Kim, M.J. Kim, Probabilistic assessment of design error
other more advanced industrial sectors (such as mitigating interoper- costs, J. Perform. Constr. Facil. 28 (3) (2014) 518–527, http://dx.doi.org/10.1061/
ability issues and clash error management) into the AECO sector; and (ASCE)CF.1943-5509.0000439.
re-evaluate the training and competence development needs of SMEs [26] S. Lee, F. Peña-Mora, M. Park, Quality and change management model for large
scale concurrent design and construction projects, J. Constr. Eng. Manag. 131 (8)
working within the PMT supply chain. (2005) 890–902, http://dx.doi.org/10.1061/(ASCE)0733-9364(2005)131:8(890).
[27] Y. Arayici, C.O. Egbu, P. Coates, Building information modelling (BIM) im-
References plementation and remote construction projects: issues, challenges, and critiques,
Electron. J. Inf. Technol. Constr. 17 (2012) 75–92. Available via: http://usir.
salford.ac.uk/22736/1/BIM_AND_REMOTE_CONSTRUCTION_PROJECTS.pdf (ac-
[1] D.J. Edwards, E.A. Pärn, P.E.D. Love, H. El-Gohary, Machines, manumission and cessed: November, 2016).
economic machinations, J. Bus. Res. 70 (2016) 391–394, http://dx.doi.org/10. [28] A. Porwal, K.N. Hewage, Building information modelling (BIM) partnering frame-
1016/j.jbusres.2016.08.012. work for public contracts, Autom. Constr. 31 (2013) 204–214, http://dx.doi.org/
[2] S.C. Park, S.Y. Ryoo, An empirical investigation of end-users' switching toward 10.1016/j.autcon.2012.12.004.
cloud computing: a two factor theory perspective, Comput. Hum. Behav. 29 (1) [29] C. Merschbrock, B.E. Munkvold, Effective digital collaboration in the construction
(2013) 160–170, http://dx.doi.org/10.1016/j.chb.2012.07.032. industry – a case study of BIM deployment in a hospital construction project,
[3] A. Lin, N.-C. Chen, Cloud computing as an innovation: perception, attitude and Comput. Ind. 73 (2015) 1–7, http://dx.doi.org/10.1016/j.compind.2015.07.003.
adoption, Int. J. Inf. Manag. 32 (6) (2012) 533–540, http://dx.doi.org/10.1016/j. [30] S. Whang, R. Flanagan, S. Kim, S. Kim, Contractor-led critical design management
ijinfomgt.2012.04.001. factors in high-rise building projects involving multinational design teams, J.
[4] D. Dutta, I. Bose, Managing a big data project: the case of Ramco Cements Limited, Constr. Eng. Manag. (2016), http://dx.doi.org/10.1061/(ASCE)CO.1943-7862.
Int. J. Prod. Econ. 165 (2015) 293–306, http://dx.doi.org/10.1016/j.ijpe.2014.12. 0001242.
032. [31] N. Forcada, A. Alvarez, P. Love, D.J. Edwards, Rework in urban renewal projects in
[5] Manyika, J., Chui, M., Brown, B., Bughin, J., Dobbs, R., Roxburgh, C., and Byers, A. Colombia, J. Infrastruct. Syst. (2016), http://dx.doi.org/10.1061/(ASCE)IS.1943-
H. (2011) Big data: the next frontier for innovation, competition, and productivity. 555X.0000332.
McKinsey Global Institute. Available via: file:///C:/Users/pc%20user/Downloads/ [32] W. Solihin, C. Eastman, Y.C. Lee, A framework for fully integrated building in-
MGI_big_data_full_report.pdf (Accessed: November, 2016). formation models in a federated environment, Adv. Eng. Inform. 30 (2) (2016)
[6] Russom, P. (2013) Managing big data. TDWI best practices report, TDWI research, 168–189, http://dx.doi.org/10.1016/j.aei.2016.02.007.
Vol., No. Available via: https://www.pentaho.com/sites/default/files/uploads/ [33] V. Peansupap, R. Ly, Evaluating the impact level of design errors in structural and
resources/tdwi_best_practices_report-_managing_big_data.pdf pp. 1–40 (Accessed: other building components in building construction projects in Cambodia, Protein
November, 2016). Eng. 123 (2015) 370–378, http://dx.doi.org/10.1016/j.proeng.2015.10.049
[7] P.B. Seddon, D. Constantinidis, T. Tamm, H. Dod, How does business analytics Available via: http://2015.creative-construction-conference.com/CCC2015_
contribute to business value? Inf. Syst. J. (2016), http://dx.doi.org/10.1111/isj. proceedings/CCC2015_45_Peansupap.pdf (Accessed: November, 2016).
12101. [34] M. Al Hattab, F. Hamzeh, Using social network theory and simulation to compare
[8] A. Shollo, R.D. Galliers, Towards an understanding of the role of business in- traditional versus BIM–lean practice for design error management, Autom. Constr.
telligence systems in oganisational knowing, Inf. Syst. J. 26 (2016) 339–367, 52 (2015) 59–69, http://dx.doi.org/10.1016/j.autcon.2015.02.014.
http://dx.doi.org/10.1111/isj.12071. [35] R.K. Allen, B. Becerik, S.N. Pollalis, B.R. Schwegler, Promise and barriers to tech-
[9] P.E.D. Love, J. Liua, J. Matthews, C.-P. Sing, J. Smith, Future proofing PPPs: life- nology enabled and open project team collaboration, J. Prof. Issues Eng. Educ.
cycle performance measurement and building information modelling, Autom. Pract. 131 (4) (2005) 301–311, http://dx.doi.org/10.1061/(ASCE)1052-
Constr. 56 (2015) 26–35, http://dx.doi.org/10.1016/j.autcon.2015.04.008. 3928(2005)131:4(301).
[10] J. Park, K. Kim, Y. Cho, Framework of automated construction-safety monitoring [36] J.M. Chevalier, D.J. Buckles, Participatory Action Research: Theory and Methods
using cloud-enabled BIM and BLE mobile tracking sensors, J. Constr. Eng. Manag. for Engaged Inquiry, Routledge, London, 2013 ISBN: 0415540321.
(2016), http://dx.doi.org/10.1061/(ASCE)CO.1943-7862.0001223. [37] L. Smith, L. Ronsenzweig, M. Schmidt, Best practices in the reporting of partici-
[11] P. Oskouiel, B. Becerik-Gerber, L. Soibelman, Automated measurement of highway patory action research: embracing both the forest and the trees, Couns. Psychol. 38
retaining wall displacements using terrestrial laser scanners, Autom. Constr. 65 (8) (2010) 1115–1138, http://dx.doi.org/10.1177/0011000010376416.
(2016) 86–101, http://dx.doi.org/10.1016/j.autcon.2015.12.023. [38] P. Mapfumo, S. Adjei-Nsiah, F. Mtambanengwe, R. Chikowo, K.E. Giller,
[12] J. Teizer, Status quo and open challenges in vision-based sensing and tracking of Participatory action research (PAR) as an entry point for supporting climate change
temporary resources on infrastructure construction sites, Adv. Eng. Inform. 29 (2) adaptation by smallholder farmers in Africa, Environ. Dev. Econ. 5 (2013) 6–22,
(2015) 225–238, http://dx.doi.org/10.1016/j.aei.2015.03.006. http://dx.doi.org/10.1016/j.envdev.2012.11.001.
[13] L. Ben-Alon, R. Sacks, Simulating the Behavior of Trade Crews in Construction [39] R. Pain, M. Finn, R. Bouveng, G. Ngobe, Productive tensions - engaging geography
Using Agents and Building Information Modeling, Automation in Construction, Vol. students in participatory action research with communities, J. Geogr. High. Educ.
74 (2017), pp. 12–27, http://dx.doi.org/10.1016/j.autcon.2016.11.002. 37 (1) (2012) 28–43, http://dx.doi.org/10.1080/03098265.2012.696594.
[14] Y. Liu, S.V. Nederveen, M. Hertogh, Understanding effects of BIM on collaborative [40] M. Kornbluh, E.J. Ozer, C.D. Allen, B. Kirshner, Youth participatory action research
design and construction: an empirical study in China, Int. J. Proj. Manag. (2016), as an approach to sociopolitical development and the new academic standards:
http://dx.doi.org/10.1016/j.ijproman.2016.06.007. considerations for educators, Urban Rev. 47 (5) (2015) 868–892, http://dx.doi.org/
[15] A.L.C. Ciribini, S. Mastrolembo Ventura, M. Paneroni, Implementation of an 10.1007/s11256-015-0337-6.
Interoperable Process to Optimize Design and Construction Phases of a Residential [41] S. Ding, S.L. Yang, C. Fu, A Novel Evidential Reasoning Based Method for Software
Building: A BIM Pilot Project, Automation in Construction, Vol. 71, Part 1 (2016), Trustworthiness Evaluation Under the Uncertain and Unreliable Environment,
pp. 62–73, http://dx.doi.org/10.1016/j.autcon.2016.03.005 (The Special Issue of Expert Systems with Applications, Vol. 39, No. 3 (2012), pp. 2700–2709, http://dx.
32nd International Symposium on Automation and Robotics in Construction). doi.org/10.1016/j.eswa.2011.08.127.
[16] E.M. Wetzel, W.Y. Thabet, Utilizing six sigma to develop standard attributes for a [42] J.M. Wittmayer, N. Schäpke, Action, research and participation: roles of researchers
safety for facilities management (SFFM) framework, Saf. Sci. 89 (2016) 355–368, in sustainability transitions, Sustain. Sci. 9 (4) (2014) 483–496, http://dx.doi.org/
http://dx.doi.org/10.1016/j.ssci.2016.07.010. 10.1007/s11625-014-0258-4.
[17] P.E.D. Love, J. Zhou, J. Matthews, H. Luo, Systems information modelling: enabling [43] T.Y.M. Lam, A performance outcome framework for appraising construction con-
digital asset management, Adv. Eng. Softw. 102 (2016) 155–165, http://dx.doi.org/ sultants in the university sector, J. Facil. Manag. 14 (3) (2016) 249–265, http://dx.
10.1016/j.advengsoft.2016.10.007. doi.org/10.1108/JFM-05-2015-0017.
[18] P. Bagwat, R. Shinde, Clash Detection: A New Tool in Project Management, [44] HM Government. (2012) Final report to government by the procurement/lean client
International Journal of Scientific Research in Science, Engineering and task group. London: Government Construction Strategy. Available via: https://
Technology, Vol. 2, No. 4, (2016), pp. 193–197. Available via: http://ijsrset.com/ www.gov.uk/government/uploads/system/uploads/attachment_data/file/61157/
paper/1637.pdf (Accessed: November, 2016). Procurement-and-Lean-Client-Group-Final-Report-v2.pdf (Accessed: November,
[19] J. Won, G. Lee, How to tell if a BIM project is successful: a goal-driven approach, 2016).
Autom. Constr. 69 (2016) 34–43, http://dx.doi.org/10.1016/j.autcon.2016.05.022. [45] HM Government. (2013) Building information modeling industrial strategy: gov-
[20] Y. Li, T. Taylor, Modeling the impact of design rework on transportation infra- ernment and industry in partnership. London: Government Construction Strategy.
structure construction project performance, J. Constr. Eng. Manag. 140 (9) (2014), Available via: https://www.gov.uk/government/uploads/system/uploads/
http://dx.doi.org/10.1061/(ASCE)CO.1943-7862.0000878. attachment_data/file/34710/12-1327-building-information-modelling.pdf
[21] R. Lopez, P.E.D. Love, D.J. Edwards, P.R. Davis, Design error classification, cau- (Accessed: November, 2016).
sation, and prevention in construction engineering, J. Perform. Constr. Facil. 24 (4) [46] C. Wheelan, Naked Statistics: Stripping the Dread from the Data, W.W. Norton and
(2010) 399–408, http://dx.doi.org/10.1061/(ASCE)CF.1943-5509.0000116. Company, London, 2013 ISBN: 978-0-393-07195-5.
[22] P.E.D. Love, C.-P. Sing, Determining the probability distribution of rework costs in [47] J.J. Schiller, A.R. Srinivasan, M.R. Spiegel, Schaum's Outline of Probability and
construction and engineering projects, Struct. Infrastruct. Eng. 9 (11) (2013) Statistics, 4th Edition, McGraw-Hill, London, 2013 ISBN: 978-0-07-179558-9.
1136–1148, http://dx.doi.org/10.1080/15732479.2012.667420. [48] D. Cousineau, S. Chartier, Outliers detection and treatment: a review, Int. J.
[23] P.E.D. Love, C.P. Sing, D.J. Edwards, H. Odeyinka, Probability distribution fitting of Psychol. Res. 3 (1) (2010) 58–67. Available via: http://revistas.usb.edu.co/index.

218
E.A. Pärn et al. Automation in Construction 85 (2018) 209–219

php/IJPR/article/view/844/601 (Accessed: November 2016). dynamics on construction, projects, Constr. Manag. Econ. 34 (3) (2016) 192–204,
[49] D. Bryde, M. Broquetas, J.M. Volm, The project benefits of building information http://dx.doi.org/10.1080/01446193.2016.1190026.
modelling, Int. J. Proj. Manag. 31 (7) (2013) 971–980, http://dx.doi.org/10.1016/ [62] L. Bottou, From machine learning to machine reasoning, Mach. Learn. 94 (2) (2014)
j.ijproman.2012.12.001. 133–149, http://dx.doi.org/10.1007/s10994-013-5335-x.
[50] S.-J. Guo, T. Wei, Cost-Effective Energy Saving Measures Based on BIM Technology: [63] C. Perlich, B. Dalessandro, T. Raeder, O. Stitelman, F. Provost, Machine learning for
Case Study at National Taiwan University, Energy and Buildings, Vol. 127 (2016), targeted display advertising: transfer learning in action, Mach. Learn. 95 (1) (2014)
pp. 433–441, http://dx.doi.org/10.1016/j.enbuild.2016.06.015. 103–127, http://dx.doi.org/10.1007/s10994-013-5375-2.
[51] S. Azhar, Building Information Modeling (BIM): Trends, Benefits, Risks, and [64] H. Son, F. Bosche, C. Kim, As-built data acquisition and its use in production
Challenges for the AEC Industry. Leadership Management in Engineering, Vol. 11, monitoring and automated layout of civil infrastructure: a survey, Adv. Eng. Inform.
No. 3, (2011), pp. 241–252, http://dx.doi.org/10.1061/(ASCE)LM.1943-5630. 29 (2) (2015) 172–183, http://dx.doi.org/10.1016/j.aei.2015.01.009.
0000127. [65] W. Solihin, C. Eastman, Classification rules for automated BIM rule checking de-
[52] J.K.W. Wong, J. Zhou, Enhancing environmental sustainability over building life velopment, Autom. Constr. 53 (2015) 68–82, http://dx.doi.org/10.1016/j.autcon.
cycles through green BIM: a review, Autom. Constr. 57 (2015) 156–165, http://dx. 2015.03.003.
doi.org/10.1016/j.autcon.2015.06.003. [66] H. Fathi, F. Dai, M. Lourakis, Automated as-built 3D reconstruction of civil infra-
[53] Y. Arayici, P. Coates, L. Koskela, K. Kagioglou, C. Usher, K. O'Reilly, Technology structure using computer vision: achievements, opportunities, and challenges, Adv.
Adoption in the BIM Implementation for Lean Architectural Practice, Building Eng. Inform. 29 (2) (2015) 149–161, http://dx.doi.org/10.1016/j.aei.2015.01.012.
Information Modeling and Changing Construction Practices, Automation in [67] J. Yang, M.-W. Park, P.A. Vela, M. Golparvar-Fard, Construction performance
Construction, Vol. 20, No. 2, (2011), pp. 189–195, http://dx.doi.org/10.1016/j. monitoring via still images, time-lapse photos, and video streams: now, tomorrow,
autcon.2010.09.016. and the future, Adv. Eng. Inform. 29 (2) (2015) 211–244, http://dx.doi.org/10.
[54] Y. Ham, M. Golparvar-Fard, Mapping actual thermal properties to building ele- 1016/j.aei.2015.01.011.
ments in GBXML-based BIM for reliable building energy performance modeling, [68] C. Eastman, C.M. Eastman, P. Teicholz, R. Sacks, K. Liston, BIM Handbook: A Guide
Autom. Constr. 49 (Part B) (2015) 214–244, http://dx.doi.org/10.1016/j.autcon. to Building Information Modeling for Owners, Managers, Designers, Engineers and
2014.07.009. Contractors, John Wiley & Sons, New Jersey, USA, 2011 ASIN: B01JXSY6Q8.
[55] M.E. Murphy, Implementing innovation: a stakeholder competency-based approach [69] S.K. Chandrasegaran, K. Ramani, R.D. Sriram, I. Horvath, A. Bernard, R.F. Harik,
for BIM, Constr. Innov. 14 (4) (2014) 433–452, http://dx.doi.org/10.1108/CI-01- W. Gao, The evolution, challenges, and future of knowledge representation in
2014-0011. product design systems, Comput. Aided Des. 45 (2) (2013) 204–228, http://dx.doi.
[56] B. Succar, W. Sher, A. Williams, An integrated approach to BIM competency as- org/10.1016/j.cad.2012.08.006.
sessment, acquisition and application, Autom. Constr. 35 (2013) 174–189 DOI [70] K.K. Han, M. Golparvar-Fard, Appearance-based material classification for mon-
http://dx.doi.org/10.1016/j.autcon.2013.05.016. itoring of operation-level construction progress using 4D BIM and site photologs,
[57] S.M. Duffield, S.J. Whitty, Application of the systemic lessons learned knowledge Autom. Constr. 53 (2016) 44–57, http://dx.doi.org/10.1016/j.autcon.2015.02.007.
model for organisational learning through projects, Int. J. Proj. Manag. 34 (7) [71] W. Zhou, J. Whyte, R. Sacks, Construction Safety and Digital Design: A Review,
(2016) 1280–1293, http://dx.doi.org/10.1016/j.ijproman.2016.07.001. Autom. Constr. 22 (2012) 102–111, http://dx.doi.org/10.1016/j.autcon.2011.07.
[58] A. Konak, T.K. Clark, M. Nasereddin, Using Kolb's experiential learning cycle to 005.
improve student learning in virtual compute laboratories, Commun. Educ. 72 [72] N. Young, S. Jones, H. Bernstein, J. Gudgel, The business value of BIM: getting
(2014) 11–22, http://dx.doi.org/10.1016/j.compedu.2013.10.013. building information modeling to the bottom line, McGrawHill Smart Market
[59] R. Eadie, M. Browne, H. Odeyinka, C. McKeown, S. McNiff, BIM implementation Report, 2009 http://www.trane.com/architect/Files/PDF/SMR%20BIM
throughout the UK construction project lifecycle: an analysis, Autom. Constr. 36 %2009%20FINAL%20rev.pdf.
(2013) 145–151, http://dx.doi.org/10.1016/j.autcon.2013.09.001. [73] Peter E.D. Love, David J. Edwards, Hunna Watson, Peter Davis, Rework in civil
[60] R. Harris, R. McAdam, I. McCausland, R. Reid, Levels of innovation within SMEs in infrastructure projects: determination of cost predictors, ASCE, J. Constr. Eng.
peripheral regions: the role of business improvement initiatives, J. Small Bus. Manag. 13 (3) (2010), http://dx.doi.org/10.1061/(ASCE)CO.1943-7862.0000136.
Enterp. Dev. 20 (1) (2013) 102–124, http://dx.doi.org/10.1108/ [74] C.B. Tatum, T.M. Korman, MEP Coordination in Building and Industrial Projects,
14626001311298439. CIFE Work 54, March 1999, CIFE, Stanford University, Stanford, CA, 1999.
[61] K.I.A. Khan, R. Flanagan, S.-L. Lu, Managing information complexity using system

219