Automation in Construction 113 (2020) 103129

Contents lists available at ScienceDirect

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

Automatic generation of building information models from digitized plans T

a b,⁎
Omar Doukari , David Greenwood
a
CESI Centre de Paris Nanterre, 93 Boulevard de la Seine BP 602 Cedex, 92006 Nanterre, France
b
Northumbria University, Department of Mechanical & Construction Engineering, Faculty of Engineering and Environment, 205 Wynne Jones Centre, Newcastle-upon-Tyne
NE1 8ST, UK

A R T I C LE I N FO A B S T R A C T

Keywords: This paper proposes a new approach to creating Building Information (BIM) models of existing buildings from
Artificial intelligence digitized images. This automatic approach is based on three main steps. The first involves extracting the useful
Automation information automatically from rasterized plans by using image processing techniques that include segmenta-
Digitized plans tion, filtering, dilation, erosion, and contour detection. This information feeds the knowledge base of an expert
Expert system
system for BIM model generation. In the second step, using the knowledge base of the expert system, the in-
Knowledge base
formation required to inform the BIM model can be deduced. The range of information thus obtainable can be
extended beyond the examples given. The paper concludes with a discussion of the final stage: the automatic
generation of an Industry Foundation Classes (IFC) information model with all the desired geometric, physical
and technical information. This can be accomplished by using one of the available open-source application
program interfaces (APIs). This stage is currently work-in-progress and will be the subject of a future publication.

1. Introduction might be reproduced automatically using artificial intelligence with

tools such as neural networks, expert systems, and genetic algorithms.
A Building Information Model (BIM) is a database containing in- Expert systems have for many years been used for different purposes
formation relating to a built asset; more fully defined by the and applied within several domain applications, for example: in-
International Organization for Standardization as a “shared digital re- formation retrieval (e.g. [36]), medical diagnosis [12], help desk
presentation of a built object” and “the shared digital representation of management [10], performance evaluation [19], loan analysis [24] and
the physical and functional characteristics of any construction works” computer virus detection [32]. Construction-related applications in-
[13]. It can represent not just the geometry of the building and its clude design optimization [1,6,18], logistical problems [38] and the
contents, but all its physical and technical characteristics [2,8]. Over enrichment or validation of BIM models [20,21,26,31].
the last ten years, as recently summarized by Zhao [39] BIM has be- In all such applications, the first step is generally to represent
come the most discussed and utilized new technological tool in the field human expertise in a machine-readable format, then to define rea-
of construction. However, though its usefulness and benefits have been soning operators that can, based on certain information, draw relevant
demonstrated in several fields of application (e.g. [14,16,17]) the conclusions. In this paper, we focus our attention on the development of
creation of a BIM model can be a laborious task that requires the col- a new approach to creating BIM models of existing buildings from di-
laboration of several modelling teams over time. This is the case with a gitized images. This automatic approach is based on three main steps.
new asset: when it comes to modelling an existing asset (for example, The first involves extracting the useful information automatically from
retro-modelling an existing building to take advantage of a digital the digitized plans (in .TIF, .JPG or .PNG image formats) by using image
model for the purposes maintenance of facilities management) this can processing techniques (including segmentation, filtering, dilation, ero-
quickly become particularly expensive. First, there is the use of addi- sion, and contour detection): this information will feed the knowledge
tional expensive equipment, such as laser scanning apparatus and post- base of an expert BIM model generation system. Secondly, using the
processing and modelling software [7,29,35]. Thereafter expert inter- knowledge base of the expert system, the information that will inform
vention is required to identify certain information and characteristics of the BIM model can be deduced. Finally, an Industry Foundation Classes
the various components such as types of materials [37]. (IFC) model can be automatically generated with all the desired geo-
On the other hand, the cognitive processes of the human expert metric, physical and technical information. We present a proof of

⁎
Corresponding author.
E-mail addresses: [email protected] (O. Doukari), [email protected] (D. Greenwood).

https://doi.org/10.1016/j.autcon.2020.103129
Received 8 November 2019; Received in revised form 26 December 2019; Accepted 7 February 2020
Available online 26 February 2020
0926-5805/ © 2020 Published by Elsevier B.V.
O. Doukari and D. Greenwood Automation in Construction 113 (2020) 103129

arguably the most important information within such models is the

semantic, non-geometric information required for auto-populating
computerized facility management systems [28]. The human expertise
that is required for providing this kind of information remains the most
difficult part to automate, however, the use of artificial intelligence
may ameliorate this situation.
In the realm of Artificial Intelligence, a knowledge base is a tech-
nology used to store complex structured and unstructured information
used by a computer system. The first use of the term was in relation to
the expert systems that were the first knowledge-based systems; com-
Fig. 1. Conceptual view of an expert system. puter systems that emulate the decision-making abilities of human ex-
perts. The term “knowledge base” was adopted to distinguish itself from
concept as well as the conceptual model of an expert system for the first the widely-used “database”, as by the 1970s, most large information
stages in the automatic generation of such BIM models. An algorithm systems managed their data stored in hierarchical or relational data-
was developed using Python programming language and existing bases. An expert system is principally composed of two modules: an
Python image processing modules (OpenCV, Numpy, SciPy, Scikit- inference engine and a knowledge base (Fig. 1). The knowledge base
image and Matplotlib) were used for extracting information from di- includes a set of defined rules that serve as a reference for extracted
gitized plans. This process is presented in the next section together with facts. The inference engine applies the rules to known facts to infer new
a knowledge base, represented in the form of rules of production. The facts and new information. In some cases, an inference engine can also
successful application of such a system would help to overcome the provide explanations for the results obtained.
constraints of time and cost when creating a BIM model of an existing Expert systems are designed to solve complex problems by rea-
built asset. soning based on knowledge representation formalisms that involve
various types of rules frames and ontologies (see, for example, [27]).
The approach that we adopted involves propositional logic encoding
2. Methodology and is shown in Fig. 2 and presented in the following text. As an ex-
ample of semantic information that we are seeking, we selected the
Many national and state governments have mandated the use of BIM categories of ‘Dwelling type’ and (predominant) ‘Construction mate-
for new buildings and infrastructure [15] and have also recognised the rials’ (e.g. ‘brick’ or ‘concrete’).
advantages of having models of built assets that already exist (see, for The functionality of the building expert system that we are aiming
example, [23]). However, the creation of models of the existing stock is to build is based upon the process stipulated in Fig. 2. Based on an
a time-consuming and expensive exercise, currently requiring a sig- expert knowledge base, it must be able to deduce useful information
nificant amount of expert human intervention [34]. The work described (for the present purposes, ‘Dwelling type’ and ‘Construction materials’)
here concerns the automatic creation of IFC BIM models of existing that would normally be provided by human experts for incorporation
buildings from digitized images. It was carried out in Paris and used test into the corresponding BIM model.
case buildings within the suburb of Nanterre. In France, as elsewhere,
very few existing buildings enjoy the advantages of having digital
models that can facilitate their operation, and fewer still have been 2.1. Building the knowledge base
originally modelled in BIM. In order to accelerate this digital transition
and assist the modelling of existing buildings, automatic and fast ap- As noted by Hayes-Roth [11] the knowledge base of any expert
proaches need to be defined and implemented. As with new buildings, system is its most important component and building this component

Fig. 2. Building expert system.

2
O. Doukari and D. Greenwood Automation in Construction 113 (2020) 103129

Table 1
Knowledge base of dwelling types with typical surface areas and construction materials.
Min. density Max. density Dwelling type Surface area (m2) Materials

1 4 Suburban villa 180 Block – Tile – Concrete - Wood

5 8 Housing estate 130 Waterproofed Insitu concrete
9 10 Individual grouped 125 Stone – Concrete – Brick – Wood
11 15 Detached town house 116 Brick – Concrete
16 50 Single terraced 108 Brick – Concrete
51 80 Intermediate 89 Brick – Stone
81 121 ‘Grand Ensemble’ 78 Brick – Concrete
122 212 Multiple occupancy 69 Insitu reinforced concrete
213 343 High-density multi-occupancy (Centre Bourg) 45 to 90 Stone – Concrete
344 1,000,000 Built-up area (Hausmannian fabric) 30 to 120 Brick – Concrete – Brick

Fig. 3. Stages in the algorithm for Housing density generation.

represents a fundamental step in creating the expert system. In this a) The Building (B1) is situated in Region R;
paper, we present a simplified prototype version of the knowledge base b) If the Housing density observed is between 5 and 8, then:
that simply allows it to deduce information, for incorporation into a Dwelling type = Housing estate and
BIM model, relating to ‘Dwelling type’ and ‘Construction materials’ (see Construction materials = Waterproofed Insitu concrete.
Table 1). We intend subsequently to develop and enrich this initial
proof-of-concept version to take into account other types of informa- A very simple representation in propositional language can be:
tion. In France, buildings are classified by geographical zone (urban,
suburban, rural, etc.) and according to dwelling type (see [30]). In − To calculate the Housing density of Region R: such as c) R6 (The
order to simplify automation, we have here classified the different Housing density of Region R is 6). Rx (The Housing density of Region
dwelling types (e.g. as ‘detached’, ‘semi-detached’, ‘terraced’, ‘dense R is x).
housing complex’, etc.) in line with the types that commonly occur − To infer consequences from rules (a), (b) and (c), on this list: B1 is a
throughout the regions of France. Housing estate and B1
Table 1 shows a simplified version of the knowledge base as far as it Construction material is Waterproofed Insitu concrete.
currently extends. For these purposes ‘Housing density’ is defined as
Number of dwellings/Site area. To enable this, we must first address ‘Housing density’ and its means
in hectares (Ha). of calculation as explained in the next section.
From the knowledge base created, the type of housing and the type
of materials that it is typically composed of can be readily deduced. 2.2. Base information: housing density
The most commonly-used representation of knowledge-based rea-
soning is propositional language. Representing human expertise with The information base of our expert system consists of a set of sa-
propositional language takes advantage of the simplicity of expression tellite images taken on regions, building facades, etc. in .TIFF, .JPEG or
of this language and, from a computational point of view, of its deci- .PNG formats. To enable us to extract information that can be used by
siveness. In addition, most of the reasoning and inference operators the expert system (example: ‘Housing density’) it is necessary to carry
defined in the area of knowledge representation are defined in propo- out image processing.
sitional calculus [9]. Thus, our approach is essentially based upon the use of program-
Consider the statements (derived from Table 1): ming and image-processing tools such as: Python, OpenCV, NumPy,
SciPy, Skimage (scikit-image), and Matplotlib. The approach follows

3
O. Doukari and D. Greenwood Automation in Construction 113 (2020) 103129

Fig. 4. Section of a cadastral map.

the stages that are shown in Fig. 3, each of which is explained in the
subsequent text.
To illustrate how this algorithm works, in the following sections we
apply it step by step on a real-life example with data taken from
Nanterre City (France).

2.2.1. Data
We selected two types of data, namely, cadastral sources (maps) and
satellite images (see Figs. 4 and 5, respectively) of a part of Nanterre, a
suburb of Paris. The original format of the data is .TIFF, .JPEG or .PNG
and each represents a surface area of 1 ha.

2.2.2. Manipulation of the data

By applying the above algorithm to the cadastral map (Fig. 3) we
obtained the results shown in Figs. 6 to 12. These figures are accom-
panied by a short description of the process.
Conversion of image from RGB to grey-scale: To simplify the data
input we have chosen to work with monochromatic images. The RGB
colour images are therefore converted to grey-level, as shown in Fig. 6.
Gaussian filter: The occurrence of random noise information in the
image reduces its sharpness. To reduce the noise an important step is to
smooth the image using the Gaussian filter (Figs. 6 and 7).
Edge detection using Canny filter application: The Canny filter
algorithm [4] is then used to: (i) minimize the error rate in edge de-
tection, (ii) minimize the distance between the detected contours and
the actual contours, and (iii) return a single response by contour. To
draw only the contours, it uses a calculation of the intensity gradient
followed by a hysteresis thresholding of the contours in order to have a
binary image; with the outlines in white and the other points in black
(Fig. 8). Unwanted electronic visual ‘noise’ is suppressed by the hys-
teresis thresholding. This typically requires the input of two user-de-
fined threshold levels, namely: minValue and maxValue. Pixels with an
Fig. 5. Satellite image (Google Earth).
intensity above the maxValue threshold are retained; while those below
minValue are removed. Where intensity falls between the two values, a
further criterion is applied: pixels are retained if they are connected to

4
O. Doukari and D. Greenwood Automation in Construction 113 (2020) 103129

Fig. 6. Source image converted to grey-level.

Fig. 7. Image after ‘smoothing’ using Gaussian filter.

other pixels classified as ‘accepted’ edges. The minValue and maxValue shapes with irregular contours that require treatment by reducing and
threshold values may need to be varied depending upon the data input weakening their shapes into a curve called a skeleton. This enables an
quality and data acquisition method in order to get reliable edge re- average contour to be obtained in cases where the size of the contours is
cognition results. not uniform. In the case shown in Fig. 7, however, the process is not
‘Skeletonization’: This stage may be necessary where there are necessary and skeletonization had no effect.

5
O. Doukari and D. Greenwood Automation in Construction 113 (2020) 103129

Fig. 8. Image resulting from application of Canny filter.

Fig. 9. Image with highlighted contours of interest (338 contours detected).

Extraction, evaluation and selection of contours: The next step is to distinguishing (using colour) those that appear to be of interest, as
select those contours that are of interest: i.e. those that are likely to shown in Fig. 9.
represent buildings (as opposed to vehicles, natural spaces and other Closing contours: There remains the potential problem of unclosed
images that do not represent built assets). This is done by calculating contours. This distorts area calculations, counting, and contour selec-
the area and perimeter of the contours before selecting and tions. An example of this is shown in close-up (zoomed) in Fig. 10.

6
O. Doukari and D. Greenwood Automation in Construction 113 (2020) 103129

2.3. Generation of further information

From the knowledge base presented in Table 1, the type of housing

and the type of materials that it is typically composed of can be readily
deduced. In the case of the previously discussed satellite image
(Fig. 12), the resulting inference is that the dwelling is part of a “col-
lective housing complex” and the material of construction is “brick and
concrete” (see Fig. 14).
Using the techniques exemplified above more information may be
derived from image processing at the urban scale; for example, the
Fig. 10. Image close-up showing unclosed contours.
images of the facade of a building of interest (see, for example, Fig. 15).
More detailed information such as the number of levels, openings,
In order to solve this problem, we have developed a new algorithm doors, rooms, etc. could readily be derived from this second category of
that can detect the ends of open contours and connect them to the images and the Building Knowledge Base can be completed with new
nearest pixels in their vicinity. Some approximations were made during expertise as shown by the examples in Tables 2 & 3.
the tests. In its current state, this algorithm allows at least 70% of open The final step in the process, namely the incorporation of auto-
contours to be closed. As a result, a total of 588 contours were detected matically generated information.
(Fig. 11) as opposed to the original 338 contours. into an IFC BIM model is discussed in the next section.
In order to retain only the contours relevant to our study, i.e. con-
tours potentially representing buildings, the results are again filtered to
keep only contours whose area is between 30 m2 and 1000 m2. (see 2.4. Creation of an IFC BIM model
Fig. 12).
An example of the results following this stage is shown in Fig. 13 The work described is aimed towards the ability to generate and
alongside the corresponding section of the original satellite image. populate an IFC BIM model of an existing building using information
Once the relevant contours have been selected, the useful and that has been automatically generated from images. The attractiveness
usable information is extracted into an Excel file, particularly the area of such an approach for modelling existing assets, and its efficiency and
and perimeter of the contours. It is also possible to extract the co- cost-effectiveness, have long been recognised (see, for example, [3]). As
ordinates of the approximated points of the contours, that is to say, the a result, several open-source application program interfaces (APIs) have
edge points of each segment. become available for this type of operation. One of the most highly-
Calculation of housing density: the housing density represents the regarded of these is the eXtensible Building Information Modelling
number of dwellings (or detected buildings) per site area (Ha). Thus: (xBIM) Toolkit [22]. The xBIM Toolkit platform (at https://docs.xbim.
Housing density = Number of dwellings/Site area (Ha). net/examples/proper-wall-in-3d.html) gives a simple example of how
to generate a 3D parametric IfcWall using its programming functions.
Given that we are working on images whose area is 1 ha then the Using basic information (e.g. length, width, height, materials) about a
number of buildings detected will correspond directly to the density of wall, all of which could be deduced using the Building Expert System
housing. described, a compliant IFC model can be created that contains a

Fig. 11. Image after closure of contours (588 contours detected).

7
O. Doukari and D. Greenwood Automation in Construction 113 (2020) 103129

Fig. 12. Image after selection of chosen contours (178 contours selected and coloured). (For interpretation of the references to colour in this figure legend, the reader
is referred to the web version of this article.)

3. Conclusions, discussion and future work

The use of BIM in the construction and property sectors is in-

creasing, and as it does so, further benefits are becoming evident.
However, the creation of a BIM models can be expensive and time-
consuming, particularly in the case of the existing stock of built assets,
where the required information (both geometric and non-geometric)
needs to be retrofitted into a model using a variety of scanning and
other techniques. Thereafter, human expert intervention is required to
fully develop the information in the model. In this paper, we have ex-
plored the possibility of automating the modelling process by devel-
oping artificial intelligence that replicates and replaces certain of the
cognitive processes that are elements of human expert intervention. We
have illustrated a set of image-processing algorithms that automatically
retrieve information about the area and perimeter of a building, and the
housing density of a one-hectare sample of a region. We have also de-
monstrated the creation of a knowledge base with an expert system able
to deduce new information such as the type of housing or the type of
materials used.
This building knowledge base can be automatically enriched using
Fig. 13. Section of the treated data (4 contours alongside original image of 4
individual houses).
new machine learning techniques, thus enhancing the inference engine.
Furthermore, better building image processing could be obtained by
using new generation deep convolutional neural networks like those
parametric 3D wall object. Iterations of the same process would result described in Cevallos et al. [5] and Mayya et al. [25]. Image pre-pro-
in the creation of other 3D parametric objects (e.g. windows, doors, cessing to remove noise and unwanted features could also be useful to
roof) and their integration within the same IFC model. This last step, enhance contour closure and building component results (see, for ex-
the integration of parametric IFC objects, can also be readily accom- ample, [33]). Using the same methodology our intention is to extend
plished using the framework for merging IFC-Based BIM Models pre- this system to a much wider range of information that could then, using
sented in Doukari et al. [8]. This process also allows checking the available APIs, be used to populate semantically-rich IFC BIM models in
overall consistency of the resulting IFC parametric model. If any in- a quick and inexpensive way. Ultimately, we aim to develop an inter-
consistency is detected, the wrong parameter and its value are high- active graphical user interface for the proposed expert system to assist
lighted in the model tree, facilitating its correction. architects, engineers and construction project managers in executing
advanced BIM-creating tasks.

8
O. Doukari and D. Greenwood Automation in Construction 113 (2020) 103129

178 usable contours

Dwelling type:
Collecve housing complex

Material type:
Brick Concrete

Fig. 14. Results of expert system query.

acknowledge the editorial contributions of Professor Bimal Kumar of

Northumbria University and Dr. Farzad Rahimian of Teesside
University in the publication of this paper.

References

[1] M. Andersen, J.M.L. Gagne, S. Kleindienst, Interactive expert support for early stage
full-year daylighting design: a user’s perspective on Lightsolve, Autom. Constr. 35
(2013) 338–352, https://doi.org/10.1016/j.autcon.2013.05.014.
[2] S. Aram, C. Eastman, R. Sacks, Requirements for BIM platforms in the concrete
reinforcement supply chain, Autom. Constr. 35 (2013) 1–17, https://doi.org/10.
1016/j.autcon.2013.01.013.
[3] Y. Arayici, Towards building information modelling for existing structures, Struct.
Surv. 26 (3) (2008) 210–222, https://doi.org/10.1108/02630800810887108.
[4] J. Canny, A computational approach to edge detection, IEEE Trans. Pattern Anal.
Mach. Intell. (6) (1986) 679–698, https://doi.org/10.1109/TPAMI.1986.4767851.
[5] J.C. Cevallos, J.A. Villagomez, I.S. Andryshchenko, Convolutional neural network
in the recognition of spatial images of sugarcane crops in the Troncal region of the
coast of Ecuador, Procedia Computer Science 150 (2019) 757–763, https://doi.org/
10.1016/j.procs.2019.02.001.
[6] J.S. Chou, J.P.P. Thedja, Metaheuristic optimization within machine learning-based
Fig. 15. Building facade image processing. classification system for early warnings related to geotechnical problems, Autom.
Constr. 68 (2016) 65–80, https://doi.org/10.1016/j.autcon.2016.03.015.
[7] Z. Ding, S. Liu, L. Liao, L. Zhang, A digital construction framework integrating
Table 2 building information modeling and reverse engineering technologies for renovation
projects, Autom. Constr. 102 (2019) 45–58, https://doi.org/10.1016/j.autcon.
Knowledge base of window types. 2019.02.012.
Windows standard size: Window type [8] O. Doukari, B. Naudet, R. Teulier, Merging IFC-based BIM models: a new paradigm
and co-design support tool, Int. J. 3-D Inf. Model. 6 (1) (2017), https://doi.org/10.
height × width (metre)
4018/IJ3DIM.2017010104.
[9] O. Doukari, W. Würbel, R. Jeansoulin, A new model for belief representation and
0.75 × 0.60 Simple window:1 door wing
belief revision based on inconsistencies locality, ICTAI 2007 (2) (2007) 262–269,
1.15/1.25/1.35 × 1 Simple window: 2 door wings https://doi.org/10.1109/ICTAI.2007.37.
1.25/1.35 × 1.20 Simple window: 2 door wings [10] L.M. González, R.E. Giachetti, G. Ramirez, Knowledge management-centric help
2.15 × 0.60/0.80 French window: 1 door wing desk: specification and performance evaluation, Decis. Support. Syst. 40 (2) (2005)
2.15 × 1/1.20 French window: 2 door wings 389–405, https://doi.org/10.1016/j.dss.2004.04.013.
2.15 × 1.80/2.10/2.40 Sliding bay: 2 door wings [11] F. Hayes-Roth, The knowledge-based expert system: a tutorial, Computer 9 (1984)
11–28 https://doi.ieeecomputersociety.org/10.1109/MC.1984.1659242.
[12] D.E. Heckerman, E.J. Horvitz, B.N. Nathwani, Toward normative expert systems:
part i the pathfinder project, Methods Inf. Med. 31 (2) (1992) 90–105, https://doi.
Table 3 org/10.1055/s-0038-1634867.
Knowledge base of maximum storey height of building types. [13] ISO, ISO 29481-1: 2016 Building Information Models - Information Delivery
Manual - Part 1: Methodology and Format, International Organization for
French building level height (metre) Building type Standardization, Geneva, Switzerland, 2016 Web page at https://www.iso.org/
standard/60553.html , Accessed date: 23 December 2019.
2.66 Parking; Hotel [14] Y. Ji, A. Borrmann, J. Beetz, M. Obergrießer, Exchange of parametric bridge models
3 Older residential building using a neutral data format, J. Comput. Civ. Eng. 27 (6) (2013) 593–606, https://
3.3 Office building; Hospital doi.org/10.1061/(ASCE)CP.1943-5487.0000286.
4 Station; Exhibition hall; Superstore [15] M. Kassem, B. Succar, Macro BIM adoption: comparative market analysis, Autom.
> 4 Cathedral Constr. 81 (2017) 286–299, https://doi.org/10.1016/j.autcon.2017.04.005.
[16] H. Kim, K. Anderson, S. Lee, J. Hildreth, Generating construction schedules through
automatic data extraction using open BIM (building information modeling) tech-
nology, Autom. Constr. 35 (0) (2013) 285–295, https://doi.org/10.1016/j.autcon.
Declaration of competing interest 2013.05.020 (2013).
[17] H. Kim, Z. Shen, I. Kim, K. Kim, A. Stumpf, J. Yu, BIM IFC information mapping to
The authors declare that they have no known competing financial building energy analysis (BEA) model with manually extended material informa-
tion, Autom. Constr. 68 (2016) (2016) 183–193, https://doi.org/10.1016/j.autcon.
interests or personal relationships that could have appeared to influ- 2016.04.002.
ence the work reported in this paper. [18] E. Laptali, N.M. Bouchlaghem, Expert systems within the construction industry in
the UK, Automation in Construction, 3(4) Elsevier, Amsterdam, 1995, pp. 321–325,
, https://doi.org/10.1016/0926-5805(94)00037-N 1995.
Acknowledgements [19] A.H. Lee, W.C. Chen, C.J. Chang, A fuzzy AHP and BSC approach for evaluating
performance of IT department in the manufacturing industry in Taiwan, Expert
This is a substantially extended and enhanced version of the paper Syst. Appl. 34 (1) (2008) 96–107, https://doi.org/10.1016/j.eswa.2006.08.022.
[20] G. Lee, R. Sacks, C.M. Eastman, Specifying parametric building object behavior
presented at the CIB W78 Annual Conference held at Northumbria (BOB) for a building information modeling system, Automation in Construction,
University in Newcastle UK in September 2019. We would like to

9
O. Doukari and D. Greenwood Automation in Construction 113 (2020) 103129

15(6) Elsevier, Amsterdam, 2006, pp. 758–776, , https://doi.org/10.1016/j.autcon. urban building structures based on discriminant analysis techniques, Trans. GIS 12
2005.09.009 2006. (1) (2008) 31–59, https://doi.org/10.1111/j.1467-9671.2008.01085.x.
[21] Y.-C. Lee, C.M. Eastman, W. Solihin, R. See, Modularized rule-based validation of a [31] N. Stuurstraat, F. Tolman, A product modeling approach to building knowledge
BIM model pertaining to model views, Automation in Construction, 63 2016, pp. integration, Autom. Constr. 8 (3) (1999) 269–275, https://doi.org/10.1016/S0926-
1–11, , https://doi.org/10.1016/j.autcon.2015.11.006 March 2016. 5805(98)00076-4 1999.
[22] S. Lockley, C. Benghi, M. Cerny, Xbim. Essentials: a library for interoperable [32] A. Sundaram, An introduction to intrusion detection, Crossroads 2 (4) (1996) 3–7
building information applications, The Journal of Open Source Software 2 (20) Web page at http://neuro.bstu.by/ai/To-dom/My_research/Paper-0-again/For-
(2017) 473, https://doi.org/10.21105/joss.00473. research/D-mining/Anomaly-D/Intrusion-detection/Intrusion-Detection-Intro.pdf ,
[23] P.E.D. Love, J. Matthews, The ‘how’ of benefits management for digital technology: Accessed date: 23 December 2019.
from engineering to asset management, Autom. Constr. 107 (2019) 102930, , [33] A. Vinay, A. Gupta, A. Bharadwaj, A. Srinivasan, K.B. Murthy, S. Natarajan, Deep
https://doi.org/10.1016/j.autcon.2019.102930. learning on binary patterns for face recognition, Procedia Computer Science 132
[24] N.F. Matsatsinis, M. Doumpos, C. Zopounidis, Knowledge acquisition and re- (2018) 76–83, https://doi.org/10.1016/j.procs.2018.05.164.
presentation for expert systems in the field of financial analysis, Expert Syst. Appl. [34] R. Volk, J. Stengel, F. Schultmann, Building information modeling (BIM) for ex-
12 (2) (1997) 247–262, https://doi.org/10.1016/S0957-4174(96)00098-X. isting buildings—literature review and future needs, Autom. Constr. 38 (2014)
[25] V. Mayya, R.M. Pai, M.M. Pai, Automatic facial expression recognition using DCNN, 109–127, https://doi.org/10.1016/j.autcon.2013.10.023.
Procedia Computer Science 93 (2016) 453–461, https://doi.org/10.1016/j.procs. [35] R. Volk, T.H. Luu, J.S. Mueller-Roemer, N. Sevilmis, F. Schultmann, Deconstruction
2016.07.233. project planning of existing buildings based on automated acquisition and re-
[26] I. Motawa, A. Almarshad, A knowledge-based BIM system for building main- construction of building information, Autom. Constr. 91 (2018) 226–245, https://
tenance, Autom. Constr. 29 (2013) 173–182, https://doi.org/10.1016/j.autcon. doi.org/10.1016/j.autcon.2018.03.017.
2012.09.008. [36] S.T. Waters, Answerman, the expert information specialist: an expert system for
[27] N.F. Noy, C.D. Hafner, The state of the art in ontology design: a survey and com- retrieval of information from library reference books, Inf. Technol. Libr. 5 (3)
parative review, AI Mag. 18 (3) (1997) 53, https://doi.org/10.1609/aimag.v18i3. (1986) 204–212 Web page at http://www.learntechlib.org/p/138219/ , Accessed
1306. date: 23 December 2019.
[28] Pardis Pishdad-Bozorgi, Xinghua Gao, Charles Eastman, Alonzo Patrick Self, [37] R. Zeibak-Shini, R. Sacks, L. Ma, S. Filin, Towards generation of as-damaged BIM
Planning and developing facility management-enabled building information model models using laser-scanning and as-built BIM: first estimate of as-damaged locations
(FM-enabled BIM), Autom. Constr 87 (2018) 22–38, https://doi.org/10.1016/j. of reinforced concrete frame members in masonry infill structures, Adv. Eng.
autcon.2017.12.004 March 2018. Inform. 30 (3) (2016) 312–326, https://doi.org/10.1016/j.aei.2016.04.001.
[29] S.A. Prieto, B. Quintana, A. Adán, A.S. Vázquez, As-is building-structure re- [38] J.P. Zhang, L.H. Liu, R.J. Coble, Hybrid intelligence utilization for construction site
construction from a probabilistic next best scan approach, Robotics and layout, Autom. Constr. 11 (5) (2002) 511–519, https://doi.org/10.1016/S0926-
Autonomous Systems, 94 2017, pp. 186–207, , https://doi.org/10.1016/j.robot. 5805(01)00071-1.
2017.04.016 August 2017. [39] X. Zhao, A scientometric review of global BIM research: analysis and visualization,
[30] S. Steiniger, T. Lange, D. Burghardt, R. Weibel, An approach for the classification of Autom. Constr. 80 (2017) 37–47, https://doi.org/10.1016/j.autcon.2017.04.002.

10