Automated Code Compliance Checking Based on

a Visual Language and Building Information Modeling

Cornelius Preidela and André Borrmanna
a
Chair of Computational Modeling and Simulation, Technische Universität München, Germany
E-mail: [email protected], [email protected]

ABSTRACT structural stability, reliability, quality of material and not

at least the safety of the user. Therefore the compliance
One of the most important issues during the check of the design planning concerning the applicable
planning of a construction project is to maintain the rules and regulations represents an essential process
quality of the design planning constantly at a high during the execution of a construction project. Due to the
level. Therefore this quality must be checked variety of disciplines and subject areas in the building
continuously in terms of accuracy and compliance to industry a large amount of codes and regulations have to
the applicable codes and guidelines throughout the be taken into account by the planning consultant, from
duration of a project. Nowadays this checking process who a high level of expertise, experience and care is
is laborious, cumbersome and error-prone since it is demanded accordingly.
mostly performed manually based on two- Case examples and requirements in guidelines, as
dimensional planning and iteratively at each planning shown in Figure 1, can be presented in many different
change by the responsible planning consultant. ways, ranging from simple and clearly structured tables
Recently, various approaches attempted to automate with limiting values over graphical representations to
this highly relevant process with the help of digital flow-text written descriptions.
methods, such as Building Information Modeling, in
order to reduce the amount of work and increase the
quality of the planning at the same time. Although this
Automated Code Compliance Checking has been
implemented using a variety of different methods,
most of the existing approaches fail because they
represent the information of rules in an insufficient or
overly complex manner.
In this paper a short analysis of the pros and cons
of selected existing approaches is given and
subsequently minimal requirements for a successful “The access route for pedestrians / wheelchair users
automation of this process are defined. To counteract shall not be steeper than 1:20. For distances of less
the lacks and insufficiencies of existing approaches, a than 3 metres, it may be steeper, but no more than
new method is introduced which enables an 1:12. The access route shall have clear width of a
automation using a flow-based, visual programming minimum of 1,8 m and obstacles shall be placed so
language, which we call Visual Code Checking that they do not reduce that width.”
Language (VCCL). Finally the practical imple-
mentation of a semi-automated compliance check Figure 1. Top: Illustration of a spatial case of fire
concerning an exemplary German fire code according to the German standard DIN 18232-
demonstrates the viability of the approach. 11:2007 [1] Bottom: Excerpt of the Norwegian
Keywords – accessibility guideline, NS 11001-1.E:2009 [2]
Automated Code Compliance Checking; Building
Information Modeling; Visual Language; ISARC Nowadays the checking process is performed to a
2015 large extent manually based on two-dimensional
technical drawings and textual documents by the
1 Introduction responsible planning consultant as well as the building
permission authorities. Due to the low level of
Standards and guidelines in construction industry are automation the common checking procedure is time-
used for standardization of requirements and secure the consuming, tedious and error-prone. This is particularly
technological standards in order to guarantee the evident when unwanted iteration cycles become
necessary due to modifications demanded by the Real Estate Network” (CORENET) with the intention to
respective authorities or errors in the construction optimize the collaboration and interaction between all the
processing. As a result, checking the code compliance of participants of a building project with special emphasis
the building design can be a major cause of delays and on incorporating the responsible authorities [7].
cost increases in construction planning. Recently Accordingly, the quality control of the design,
significant delays of several major projects in Germany including code compliance checking, is of particular
have shown the impacts of a wrong execution of importance for CORENET and was introduced as a
checking processes. Therefore the public and political separate module called “CORENET e-Plan Check” in
interest has grown and have contributed to an increased 2002. The provided checking functionalities focus on the
demand for an optimization of construction processes national applicable codes in the areas of building control,
with the help of modern digital tools [3]. barrier-free access und fire safety [8].
Due to the continuing development of new The main component of the code checking system is
technologies in the recent years, the construction industry the FORNAX library which has been developed and
is undergoing a fundamental transformation, which was maintained by a private company [9]. As a consequence
initiated in particular by the methodology of Building the correctness of the implementation is not verifiable
Information Modeling (BIM). By means of new digital since the hard-coded checking routines are not
methods and the rapid crosslinking of increasingly transparent for the user. Therefore this checking
powerful computers new practices and research areas functionality is called a black-box method [10,11].
arise offering a range of new approaches to make Nevertheless CORENET represents one of the most
building processes more efficient [4,5]. As a result the comprehensive approaches in the area of Automated
construction industry has gained the necessary resources Code Compliance Checking since it covers a large part
to automate and thereby optimize the checking process in of the Singaporean guidelines and is used in over 2500
terms of effort, time and cost. During the BIM planning companies of the AEC sector [9].
process all information is stored in a central digital At this point it should also be mentioned that the
building model, which in turn provides all the current introduction of CORENET was heavily promoted by the
information for all project participants throughout the Singaporean government and accompanied by
entire life cycle of the building. It is recommendable to appropriate legislation. As a consequence construction
use these already bundled data for a full- or semi- companies in Singapore were brought to the use of
automatic review of a model for compliance with CORENET as they may otherwise not receive a building
standards, so that finally the overall process achieves a permit for the construction projects. Taking the special
higher level of efficiency. political, economic and demographic structure of this
country into account, it remains doubtful that such an
2 State of the Art approach would prevail in the same way in a European
country.
In the recent years various efforts were undertaken in
order to develop a method for Automated Code 2.2 Solibri Model Checker (SMC)
Compliance Checking. In this paper some important
representatives of such methods are presented, and In the year 2000, the Finnish software company
subsequently the challenges and difficulties are discussed. Solibri introduced the Java-based Solibri Model Checker,
In Figure 2 chronological sequence of the treated which was intended to be a validation and optimization
approaches is shown. tool for digital building models stored in the Industry
Foundation Classes (IFC) data format. In the Rule
Solibri Model Checker
Manager component, the SMC provides a library of rules
CORENET BP-Expert CORENET ePlanCheck
and guidelines, from which the user can select and build
Fornax up an individual review process according to his
BERA
requirements [9].
Next to basic rules, which check the quality of the
1995 2000 2005 2010 2015 imported IFC model, the manager provides mainly
geometry-oriented rules e.g. in the field of space
Figure 2. Representative approaches for management and accessibility. The rule sets within the
Automated Code Compliance Checking in the SMC are implemented as hard-coded functions, which
recent years (inspired by [6]) access the information of the data model via a native
programming interface. Since this interface is not
2.1 CORENET & FORNAX available to the public, also the SMC implements a black-
box method, which makes no information of the process
In 1995, the Singaporean building construction
visible for the user. An external development of new or
authority (BCA) started the platform “Construction and
custom rule sets is only possible in cooperation with the Building Model
company Solibri. Preparation

2.3 Building Environment Rule and Analysis Report

Language (BERA) Rule
Rule Interpretation Checking
Execution
Results
Next to the to many directly implemented black box
methods, there are approaches that introduce a language-
Figure 3. Common structure of an Automated
based Code Compliance Checking. A significant
Code Compliance Checking process (inspired
representative is BERA, a domain-specific programming
by [9])
language, which has been developed not only for
querying but also for the formulation of checking
In a next step the digital rules are read from an
processes for digital building models [12]. Since this is a
executing instance, interpreted and processed on base of
language-based approach, which grants much more
information, which is provided by the digital building
manoeuvrability for users, it distinguishes itself clearly
model. Various investigations have shown that this step
from the methods discussed above. The transparent
can be a major error source due to inconsistencies,
description of procedures and the direct influence of the
contradictory, false or non-existent information in the
user result in a higher potential to encode more complex
building model [14]. A direct use is therefore not
rules. As a proof of concept, BERA has already been
recommended, but to prepare the required information
applied for first evaluations of buildings circulation and
separately and in a pre-processing step.
spatial programmes. To this end the language provides a
In a last step, the results of the review are finally
set of spatial operators for the definition of rules in the
processed and retained for the user. In the more modern
context of these application areas [9].
approaches of an Automated Code Compliance Checking
In summary BERA shows the high potential of
this process step is designed primarily graphically.
language-based methods and that this can be an important
point of departure. Nevertheless this approach lacks a
3.2 Major Challenges
generality in the logic base, which is necessary to achieve
a higher versatility especially to define more complex As shown in Section 2, there are a lot of different
structures. approaches for an automation of the compliance
checking. Nevertheless several factors can be found,
3 Challenges of an Automated Code which relativize this already reached degree of
Compliance Checking automation:
Most of the existing approaches lack because of the
3.1 Common Structure insufficient transparency and visibility of the processing
steps for the user. Many methods focus too much on the
In order to demonstrate the challenges of the technical automation of the checking process and do not consider
implementation of an Automated Code Compliance the incorporation of the user and therefore the practical
Checking the basic structure of the overall process has to applicability. The correctness and accuracy of a
be identified first. Although each of the presented compliance checking is the responsibility of the reviser
approaches is characterized by its individual features, a and cannot be transferred to a machine because of legal
common structure of the compliance checking process reasons. Therefore in building practice it is common to
can be found. Eastman [9] defines the overall process as manually verify and validate the results with
a flow and interaction of four single process steps, shown simultaneous or trailing plausibility checks, e.g. by rough
in Figure 3. calculations by hand or comparison with empirical values.
The main requirement and first step of any Such a validation based on the experience of the reviser
compliance review is the translation of rules into a is not possible if the transparency of each single process
machine-interpretable language. The idea of the step is not given. Because of its hardcoded machine-rules
digitization of language in oral or written form exists the SMC represents such a black-box method and so it
since the early days of computer science and is still a fails in this point [8]. As a result an automated checking
highly relevant topic in the various application areas. must be a dynamic and semi-automated process that
Basically it is about to translate the content of spoken or moves the user into the focus and incorporates him into
written word as precisely as possible into binary code the process. Although the reviser has usually no
[13]. Since guidelines and standards describe the programming skills, the human-readability of the
contented information usually in many different ways, it translated codes must be maintained. Such a method is
is a major challenge to standardize this process. called white box and is schematically shown in Figure 4.
 Black-Box languages. In recent years visual languages, which are
also known as Visual Programming Languages (VPL),
Input Output
have been established particularly in the field of control
and modification tools of digital information systems.
Known software products in the context of building
White-Box
design are in particular the plug-in Grasshopper for
Rhinoceros3D [15] or Dynamo for Autodesk Revit and
Input Output Autodesk Vasari [16].
By adapting such a visual language to the specific
needs of the Code Compliance Checking, the presented
insufficiencies discussed in Section 3 can be overcome.
Figure 4. Schematic representation of a black-box The approach focusses in particular on the human-
and a white-box method (inspired by [11]) machine-communication, which represents a previously
defined, mandatory requirement for the success of an
Furthermore, many of the approaches focus on automation of the Code Compliance Checking. At any
comparatively simple and straightforward rules. A time and degree of completion of the visual processing
transfer of this method to a higher level of complexity system, the user is able to understand and inspect every
will be probably difficult, since the hard-coded and fixed single processing step, what is particularly important,
implementation of rules cause a rigidity, which inhibit since the reviewer is responsible for the accuracy and
the formulation of complex rules. As a representative for correctness of all compliance check results. If errors are
language-based approaches, BERA has shown that the
identified in the processing chain, the system can be
problem can be solved by creating a larger space of action adjusted very quickly and simply according to the user’s
for the translation step. According to the modular requirements. With the help of such a visual language, it
principle, complex structures can be build up by is possible to describe any compliance check without
composing simple elements with a low degree of sacrificing the transparency for the user.
complexity, which are already well-defined and can be
reused by the user. For this purpose it is necessary that
the individual elements of the language are defined
within a fixed logic frame, in order to implement a formal Finest
rigidness at the base. Genericity Granularity
VCCL
4 Visual Code Checking Language
The approaches in Section 2 have shown, that there
Maximal
are a lot of ways to automate the compliance checking,
Flexibility
but there are still a lot of inadequacies. To overcome
these insufficiencies we introduce a new approach, which
uses a visual language for representing the Code
Compliance Checking process. Figure 5. Schematic illustration of the
principles of the VCCL
4.1 Methodological Basis
As shown in Figure 5 the VCCL follows three
In general a visual language can be defined as a principles: genericity, finest granularity and maximal
“formal language with a visual syntax and visual flexibility. The genericity describes the property of the
semantics”, which means that it represents a modular VCCL that all elements must be defined as generic as
system of signs and rules using visual elements instead possible regardless of the level of complexity. As a result
of textual ones on the semantic and syntactic level [13]. each elements can be used in any situation and on any
Information systems, which are described by a visual point of a desired structure. At the same time it must be
language can be interpreted much faster and easier by possible to break down each element to its lowest level.
humans. Visual languages are often also called flow- This property of the VCCL is called finest granularity.
based, since they display the complicated structures as a These two features cause a maximum of flexibility for the
flow of information. The reason for the higher user, who can formulate the desired content. Furthermore,
interpretation capability can be found in cognitive a visual language is ideal for building a library of simple
psychology, which states that visual information can be base elements, as it was introduced in Section 3.2. A
processed with two instead of only one hemisphere of the schematic structure of such a library for the VCCL is
human brain in parallel. Schiffer [13] performs a detailed shown in Figure 6.
discussion of the advantages and disadvantages of visual
 VCCL Node Library Walls :
Room X : Room
set<Wall>
Operators Objects
0 Basic Ops Data objects Relation*
Other
O
Relation Op* e.g. Table
Area : Float Wall 1
Iteration* set
set
Wall 2 <Wall>
Height : Float
Base Graph Base Graph Base Graph Wall n
1

degree of complexity
Figure 8. Illustration of an attribute and set node
Complex Complex
2
Graph Graph
A special form of an object node is a set of objects. In
this node multiple objects of the same type can be stored.
Complex Principally it represents thus a special data type for an
n
Graph
object node. An example of a set object is shown in
Figure 8.
VCCL Graph Library *planned

label description
Figure 6. Schematic illustration of the VCCL node
library and its resultant VCCL graph library with Operator Node
its ascending degrees of complexity

The principle behind it is to make the overall process

of compliance checking visible, by structuring this Figure 9. Illustration of an operator node on
process in a compilation of composite process steps. VCCL
Each of these elements is a single white box, which can
be considered as a small module of the whole process. For the description of processes between object nodes
Therefore we introduce a modular principle, which can the VCCL uses the operator node (see Figure 9). Such a
be used by any user even without profound programming node describes a well-defined operation on a specified
skills. In this way we allow that any engineer can bring number of input variables - the operands - and generates
his professional skills and his experience into the process. a corresponding result.

4.2 Elements of the VCCL

To define a new language, both base aspects -
semantics and syntax - need to be defined. In the
following, the elements of each level and their graphical
representation are presented for the VCCL.
Input-Port Output-Port

Object Node : datatype Figure 10. Illustration of a generic node and its
interfaces (called ports) and directed edges
label description label datatype
The elements of the syntactic level are represented as
edges and interfaces. A directed edge links two VCCL-
Figure 7. Illustration of an object node on VCCL
nodes and thus builds up a processing chain. By defining
the unambiguously direction of the edge, the
All elements of the semantic level are represented as
representation of the information-flow is specified (see
nodes. The object node, as shown in Figure 7, describes
Figure 10). To describe the connection between VCCL-
an object that can be individually and clearly identified
nodes precisely, each semantic object has a certain
in a real-world system. Therefore it must be of a
number of interfaces (called ports) defining which
particular, unambiguously data type. For a better
information can be passed. In this way, the transmission
interpretation, attributes of objects can also be visualized
of information across the processing chain is given a
as single elements (see Figure 8). The visualization of
fixed frame and inconsistencies can be prevented. A port
attributes is not obligatory and is only intended to let the
on the left side of a VCCL element is responsible for the
user capture a complex flow of information.
incoming information and therefore is called Input-Port.
Accordingly, ports on the right side are responsible for criteria “Wall”, which means that the operator extracts all
outgoing information and therefore called Output-Port. A building elements of type “Wall” from the building
schematic illustration of the different types of ports is model. This is a common procedure, which can be
shown in Figure 10. applied for any attribute which is stored in the respective
object. The result is finally transferred to the resulting set
4.3 Application of the VCCL node.
With these elements, it is possible to build up a VCCL
graph, which describes a certain checking procedure. At
the same time the process remains transparent and each
process step visible as a single element of the information
system.

Building Model :
Model
Wall : Model
String

String Figure 13. Excerpt of the data table for the

Get required smoke ventilation area in m² [1]
set
Access
<Wall>
Since this is a very simple example, the application
Wall 1
set<Wall> will now be shown for a more sophisticated case. In order
Set : to demonstrate the potential and versatility of the
Wall 2
set<Wall>
approach, a semi-automated compliance check regarding
Wall n an applicable German standard is shown. A central
regulation of the DIN 18232-2:2007-11, a German
Figure 11. VCCL graph describing the access to a standard for the design of buildings in terms of smoke
certain data member and fire protection represents a spatial case of fire (see
Figure 1). Depending on the height of the room, the
An example of the VCCL for accessing wall height of the smoke layer and the fire classification, the
components of a building model is illustrated in the guideline requires a minimal smoke ventilation area that
following example, shown in Figure 11. In this example, is listed in a data table, which is shown in Figure 13. The
both initial object nodes hold certain information of a translation of this central regulation is shown in Figure
defined data type and transfer it via the ports and the 12 as a VCCL-graph. In this processing graph a single
edges to the operator node. Inside of this operator the room is identified and afterwards its attributes are used to
information of both elements is processed according the capture both values “Target area” and “Actual area”. The
instructions that were assigned to the operator node. In final result of this check is the comparison of these values,
this case, the model instance is accessed for the filter to check whether the limit value is met or not.

Room X : Building Model :

String Model (*)
Area : Automatic sum up of all
Model
String String Float-attributes of the
Opening : elements, which are stored
Access
String String in the object node
Area : Float Float(*)
set set
Get (**)
Room String
<Opening> <opening> Special node, which can
Height : Float
Openings : access a datatable and
set<Opening> Room X : Room Get set<Opening> search for limiting
Room Actual area:
Opening X Float boundary values according
Area : Float to certain criteria
Room
Opening Y Float
Float Necessary smoke
Get Target area : Bool
Height : ventilation area
Float Float Float Comparison Result : Bool
String String (**) Float

Height smoke layer :

Float
Fire classification : String
String

Figure 12. VCCL graph describing the central regulation of the DIN 18232-2:2007-11
4.4 Proof of Concept and the graph can be re-processed immediately.
Furthermore the CodeBuilder focuses particular on the
In order to validate the concept of the VCCL a incorporation of the user in the checking procedure. Most
practical implementation of this language-based method of the nodes are able to display intermediate results of the
was carried out and is presented in this section. processing procedure. As an example, relevant building
elements, which fulfil or fail a certain check can be
highlighted directly in the geometry view as shown in
Figure 15. In this way, the user is able to check, if the
processed result meets his expectations and requirements.
The developed tool was examined for its practical
applicability. To this end the central regulations of DIN
18232-2 discussed above were translated into VCCL and
successful semi-automated. Exemplarily, the user
interface as well as a result of a geometric checking,
which stated an intermediate result of the VCCL
processing, is shown in Figure 15.

5 Conclusion and Outlook

Figure 14. User interface of the bim+ viewer [17] The large number of existing approaches have shown
the extraordinary relevance and importance of an
The application of the VCCL was developed and Automated Code Compliance Checking for the
designed in close cooperation with the German software construction industry. The introduction of VCCL has
vendor Nemetschek [17]. As basis for the VCCL serves demonstrated that it is possible to automate the
bim+, which is a standardized central platform for compliance checking of a building information model
building information exchange. Next to a large number using a visual language and BIM. At the same time
of basic functionalities, such as a web project various requirements that were not met adequately in
management and a web viewer (see previous research approaches can be fulfilled. By
Figure 14), the platform provides an open REST-API, implementing the VCCL within the CodeBuilder plugin,
which enables developers to use the platform for their we proved the practical viability of the approach.
own purposes, such as the building information handling In building practice, there is a variety of codes and
by the VCCL. many different ways of presenting information.
We developed the CodeBuilder plugin, which allows Therefore it is necessary to develop more VCCL
the user to build up a VCCL graph using a library of elements, which are able to represent this information
elementary nodes. Since bim+ is used as a database, within a node and in a VCCL graph. Based on an analysis
which allows a fast loading and switching of different of other standards, these representations can be identified
models, especially the principle of generality of a VCCL- and serve as a basis for further development. In this way,
graph comes into effect. Each graph, which was built up a library of VCCL elements progressively grows, which
by the user, is valid for any building model stored on the captures more different applications.
platform. Therefore the model can be changed on the fly The introduction of a visual, flow-based language for

Figure 15. Left: User interface of the CodeBuilder plugin in bim+ [17] Right: Visualized result of a check:
Automatic identification, which building elements contain an opening and belong to a certain Room
Automated Code Compliance Checking creates the basis [7] T.A. Lin, Building Smart - A Strategy for
of a new genre for the automation of processes in Implementing BIM Solution in Singapore,
construction industry. This approach opens up a variety Synthesis Journal 2006. 5 (1995) 117–124.
of opportunities for other developments in the field of
construction. An example is the bidding process for a [8] R.X.R. Xu, W. Solihin, Z.H.Z. Huang, Code
construction project, which is subject to individual Checking and Visualization of an Architecture
company rules and specific guidelines for the quantity Design, IEEE Visualization 2004. (2004) 10p–
take-off. 10p.
In summary it can be stated that the introduction of
the VCCL represents a step towards the automation of [9] C. Eastman, J. Lee, Y. Jeong, J.-K. Lee,
many processes in construction industry and can serve as
Automatic rule-based checking of building
a base for several following approaches.
designs, Automation in Construction. 18 (2009)
1011–1033. doi:10.1016/j.autcon.2009.07.002.
Acknowledgements
The authors gratefully acknowledge the support by [10] N. Nisbet, J. Wix, D. Conover, The future of
Nemetschek Allplan GmbH for the presented research. virtual construction and regulation checking, in:
P. Brandon, T. Kocatürk (Eds.), Virtual Futures
References for Design, Construction and Procurement,
Blackwell Publishing Ltd, 2008: pp. 241–250.
[1] DIN 18232:2007-11 - Smoke and heat control
systems - Part 2: Natural smoke and heat [11] L. Von Bertalanffy, The History and Status of
exhaust ventilators; design, requirements and General Systems Theory., Academy of
installation, 2007. Management Journal. 15 (1972) 407–426.

[2] NS 11001-1, issue date: 2009, Universal design [12] J.K. Lee, Building environment rule and
of building works - Part 1: Buildings open to the analysis (BERA) language and its application
public, 2009. for evaluating building circulation and spatial
program, Georgia Institute of Technology,
2011.
[3] K. Kammholz, So werden Baudesaster wie der
BER künftig vermieden, Die Welt. (2014).
http://www.welt.de/politik/deutschland/article12 [13] S. Schiffer, Visuelle Programmierung -
7986675/So-werden-Baudesaster-wie-der-BER- Grundlagen, Potentiale und Grenzen, Addison-
kuenftig-vermieden.html (accessed January 13, Wesley, 1998.
2015).
[14] J. Beetz, B. de Vries, J. van Leeuwen, IfcOWL:
[4] S. Mihindu, Y. Arayici, Digital construction A case of transforming EXPRESS schemas into
through BIM systems will drive the Re- ontologies, Artificial Intelligence for
engineering of construction business practices, Engineering Design, Analysis and
Proceedings - International Conference Manufacturing. 23 (2009) 89–101.
Visualisation, VIS 2008, Visualisation in Built doi:10.1017/S0890060409000122.
and Rural Environments. (2008) 29–34.
[15] Robert McNeel & Associates, Rhinoceros3D,
[5] N.W. Young Jr., S. Jones, H. Bernstein, The (2014). http://www.rhino3d.com/ (accessed
business value of BIM: Getting building April 29, 2015).
information modeling to the bottom line,
Bedford, MA: McGraw-Hill Construction. [16] Autodesk Inc., Autodesk Homepage, (2015).
(2009) 51. http://www.autodesk.com/ (accessed April 29,
2015).
[6] J. Dimyadi, R. Amor, Automated Building Code
Compliance Checking – Where is it at ?, in: [17] Nemetschek Allplan Deutschland GmbH, bim+,
Proceedings of CIB WBC 2013, Brisbane, (2015). http://www.bimplus.net/ (accessed
Australia, 2013: pp. 172–185. January 13, 2015).