Professional Documents
Culture Documents
net/publication/318826829
CITATIONS READS
70 2,770
3 authors:
Markus König
Ruhr-Universität Bochum
310 PUBLICATIONS 4,524 CITATIONS
SEE PROFILE
All content following this page was uploaded by Christian Koch on 18 December 2017.
a
6 Associate Professor, Department of Civil Engineering, University of Nottingham, University
b
9 Research Associate, Computing in Engineering, Ruhr-Universität Bochum, Universitätstraße
c
12 Professor, Computing in Engineering, Ruhr-Universität Bochum, Universitätstraße 150, 44801
13 Bochum, Germany;
15
*
Corresponding author
16 ABSTRACT
17 A transparent, holistic and detailed design of individual systems and processes in mechanised
18 tunnelling is essential for a robust and low-risk construction of tunnels. In this context, the complex
19 interactions between the ground, the boring machine, the tunnel lining and the built environment
20 play a significant role. Traditionally, the entire tunnel design information is available in the form
21 of independent, dispersed and heterogeneous data files. Since these data sources are barely linked
22 in practice, unilateral decisions are made that do not consider all relevant aspects. Existing research
23 has focused either on very general approaches of multi-model container or linked data models that
24 have not been adapted to tunnelling projects, or on semantic tunnel models that solely cover small
25 parts of the entire projects. In this paper a tunnel information modelling framework, basically
26 integrating four interlinked subdomain models and linked project performance data, is presented.
27 Due to their distinctive impact on the tunnel design and construction process a ground model, a
28 boring machine model, a tunnel lining model, and a built environment model are first individually
29 created, and then linked within an open IFC environment using the concepts of Proxies, Property
30 Sets and Model View Definitions. Based on the proposed framework selected case studies are
31 presented to verify its potential and advantages when (1) interactively visualising time-dependent
33 numerical simulation models to predict settlements. These case studies are conducted using real
35
36 KEYWORDS
39
2
40 1. INTRODUCTION
41 Due to the ongoing expansion of urban areas worldwide, sustainable solutions must be found to
42 face the challenge of increasing mobility in an efficient and environment-friendly manner. One
43 option is the extension of underground transportation systems using tunnels. The stable, economical
44 and sustainable design and construction of tunnels requires reliable knowledge regarding the
45 expected impacts of the construction method on the built environment. In this context, insights into
46 the dynamic interactions between the geological conditions, the existing infrastructure and the
47 tunnelling advancement process are essential. Mechanised tunnelling is an established flexible and
48 economical construction method for underground structures that is characterised by its trend
49 towards larger shield machine diameters, meanwhile up to 19.25 m, and its constantly growing
51 The construction process using Tunnel boring machines (TBM) involves risks that are related to
52 various factors like surface settlements, gap grouting, face stability etc. Since risk is a combination
53 of the aforementioned factors, their interactions have to be studied and simulated by specialised
54 project teams while designing and constructing the tunnel. Consequently, coordinated interactions
55 of machine operations, surveying, logistics and preliminary investigation processes are very
56 important, especially in case of difficult situations such as the removal of obstructions or tunnelling
57 under sensitive structures. In particular, interactions between the soil, the TBM, the tunnel support
58 system, and the above-ground buildings play an important role in achieving successful project
59 completions. Therefore, the team members need to collaborate intensively. As a result, large
60 amounts of data are generated, including data gathered during previous site investigations, data
61 from the design stage, and data obtained from measurements made during the advance of the
62 machine.
63 The project data that is shared among the team members varies in terms of type, scale, format and
64 life cycle phase. While some data is used to describe the structural behaviour of the machine, the
65 tunnel lining and the soil, other data is related the site logistics of the entire tunnelling process (type
3
66 of data). Differences in model scale refer to the resolution of data items, both in terms of space and
67 time. On the one hand, for example, multi-scale infiltration models require data items in a spatial
68 and temporal resolution of micrometres and seconds, respectively, whereas, on the other hand, site
69 logistics models deal with data items of centimetre/metre and hours/days resolution, respectively
70 (scale of data). In addition, the project-related data might be available in different formats and from
72 dimensional partial models (format of data). Finally, it is important to recognise the life cycle phase
73 in which the data is being created, maintained and used to make assumption about the reliability or
74 uncertainty of this data. For example, during the design phase engineers usually work with
75 uncertain soil parameters, whereas during the construction phase they rely on real-time
77 To use and analyse this diverse data in an efficient and practical manner, a consistent and holistic
79 obtain required data automatically, efficiently, and in correct type, scale, and format from the
80 underlying information model. Also, results from numerical driving simulations using structural
81 models can be provided by an integrated information model. A further advantage of such a model
82 is the ability to provide a 4-dimensional visualisation of the relationships between stored data
84 With regard to tunnel information modelling, existing research has either focused on (1) semantic
85 tunnel models that solely cover small parts of the entire project, e.g. only focus on the tunnel lining
86 structure neglecting essential parts such as the ground and/or the boring machine; or on (2) very
87 general multi-model container and linked data approaches that have not yet been adapted to
88 tunnelling projects. To address current shortcomings, this paper presents the development of an
89 integrated tunnel information modelling framework. The overall framework suggests a tunnel
90 information model that consists of several subdomain models, which represent the various data
91 sources of a tunnelling project. To verify the presented approach, a specific tunnel information
4
92 model is presented with data from a real-world reference project of a subway tunnel in Düsseldorf,
93 Germany. Based on this model selected case studies highlight its potential and advantages when
95 context and (2) generating advanced numerical simulation models to predict settlements.
96
97 2. BACKGROUND
99 Tunnelling project documents are usually available in many different types, scales, and formats,
100 originate from loosely coupled and dispersed resources, but are strongly interdependent. This issue
101 is illustrated in Figure 1. The type and format of data differs widely (e.g. CAD drawings, text
102 reports, spreadsheets, diagrams and images (Fig. 1)). This complicates their integration in design,
103 simulation and visualisation models, because required data has to be re-organised from multiple
104 resources, and documents have to be scanned for appropriate information and parameters.
105 Additionally, these parameters usually have to be manually integrated and updated in case of design
106 changes. Thus, in tunnelling practice, numerical simulations, for example, are still not being used
107 to the extent that the possibilities of current simulation models would suggest. This is mainly due
108 to the enormous efforts in the modelling and, in particular, in the collection and integration of all
109 available information on the project in a form that provides automatic model generation (pre-
111
5
112
113 Figure 1: Loosely coupled, unformatted, dispersed, yet interdependent tunnel project information
114 items in practice. The type, scale and format of data varies among CAD drawings, text
116
117 In practice, data exchanges between design calculations and simulations during the planning phase
118 of tunnel projects are often performed manually and, therefore, very rarely. However, various data
119 management systems exist that are used by construction companies, engineering firms and
120 equipment manufacturers to manage data for large tunnel construction projects, usually with the
121 goal of monitoring TBM performance data, monitoring geological data, recording material
122 consumption and deliveries, and controlling project costs (e.g. [4]). The focus of these systems,
123 however, is mainly to efficiently structure the large amount of raw data generated during a
124 tunnelling process and provide a basis for further analysis. Performing project data analyses that
125 cover more than a single project component, e.g. settlement visualisation in the context of machine
127
129 To cope with current limitations in practice, several research efforts have been devoted to
130 information modelling of large construction projects. Research activities highly related to the work
6
131 presented in this paper can be subdivided into infrastructure information modelling, and multi-
133
135 Building Information Modelling (BIM) is an up-to-date modelling concept involving the generation
136 and the management of a three-dimensional (3D) digital representation of physical and functional
137 characteristics of a building or construction facility during its entire life-cycle [5]. Building
138 information models are commonly used as shared data and knowledge resources to support
139 planning, construction, management, utilisation, revitalisation, and demolition activities. Although
140 the BIM concept is currently predominantly applied to building construction projects, in many other
141 construction domains it has been considered suitable to provide a methodology for defining
142 information models and supporting a semantically coherent exchange using standardised exchange
144 The use of standardised exchange formats is particularly helpful during the design phase, when
145 many project participants must work simultaneously on different aspects of the tunnelling project.
146 Information can then be exchanged quickly and uniformly. The visualisation capabilities using BIM
147 also enable complex relationships to be easily identified. IFC is based on an object-oriented data
148 model and is therefore adaptable and easily extensible. Using IFC, objects to be modelled are first
149 organised into spatial regions, such as building floors, before other physical elements are created
150 and linked to the spatial objects. An element basically consists of a visualisation (geometry)
151 component and a set of semantic information attached to this element. IFC was originally developed
152 for the modelling of buildings, but has now been extended to other fields of application in civil
153 engineering, including bridges [7] [8] and roads [9] [10].
154 The modelling of tunnels in shield tunnelling using IFC-based multi-scale product model has been
155 presented in [11] [12] [13] [14]. These product models provide a minimum number of new IFC
7
156 classes required to represent tunnels and their alignment, and also makes it possible to model
157 various geometric spaces, such as work spaces, to define the complete interior of a tunnel in a
158 hierarchical manner. However, the management, analysis and preparation of measured data, such
159 as settlements or machine performance data, or the modelling of the tunnelling machine itself is
160 not provided. Vossebeld and Hartmann [15] have proposed an information model to support safety
161 professionals that is aimed at maintenance governance and road-user safety evaluation. However,
162 tunnel design support and context-aware visualisation of construction risk have not been addressed.
163
164 To model and analyse large-scale infrastructure projects, Geographic Information Systems (GIS)
165 have been used. However, GIS is predominantly aimed at the management of spatial and
166 geographical data, rather than at the modelling of individual structural details as with BIM. GIS
167 uses the Open Geospatial Consortium (OGC) standard GML as a data model [16]. Several
168 modelling approaches based on GML are GeoSciML [17], GroundXML based on LandXML [18]
169 and CityGML [19]. Since 2012, the CityGML [16] has been accepted as an international standard
170 by the OGC. CityGML was developed for the storage and interoperable access of 3-dimensional
171 models of cities and includes geometrical, semantic and topological aspects [19]. Although
172 CityGML is designed to represent cities, Tegtmeier at al. [20] have proposed an extension of
173 CityGML, which provides further classes for surface-based modelling of geological conditions.
174 Techniques to combine IFC and CityGML are presented in [21] and [22]. These approaches mainly
175 deal with the modelling of building structures and the land-use planning of building construction.
176 A methodology for modelling soil and underground structures is presented in [23]. There, a
177 distinction is made between geological features, such as layers of soil or groundwater, and
178 underground structures, such as sewers or tunnels. The model, however, is designed exclusively
179 for the final product and does not deal with construction related equipment (e.g. TBM) and
181
8
182 2.2.2 Multi-model data management and linked data
183 Based on the heterogeneous nature of construction projects, Scherer & Schapke [24] have presented
184 a process-centric, multi-model based and distributed Management Information System with the aim
185 of better utilising design and control models, and supporting decision-making on all different
186 management levels. They use multi-models as they seem to be a promising approach to support
187 information analysis and collaborative work over multiple application domains. According to [24],
188 the fundamental idea of a multi-model is to combine distributed application models, or selected
189 views of them, in a single exchangeable information resource, a so-called multi-model container
190 [25]. Within this container, the application models, e.g. building model, cost model, project
191 schedule model, are bound together by a link model that explicitly specifies the interdependencies
192 among the different application models referencing the respective model elements by their unique
193 identifiers. The idea of linking information that is stored in several domain models is one of the
194 key concepts used and presented in this paper. However, existing multi-model container approaches
195 have only been dealing with general construction projects, thus have never been designed nor
197 To improve the information exchange with various sources outside traditional BIM environments
198 Linked Data and Semantic Web technologies have been applied to Architecture, Engineering and
199 Construction (AEC). Curry et al. [26], for example, have proposed cross-domain AEC data sharing
200 and integration by means of Linked Data as a technology for cloud-based building data services.
201 They have demonstrated their approach in an owner-occupied office building within the context of
202 building energy performance. Corry et al. [27] have used Semantic Web technologies to access soft
203 AEC data across various stakeholders for the purpose of optimising building performance. They
204 conclude that much more research is needed to arrive at a robust building management framework
206
9
207 2.3 Problem statement and objectives
208 With regard to tunnel information modelling, existing research has either focused on (1) semantic
209 tunnel models that solely cover small parts of the entire project, e.g. only focus on the tunnel lining
210 structure neglecting essential parts such as the ground, the boring machine, and/or performance and
211 settlement data or on (2) very general multi-model container approaches that have not yet been
212 adapted to tunnelling projects. To overcome these limitations, the objective of this research is to
213 create and verify a tunnel information modelling framework that basically contains and interlinks
214 all relevant tunnelling project data. Due to their distinctive impact on the tunnel design and
215 construction process the ground, the boring machine, the tunnel lining structure and the built
216 environment need to be integrated and complemented with machine performance and settlement
217 data to provide a basis for several different engineering applications. These applications are
218 designed to support decision-making and risk assessment, for example, settlement data analysis
219 and visualisation as well as numerical simulations of the machine driving process.
220
221 3. METHODOLOGY
223 Existing data management systems in tunnelling are designed to store and manage raw data that
224 accumulates in the course of tunnelling projects. Because this data is mainly given in the form of
225 simple text documents, spreadsheets, diagrams and images, this makes them difficult to interpret
227 comprehensive view of the construction processes in combination with the various measurement
228 data and simulations is usually provided. In a tunnelling project, however, it is of great importance
229 to have a consistent and complete data management strategy throughout the duration of the project.
230 The research projects discussed above often consider only the design phase or only the construction
10
231 phase. Although methods for modelling soil or tunnels already exist, the modelling is often based
232 on separate models that represent only one aspect of the tunnelling process. For the modelling and
233 data management of an entire tunnelling project, the computational and interactive coupling of
235 Based on its success in data modelling for infrastructure projects, the BIM concept has been applied
236 by the authors to develop an integrated information model for tunnelling. To create a uniform access
237 to all relevant data needed for the planning and execution of tunnel projects, relevant project
238 information has been collected, classified, structured and linked into a holistic, object-oriented
239 Tunnel Information Model that is organised as a multi-model container (TIM, see Figure 2a). In
240 contrast to approaches currently found in the context of Building Information Modelling (BIM),
241 this approach also accounts for data that is inherently dynamic and time-dependent. For example,
242 besides capturing data gained from preliminary exploration, which is performed during the entire
243 tunnelling advancement, data from settlement measurements and on-line TBM monitoring are also
244 recorded and processed as relevant project data. The continuous adaptation of soil conditions due
245 to various measurements and simulations represents a major challenge, but is of crucial importance
247 To uniformly access tunnel project data, an interaction platform is proposed (Figure 2a). This
248 platform contains the actual data sources, the multi-model tunnel information container (mentioned
249 above) and an integration layer. The integration layer provides a unified interface to various
250 tunnelling software applications, ranging from data management and visualisation tools up to
11
252
253 Figure 2: Tunnel information modelling framework based on a unified interaction platform and
254 an application layer (a). The tunnel information model contains of four main subdomain models
255 (b).
256
257 3.2 Tunnel information model container
258 All relevant data needed for the planning, construction and maintenance of tunnels is collected,
259 classified, structured and linked into a holistic, object-oriented Tunnel Information Model (Figure
260 2a) using the concept of a multi-model container [24]. This fundamental information model forms
261 the basis for all possible interactions of the project teams and software applications. In particular,
262 four main subdomain models were specified and linked: a ground model, a boring machine model
263 (TBM model), a tunnel lining model and a built environment model (Figure 2b). These models are
264 chosen to represent the tunnelling design and construction process due to their significance. The
265 built environment model influences, for example, the speed of advancement, the possible
12
267
269 The ground model provides a fundamental basis for the detailed analysis of the entire excavation
270 process. Based on boreholes set at strategic places along the proposed alignment, profiles are
271 created, which can then be analysed and evaluated to determine individual soil layers and
272 parameters to obtain a geologist's view of the underground for the proposed tunnel alignment. The
273 soil layers are stored in the form of so-called boundary surface models (Boundary-representation,
274 B-rep) attached with relevant soil properties (attributes) in the ground data model (Figure 2b).
275 During tunnelling, advance exploration in the near field of the TBM is usually carried out
276 continuously and results are carefully evaluated. The resulting information is usually available in
277 high resolution, either as a tensor field or as an attributed point cloud and it is stored as voxel data
278 within the ground model [28]. Consequently, the ground model must support both coarse resolution
280 To support this, methods for integration, consistency checking, transfer and versioning of surface
281 data as well as point data have been developed by the authors to properly manage the hybrid ground
282 data model (Figure 3, [28]). Analytical models as well as simulation models can selectively and
283 efficiently query information concerning the state of the underground based on the developed
284 hybrid ground data model. Furthermore, calculation results of advance exploration simulations that
285 alter existing ground data can be stored in a versioned manner in the hybrid ground data model, so
286 that downstream analyses have access to possibly updated values. The methods developed for
287 updating and querying the ground data include: updating ground properties in a surface model (B-
288 rep), updating ground properties in a point model (voxel), querying ground parameters for a surface
13
289 model (B-rep) and querying ground parameters for a point model (voxel). The authors have
291
293
294 Figure 4 illustrates the information model used to model the hybrid representation of the ground.
295 In this model, the ground (Ground) consists of several soil layers (Layer), each represented by a B-
296 Rep geometry (B-Rep) and assigned to a set of soil parameters (ParameterSet), such as the friction
297 angle and cohesion. On the other hand, the ground is also represented by an octree geometry
298 representation (Octree-Rep), starting with a root note (OctreeNode). Following the octree
299 decomposition scheme, each octree node is recursively subdivided into eight child nodes. Each
300 octree node can have assigned a set of parameters (ParameterSet) that describe the soil condition.
301
14
303
305 In mechanised tunnelling, tunnels are usually constructed using large Tunnel Boring Machines
306 (TBM). As they have a very significant impact on the overall project performance it is essential to
307 model it separately (compare Fig. 2b). However, this paper only focuses on the modelling of Earth
308 Pressure Balance (EPB) shield machines. The TBM model is built upon the IFC standard. It extends
309 the existing IFC structure by adding additional classes for modelling TBM components and
310 differentiates between spatial classes and element classes. While spatial classes define different
311 spatial extents of a TBM, comparable to the IfcBuilding or IfcBuildingStorey classes, the element
312 classes capture specific physical components of a TBM, comparable to the IfcWall or IfcColumn
314 To represent the spatial structure of a soft soil compatible TBM, three classes are added: the
315 IfcTunnelBoringMachine, the IfcTbmHead and the IfcTbmBackupEquipment class (Figure 5). Each
316 class inherits from the IFC class IfcSpatialStructureElement. The IfcTunnelBoringMachine class
317 represents the whole TBM and captures the other two spatial classes IfcTbmHead and
318 IfcTbmBackupEquipment. It contains two attributes, the excavation diameter of the manufacturer
319 of the TBM. These attributes should help the user to get a first impression of the TBM. While the
320 TBM head varies from project to project, the backup equipment usually remains the same except
321 for the supply system of supporting infrastructure elements. This is why the IfcTbmHead class
322 contains an attribute of the type IfcTbmHeadType to define the type of the TBM head. This type
323 points to the enumeration IfcTbmHeadEnum, which includes various types of heads such as
15
(ABS) (ABS) (ABS) (ABS)
IfcRoot IfcObjectDefinition IfcObject IfcProduct
(ABS) (ABS)
IfcElement IfcSpatialElement
A (ABS)
IfcSpatialStructureElement
machine_diameter
length_measure_with_unit
machine_manufacturer IfcTunnelBoringMachine
organization
tbmHeadType
IfcTbmHeadEnum IfcTbmHead
IfcTbmBackupEquipment
325
326 Figure 5: EXPRESS-G diagram of the overall IFC based TBM data model
327
328 The element classes are structured differently compared to the spatial classes. The abstract super
329 class IfcTunnelBoringMachineElement represents the parent of all element classes of the TBM data
330 model. Due to the fact that it inherits from the IfcElement class, each TBM element class can
331 contain a geometric representation of its element. A geometric representation is needed, for
332 example, by planners to identify spatial conflicts during the planning process. Furthermore, the
333 geometric representation gives evidence on the dimension and the position of the specific element.
334 Besides the geometric representation of an element, there is also semantical information attached
335 to each element. This information contains numeric operational values in a key-value style (e.g. the
336 torque of the cutting wheel). They are stored in IfcPropertySets and attached to the element. The
337 IfcPropertySet class is provided by the IFC standard and can contain an arbitrary amount of values
338 of various types (e.g. enumerations, lists or single values). While Figure 6 depicts an extract of the
16
341 IfcTbnScrewConveyorElement, Figure 7 illustrates the physical meaning and the function for some
342 of them.
(ABS)
IfcTunnelBoringMachineElement
IfcTbmCenterPlateElement IfcTbmGearElement
IfcTbmBulkheadElement IfcTbmGearMotorElement
IfcTbmCuttingWheelElement IfcTbmScrewConveyorElement
IfcTbmExcavationChamberElement IfcTbmSegmentErectorElement
IfcTbmManLockElement IfcTbmShieldElement
IfcTbmThrustJackElement
…
343
344 Figure 6: Extract of the EXPRESS-G diagram containing the added element classes within the
346
347
17
349
351 The shield tunnel lining model is also based on the Industry Foundation Classes and has been
352 developed in cooperation with the Technical University of Munich [12] [14]. In analogy with the
353 machine data model, classes are designed to model both spatial regions (IfcSpatialElement) and
354 physical tunnel elements (IfcElement) (Figure 8 and 9). While the spatial elements represent such
355 regions as the tunnel support system (IfcTunnelSpace), the concrete building elements describe
356 physical components such as the ring segments (IfcRingSegment). However, the tunnel model is
357 designed parametrically, in contrast to the machine model. The explicit geometry of the individual
358 tunnel elements will be generated only at program runtime based on the defined alignment.
359 Furthermore, IFC classes have been assigned additional characteristics in order to take into account
360 the temporal aspect of the TBM advancement. This, in turn, supports time-based queries for the
(ABS) (ABS)
IfcElement IfcSpatialElement
(ABS)
B IfcSpatialStructureElement
IfcTunnel
IfcTunnelSpace
IfcCrossway
IfcTunnelPart
362 Figure 8: EXPRESS-G diagram of the overall IFC based tunnel lining data model
18
363
364 While Figure 9 depicts an extract of the EXPRESS-G diagram of the actually modelled element
367 illustrates the physical meaning and the function of IfcRingSegmentElement and
368 IfcGroutingElement.
(ABS)
IfcTunnelElement
IfcRingSegmentElement
IfcRingSegmentReinforcementElement
IfcGroutingElement
IfcTunnelInstallationElement
IfcTemporaryTunnelInstallationElement
IfcTunnelSafetyElement
IfcTunnelWarningUnitElement
IfcTemporaryTunnelSupplyElement
369
370 Figure 9: Extract of the EXPRESS-G diagram containing the added element classes within the
372
19
373
374 Figure 10: Extract of modelled tunnel lining elements and their functions
375
377 The built environment model includes the above-ground and underground structures in the area
378 influenced by the tunnel construction (compare Fig. 2b). With regard to above-ground buildings,
379 particularly their stiffness and mass have a great influence on the size and shape of the settlement
380 trough. Special attention is paid on how to model various levels of detail (LoDs). Therefore, the
381 well-established standard CityGML [19] [16] has been incorporated. The building structures are
382 represented by coarse surface geometries and reduced characteristics (CityGML models with a
383 LoD2 and LoD3) as well as by detailed building models consisting of individual components with
384 detailed properties for each component (IFC models). In this context, urban models in the
385 CityGML format and building models in the IFC format are integrated.
20
386 Figure 11 illustrates a built environment model using CityGML LoD2.
387
388 Figure 11: Example of a built environment model using CityGML with LoD2
389
390 Settlement data is commonly provided in the form of point data, which are time-dependent
391 measurements always assigned to a single point of location. Accordingly, there are a series of
392 measurements over time for a given point, taken from various sources such as terrestrial or radar
393 measurements [29] or even results of numerical simulations [3]. This kind of data remains in the
394 data source layer (see Figure 2a) and is linked on-demand to and referred from the built
395 environment model, as most of the inner city settlement measurements are taken from parts of the
396 built environment, e.g. building foundations, building roofs or road surfaces [29].
398 Within the frame of an integrated visualisation for project navigation, the four subdomain models
399 are linked based on the Industry Foundation Classes. Subdomain models that are not specified using
400 the IFC, namely the ground model and low LoD parts of the built environment model (e.g.
21
401 CityGML LoD2 and LoD3), can be converted into the IFC format using the IFC concept of generic
402 proxies and user-defined properties (IfcProxy, IfcProperty, IfcPropertySets). While proxy elements
403 of type IfcProxy are used to attach geometric representations, properties of type IfcProperty are
404 used to model semantic information in the form of key-value-pairs that are organised in sets of type
405 IfcPropertySet. This provides a unified access to all relevant tunnel project data through IFC.
406 Moreover, this procedure offers the advantage of re-using IFC-based viewing and querying tools,
408 In addition, different model views can be specified to define the subset of information in the multi-
409 model Tunnel Information Model (TIM) container required to perform a specific analysis or to
410 support a specific application, for example a context-aware settlement visualisation and an
411 advanced numerical simulation. For this purpose, the buildingSMART Model View Definition
412 (MVD) concept is applied by defining a dedicated model view for each specific application.
413 According to [31], a model view is similar to a database view that is defined as a specialised and
414 structured subset of data and is stored permanently as kind of a query or filter. Although, the main
415 purpose of a model view is to specify a particular set of information need within a particular
416 information exchange scenario, we use this concept to specify what information is needed and has
22
418 Similar to [31], an example of a conceptual and generic model view is depicted in
419 Figure 12. It specifies what domain model, what model objects and what object attributes are
420 needed to support an application A. Specific model views are presented in the case study section.
Object A ributes
Model Object
Geometry Brep
Brep Geometry Domain Model
Building Layer
Mass Parameters Name
Parameters
Built Mass
Ground
Environment
421
422 Figure 12: Conceptual and generic model view defining a subset of information across all four
424
426 Using the unified IFC-based access and the MVD concept, different applications ranging from data
427 management software, visualisation clients up to numerical simulations can now read the
428 corresponding data from the tunnel information model container and provide analysis results in a
430 As described above, each application has assigned a particular model view, which defines how to
431 access the application-relevant pieces of information stored in the tunnel information model
432 container (cf. Figure 2a). On this basis, different applications can be designed and implemented to
23
434
435 4. IMPLEMENTATION AND CASE STUDIES
437 The “Wehrhahn-Linie” (WHL) is a subway tunnel construction project in the city of Düsseldorf,
438 Germany, and connects the southern district “Bilk” with the district “City Center”. In total a 3.4
439 km long track is built by a shield machine (hydro-shield). The outer diameter D of the tunnel is
440 9.49 m; the average overburden is 1 to 1.5 x D. The tunnel section is constructed with a 45 cm thick
441 reinforced concrete segmental lining. During the approximately 1 km long advance of the east
442 branch (“Ostast”) of the project, comprehensive settlement measurements were carried out. For the
443 two case studies described below, a tunnel section between the subway stations “Schadowstraße”
444 and “Jacobistraße” with a footprint area of 730 m x 340 m is examined (Figure 13). The project
445 period considered contains the shield drive from June to December 2011.
446
447 Using the reports on geotechnical investigations, data of 18 boreholes, laboratory test and
448 groundwater measurements along the tunnel route three main soil layers and the groundwater level
449 are identified, each both in terms of geometry and mechanical properties, and implemented in the
450 WHL ground model. The boring machine model consists of a total of 86 elements, of which 54
451 elements belong the backup equipment part and 32 elements are part of the machine itself. The
452 cutting wheel geometry (diameter of 9.49 m) and the total weight of the implemented machine
454
24
455
456 Figure 13: Wehrhahn-Linie subway tunnelling project: Map of the city of Düsseldorf (adapted
458
459 For each machine element, both relevant geometric and semantic information is stored. In this case,
460 about 200 to 300 values from about 250 data sources are collected for each ring. This means that
461 about 50,000 to 75,000 values need to be recorded for a single ring. Therefore, only average values
462 from measurements and operating conditions (driving, stopping, ring building, etc.) are stored in
463 the WHL machine model. The WHL tunnel lining model consists of a total number of 534 rings
464 with a mean ring width of 1.5 m. The rings are composed of 7+1 segments and the annular gap
465 grout. The outer diameter of a ring is 9.2 m with a thickness of 0.45 m per segment. Each ring is
466 additionally assigned information about the date of installation and the installation time, including
467 the TBM driving time, length of ring construction and any downtime.
468 The built environment model is generated on the basis of a 3D CAD model of the Capital City of
469 Düsseldorf. In this model, individual buildings are initially separated from each other and buildings
470 outside the model range are removed, resulting in a model consisting of approx. 200 buildings. To
25
471 account for the buildings above the tunnel axis, for example in numerical simulations, equivalent
472 replacement (surrogate) models of buildings are provided. These equivalent models contain
473 semantic information such as the effective stiffness of the structures or the masses of the buildings.
474 The relevant information is collected, for example, from an extensive analysis of original records
475 of house construction plans. With regard to settlement data, a total of approximately 6,000
476 settlement data points is available, gained from both terrestrial measurements using classical
477 methods such as tube water levels (550 points) and from satellite measurements (5,500 points).
478 More detailed information on how satellite measurements have been collected can be found in [29].
479
480 To improve collaboration, the presented platform is implemented within a web-based environment.
481 Such environment provides a platform-independent solution, which can easily be shared within a
482 project team to provide a common data environment. For data management and project navigation
483 a client-server application has been developed. A dynamic website provides the client-side
484 graphical user interface and utilizes X3dom [33, 34] for visualizing geometric extent. The
485 integration layer is implemented as application server, which provides information at hand, when
486 models or properties are requested. To parse and interpret information from IFC based models, the
487 server uses the Open IFC toolbox [30] that supports the generation and use of extensions to IFC
488 schema. This is needed to implement the IFC based TBM model and the tunnel lining model (see
489 section 3). Figure 14 depicts an extract of the client application, where the left-hand side depicts
490 the hierarchy of the subdomain models and the right-hand side offers the corresponding
491 visualisation. In addition, by selecting or picking a specific object, related properties can be
493
26
494
495 Figure 14: Screenshot of the web-based management and navigation tool showing the entire
498 Using the integrated tunnel information model of the Wehrhahn-Linie the development of advance-
499 driven settlements can be visualised both in the context of the built environment (static) and the
500 dynamic position and performance of the TBM. This allows, on the one hand, to easily identify
501 potentially endangered buildings in the above-ground built environment that are exposed to large
502 settlements. On the other hand, the TBM’s operational performance data can be visualised and
503 analysed over time together with the time-dependent settlement values. For example, in case a
504 settlement value is unexpectedly high, the corresponding thrust force and grouting pressure values
506 For this purpose, a dedicated model view is defined as depicted in Figure 15. The specific model
507 view includes the geometry of the boring machine model, the geometry of the grouting elements
27
508 from the tunnel lining model and the geometry of the built environment model indicating the spatial
509 extend of the current tunnelling progress. With regard to the ring segments, only the ring number
510 and the built finish time are defined to reference both settlement parameters and machine data
511 parameters. First, settlement data is included in the model view. It contains positions and vertical
512 displacement values over the period of the tunnelling process. Second, machine data is part of the
513 model view, where the specific parameters can be referenced by built finish time or ring number.
Object A ributes
Model Object
Geometry Brep
Brep Geometry Domain Model
Building Layer
Mass Parameters Name
Parameters
Built Mass
Ground
Environment
Displacement Value
514
515 Figure 15: Model view for the settlement visualisation application (required parts are black and
517
518 The settlement data points are visualised using 2D mesh grids, which indicate settlement values by
519 colour. For generating the meshes two different interpolation methods are utilized. On the one hand,
520 a triangular interpolation can be applied, which creates an irregular triangular grid using Delaunay
521 triangulation and which determines corresponding colour values by linear interpolation (Figure
522 16a). Such a technique can provide results in real-time, even when processing large data sets. The
28
523 accuracy of triangular is sufficient when the measurement points are regularly distributed, for
524 example, interpolating terrestrial settlements. But when measurement points are unlikely scattered
525 and spatial variations are at hand, geostatistical methods are more adequate. Therefore, on the other
526 hand, a Kriging interpolation is applied to provide more reliable results. In contrast to the first
527 interpolation method, this interpolation estimates linear unbiased results by constraining all point
528 weights to sum up to 1. The specific weighting is defined by using a variogram. By minimizing the
529 variance of the system, a minimal value is obtained, and additionally, also uncertainty information
530 is provided. Mathematical descriptions of various Kriging approaches can be found in [35]. The
531 default approach is Ordinary Kriging, however, it can be less efficient on large data sets. Here, the
532 so-called Simple Kriging has been utilised, which differs in the way that the expectation value is
533 assumed zero constantly over the entire domain. The interpolation is sampled over a regular grid
534 where the vertices correspond to the kriging estimates. The function values are then visually
536
537 Figure 16: Visualisation of settlement data in context of existing buildings (grey) using coloured
538 mesh grids generated from either triangular interpolation (a) or Kriging interpolation (b)
539
540 Figure 17 illustrates a specific real project situation, in which the high thrust force value (Figure
541 17, top left zoom box, peak in curve) is correlated with a heaving above the TBM (Figure 17, top
542 right zoom box, blue colour). According to internal project information, the TBM had to drive
29
543 through a bearing slurry wall that caused heaving on the surface. This exemplifies and verifies the
544 high potential of an integrated information model when visualizing different interacting aspects of
545 a tunnelling project in order to communicate potential problems and identify correlations. More
546 details on the aspect of visualisation for settlement monitoring purposes can be found in [29].
547
548 Figure 17: Time-dependent visualisation of settlements in the context of the (static) built
549 environment and the (dynamic) TBM performance. Note the high thrust force that is correlated to
551
552
30
553 4.4 Case study on generating numerical simulations
554 The integrated tunnel information model of the Wehrhahn-Linie is also used to generate Finite
555 Element (FE) simulation models. For this purpose, a second dedicated model view, depicted in
556 Figure 18, is defined. This model view particularly specifies the required information regarding the
557 topology of the ground, the geotechnical parameters of the soil layers, the dimensions and
558 parameters of the surrogate building models, the advance rates of the boring machine as well as
559 measured support pressures and grouting pressures. More details on how the data has been collected
Model Object
Geometry Brep
Brep Geometry Domain Model
Building Layer
Eqivalent Mass Parameters Mass
S ffness Parameters Cohesion
Built Fric on
Ground
Environment Angle
561
562 Figure 18: Model view for generating the numerical simulation model (required parts are black
564
565 The procedure of generating FE simulations within the advanced FE framework ekate [3] is
566 depicted in Figure 19. In the first step, the relevant simulation is defined within the context of the
567 tunnel information model to specify model dimensions, number of rings and boundary conditions.
568 Based on this, all relevant FE input parameters for the soil, the existing buildings, the TBM and the
31
569 tunnel lining are automatically extracted. Next, this information is used to prepare an enhanced
570 CAD model that includes both geometry and material information as well as time-dependent
571 operational process data. Following that, the advanced FE framework ekate [3] [32] is used to
572 discretise the domain, solve the system and visualise the simulation results (Figure 19). It is noted
573 that the focus of this paper is not on the numerical analysis itself, but on how to use the proposed
574 tunnel information modelling framework to structure, link and prepare relevant information to
575 generate a numerical analysis model. More detailed information on the use of the FE framework
576 ekate to numerically simulate the mechanised tunnelling process can be found in [3].
577
578 Figure 19: Generating Advanced Finite Element Simulations to predict settlements using the
580
32
581 5. CONCLUSION AND FUTURE WORK
582 The construction process of tunnels using Tunnel boring machines (TBM) involves various kinds
583 of risks that are related to surface settlements, driving speed, gap grouting, face stability, etc. Since
584 risk is a combination of the aforementioned factors, their interactions have to be considered during
585 the design, the construction and the operation of a tunnel. In particular, interactions between the
586 ground, the TBM, the tunnel lining system, the above-ground buildings and the material flow have
588 In this paper a tunnel information modelling framework, basically providing four interlinked sub-
589 domain models and linked project performance data has been presented. Due to their distinctive
590 impact on the tunnel design and construction process a ground model, a boring machine model, a
591 tunnel lining model, and a built environment model have been individually created, and then linked
592 within an open IFC environment using the concepts of Proxies, Property Sets and Model View
593 Definition. Based on the proposed framework selected case studies on a real-world project have
594 been presented to verify its potential and advantages when (1) interactively visualising real time-
595 dependent settlement monitoring data in an environment-aware context and (2) generating
597 The main contribution of this paper is the conceptualisation, implementation and verification of a
598 tunnel information modelling framework that allows to (1) model the most important
599 interdependent components of a tunnelling system and (2) to capture and maintain the semantic
600 links between them to support more holistic studies related to risk evaluation and assessment in
601 mechanised tunnelling projects. This framework is set to be the basis for different engineering
602 applications and case studies, two of which (settlement visualisation, generate numerical analysis
604 While the contribution produces significant benefits for managing tunnelling projects, the presented
605 platform also conducts some computational and collaborative limitations. As the visualisation
606 component is essential for interacting with the model, fine detailed geometries or highly itemised
33
607 model assemblies, particularly considering larger projects, can interfere real-time interaction and,
608 therefore, limit the overall applicability. Thus, concepts for different geometric levels of detail and
609 hierarchical boundary representations should be considered, especially supporting such definitions
610 for the IFC Exchange Format. Also, exchanging IFC Files between projects members can suffer
611 from large file-sizes sent over limited bandwidths channels. For instance, tunnel models can result
612 into hundreds of Megabytes caused by redundancy. Therefore, the model mapping on import/export
614 Finally, collaboration is actually realised by (a) discussing within a project meeting, where the
615 participants stand in front of a common presentation device using the interaction platform, or (b)
616 exchanging digital information, including IFC-Files and related comments, using Internet
617 platforms. But targeting a broader set of devices and using limited bandwidth channels, this again
618 requires the need of computational optimisation. Moreover, future research aspects will consider
619 an implementation of this framework within the environment of BIM model servers, e.g. the Open
620 BIM server (www.bimserver.org). Besides web-based access, also the integration of the BIM
621 Collaboration Format (BCF) is of significant impact, enabling the recreation of synchronous
623
624
625 ACKNOWLEDGEMENTS
626 The authors gratefully acknowledge the financial support by the German Research Foundation
627 (DFG) for this work within the subproject D1 of the Collaborative Research Center SFB 837
628 “Interaction Modeling in Mechanised Tunnelling”. Special thanks to the City of Düsseldorf for the
630
34
631 REFERENCES
632
[1] M. Maidl, M. Thewes and U. Maidl, Handbook of Tunnel Engineering Part 1: Structures
and Methods, Part 2: Basics and Additional Services for Design and Construction, Wiley -
[2] V. Guglielmetti, P. Grasso, A. Mahtab and S. Xu, Mechanised Tunneling in Urban Areas,
[5] C. Eastmann, P. Teicholz, R. Sacks and K. Liston, BIM Handbook: A Guide to Building
Information Modeling for Owners, Managers, Designers., New Jersey: Wiley & Sons, Inc.,
2008.
http://www.buildingsmart-tech.org/specifications/ifc-overview/.
[7] E. Lebegue, B. Fies and J. Gual, “IFC-Bridge V3 Data Model - IFC 4, Edition R1,” 2012.
[8] Y. Ji, A. Borrmann, J. Beetz and M. Obergriesser, “Exchange of Parametric Bridge Models
Using a Neutral Data Format,” Journal of Computing in Civil Engineering, vol. 27, no. 6,
application of a road product model,” Automation in Construction, vol. 17, no. 6, pp. 719-
728, 2008.
35
[10] S. H. Lee and B. G. Kim, “IFC Extension for Road Structures and Digital Modeling,”
[11] N. Yabuki, T. Aruga and H. Furuya, “Development and application of a product model for
shield tunnels,” in Proc. of the 30th Intl. Symposium on Automation and Robotics in
Refined Product Model for Shield Tunnels Based on a Generalised Approach for Alignment
Representation,” in Proc. of the 1st Intl. Conference on Civil and Building Engineering
geometric-semantic modeling of shield tunnels for GIS and BIM applications,” Computer-
Aided Civil and Infrastructure Engineering, vol. 30, no. 4, pp. 263-281, 2015.
[14] A. Borrmann, M. Flurl, J. Jubierre, R.-P. Mundani and E. Rank, “Synchronous collaborative
[15] N. Vossebeld and T. Harmann, “Modeling Information for Maintenance and Safety along
the Lifecycle of Road Tunnels,” Journal of Computing in Civil Engineering, vol. 30, no. 5,
2016.
[16] G. Gröger, T. H. Kolbe, C. Nagel and K.-H. Häfele, “OGC City Geography Markup
Lanuage.,” Computers & Geosciences, vol. 31, no. 9, pp. 1095-1103, 2005.
36
LandXML scheme,” in Prof. of the 12th International Conference on Civil, Structural and
[19] T. H. Kolbe, “Representing and Exchanging 3D City Models with CityGML,” in 3D Geo-
2009.
[21] U. Isikdag and S. Zlatanova, “Towards Defining a Framework for Automatic Generation of
[22] I. Hijazi, M. Ehlers, S. Zlatanova, T. Becker and L. Berlo, “Initial Investigations for
Projects,” in Proc. of the 3rd Workshop on Formal Aspects of Virtual Organisations, Sao
37
[26] E. Curry, J. O'Donnell, E. Corry, S. Hasan, M. Keane and S. O'Riain, “Linking building data
in the cloud: Integrating cross-domain building data using linked data,” Advanced
[27] E. Corry, J. O'Donnell, E. Curry, D. Coakley, P. Pauwels and M. Keane, “Using semantic
web technologies to access soft AEC data,” Advanced Engineering Informatics, vol. 28, no.
[28] F. Hegemann, P. Manickam, K. Lehner, C. Koch and M. Koenig, “Hybrid Ground Data
[Accessed 16 05 2016].
[31] M. Venugopala, C. Eastman, R. Sacks and J. Teizer, “Semantics of model views for
information exchanges using the industry foundation class schema,” Advanced Engineering
[32] S. Schindler, F. Hegemann, A. Alsahly, T. Barciaga, M. Galli, K. Lehner and C. Koch, “An
Duesseldorf (Germany),” Geomechanics and Tunneling, vol. 7, no. 1, pp. 72-86, 2014.
model“, in Proc. of the 14th International Conference on 3D Web Technology, pp. 127-135,
Darmstadt, 2009.
38
View publication stats
633
39