Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 9

Diperiksa Indeks kinerja objek = (3) Eq. (3) can be customized using a user-defined learning rate function.

If
f(LearningRate) = 1 (default), then the productivity rate is assumed to be constant. To avoid unintended extensions to the
ScheduleFinish very early in the start of a task, a threshold may be set such that the ScheduleFinish will not be adjusted if
the threshold has not been achieved. The threshold may be specified as a percent complete or a fixed time, and is highly
depending on the SDL, the nature of the project, and the user's needs. The default threshold = 0. It should also be noted that
the method used for calculating the anticipated new duration is not the focus of this paper. Eq. (3) is used as a proof of
concept, and it can be replaced with any other custom-defined equation without jeopardizing mod- ule 3's performance. This
work intends to automatically commu- nicate the new duration into 4D BIMs and adjust the schedule.
For each task, the ScheduleDuration and ScheduleFinish attributes of the IfcScheduleTimeControl are updated using the
calculated duration [Eq. (3)] and the new finish date. The algorithm Actual progress ratio(input) Expected progress[Eq:(1)]
updates each task's parent task based on its subtasks following the (2) same approach. To achieve this, the value of the
ScheduleFinish for the parent task is replaced with the latest finish date of its subtasks.
With respect to the previous example, Fig. 6(d) illustrates five color-coded wall objects, the actual and expected progress
of which are shown in Fig. 5(d). In this example, the expected progress, calculated based on the user-specified inspection
date, is equal to 50%. Hence, wall objects with actual progress equal to 50% (G1-1 and G1-5) are color coded green (on
schedule). Based on the performance index [Eq. (2)], the color representing a task behind schedule ranges from green to red,
whereas it ranges from green to dark blue if it is ahead of schedule [Fig. 6(d)]. Besides the approach employed in this work,
the algorithm can be easily modified to color code the 3D objects based on a subtask's progress (eg, insulation for the
partition example) or any other user-defined conditions.
As a result, if a subtask falls behind schedule, the parent task will also fall behind schedule. This bottom-up approach helps
reflect the effect of progress data incorporated into subtasks to their parent. It is imperative to adjust a parent task's
IfcScheduleTimeControl data based on its subtasks. Hence, the program iterates over the subtasks of a parent task and
reflects their effect in the scheduling attributes of the parent. To control the schedule performance of a parent task, vital to
project control, its percentage complete can be calculated using Eq. (1), in which the denominator should be equal to new
updated duration. In other words, when updating a parent task, its percentage complete is indirectly updated because of the
changes to its ScheduleFinish and duration.
To provide the user with flexibility in terms of rescheduling, only the tasks assigned to inspected objects [eg, WallsG1
and WallsG2 in Fig. 5(a)] are updated in terms of finish date and du-
Module 3: Schedule Updating ration. Hence, the user chooses how to reschedule the other affected tasks in coordination with
changes made to the schedule's logic or In module 3 [Fig. 3(c)], the schedule is updated based on the actual resource usage.
It is also important to note that this process uses the and expected progress. Even though it follows the same algorithm task-
object and task-task relationships and does not depend on the structure as the first and second modules (ie, it iterates over the
type of objects being inspected and updated. tasks assigned to inspected objects and on the inspected floors), it Fig. 5(f)
illustrates the updated schedule for the partition was separated from the second module so that a version of the example, with
f(LearningRate) = 1 and threshold = 0. Because updated 4D BIM model can be generated before the as-designed the work is
expected to be 50% complete, the new task durations schedule data are modified. are now shorter, longer, or the same as
before if the actual percent To update the schedule for in-progress tasks, a bottom-up complete is above 50% (tasks G1-3 and
G1-4), below 50% (task approach is employed in which changes are reflected from the G1-6), or equal to 50% (tasks G1-1
and G1-5), respectively. The bottom task layers, operation-level tasks, and tasks assigned to model's color coding [Fig. 6(d)]
corresponds to the updated sched- objects, to the top ones. This is important because 4D BIM tools ule [Fig. 5(f)]. Also, the
parent task (task G1) has a new duration calculate the schedule-related attributes of a task based on its sub- and finish date,
both automatically calculated based on the changes tasks. The algorithm updates the schedule for the tasks assigned to in its
subtasks. This way, the changes are reflected from the bottom objects by modifying their scheduled duration and finish
dates.
layers up to the main tasks, regardless of the number of subtasks First, the new task durations are calculated based on the
actual
and their sequencing. This is vital to the method's generic appli- and expected progress [Eq. (3)]. The new projected
duration,
cation, because each object type has its own operation level tasks, Eq. (3), takes into consideration both the as-designed
duration
resulting in a different subtask configuration.
ASCE 04017012-8 J. Comput. Kebudayaan. Eng.
J. Comput. Kebudayaan. Eng., -1--1
. devresersthgirlla; ylnoesulanosrepro F. ECSA thgirypo C. 7 1 / 4 2 / 2 0 no ELTSACWENFOYTISREVINU EHT ybgro .
yrarbilecsa
morfdedaolnwo D
For every newly generated entity, its reference ID is calculated based on the last existing or created entity instance in the
data model, and a new GUID is generated for it. For details of the GUID generation process (BuildingSMART 2016a).
Validation and Results
The developed 4D BIM updating method was implemented using IfcOpenShell open source library (IfcOpenShell 2015) and
Python 2.7, and it was tested on a Windows 64 bit platform with a 3.40 GHz core i7 CPU and 8 GB of memory.
In this work, the state of progress information is acquired using an image processing-based algorithm (Hamledari and
McCabe 2016; Hamledari et al. 2017) that receives as input a 2D image of an under-construction partition and returns its
actual state of work. Although the development of this progress assessment method is outside the scope of this work, it is
briefly explained to clarify how the actual progress data are generated. It is noted that BAUS's performance is independent of
the choice of reality capture technology used for progress assessment.
As shown in Fig. 7, this image processing-based method consists of four modules that are designed to visually recognize
a partition's studs, insulation, state of drywall work (installed, plastered, and painted), and electrical outlets. An image,
depicting an under-construction partition, is passed to the system without a priori information with respect to existing
components in the scene, and the above-mentioned modules are applied on it in a cascade scheme (Fig. 7). These modules
detect instances of studs, insula- tion, state of drywall work, and electrical outlets. Based on the detection of components, an
overall actual state is assigned to that partition. These states correspond to the work in progress for framing and electrical
rough-in (A), insulation (B), installed drywall (C), plastered drywall (D), and painted partition (E).
An estimated percentage complete can be assigned to each state (A to E). These percentages should be defined by
experienced project managers to mirror their standard practices. Because of the visual recognition of components, a
partition's progress infor- mation can also be generated for operation level work in the form of a 1D vector holding
percentage complete data for its subtasks.
Input Images
Fig. 7. (Color) Algorithm of the image processing technique used for progress assessment of partitions (images by Hesam
Hamledari)
ASCE 04017012-9 J. Comput. Kebudayaan. Eng.
J. Comput. Kebudayaan. Eng., -1--1
No
Insulation Module
Yes
Yes
Stud Module
No
Drywall Module
No
Insulation Module
Yes
In this case, the wall subtasks consist of: (1) framing; (2) electrical rough in; (3) insulation; (4) electrical finish; (5) drywall
(hanging and plastering); and (6) painting. The detailed assessment of progress can be beneficial to subcontractors and
provide each trade with a clearer view of progress.
Experimental Setup
A construction site was visited twice to capture the digital images and videos required for the visual assessment of progress.
The first visit took place during working hours, 21 days after which the site was visited again outside of working hours. The
3,400 m2 construction site was the 14th floor of a building project at the University of Toronto (Fig. 8). During both visits, a
manually- controlled quadcopter, equipped with an on-board camera, was used to capture videos of indoor partitions at
various locations at the site. Because the employed progress assessment method is designed for partitions, the validation
process focuses on the inspection and model updating of this object type. However, as discussed previously, the process is
independent of object type.
The project's 3D BIM, developed in Revit software by the architects, was extended to a 4D BIM [Fig. 9(a)] with 11 tasks
assigned to the model's 3D components including columns, slabs, and walls. The wall objects were assigned to two tasks
[Fig. 9(b)]. Both tasks comprised a number of uninspected partitions along with partitions inspected during the first visit,
second visit, and both visits. Each visit's inspected partitions were randomly divided between the two tasks. BAUS's
performance is independent of the number of wall-related tasks and their names. Synchro software was used in the
development of the as-designed IFC-based 4D BIM [Fig. 9(a)] because of its capabilities in both model import and ex- port
in IFC file format. However, any software application capable of reading IFC data models should be able to visualize
BAUS's outputs. The BAUS's outputs were directly imported into Synchro and no further processing was involved. They are
IFC-conformant data models and can directly be visualized by IFC-friendly software applications.
A progress monitoring system must be capable of updating the as-designed BIM regardless of the number of inspected
compo- nents and their spatial distributions over a site. As a result, of
Electrical Module
Determine Insulation Extent of Progress
. devresersthgirlla; ylnoesulanosrepro F. ECSA thgirypo C. 7 1 / 4 2 / 2 0 no ELTSACWENFOYTISREVINU EHT ybgro .
yrarbilecsa
morfdedaolnwo D
Fig. 8. (Color) Floor plan of the visited construction site with locations color coded based on the inspection time the model's
361 objects of type IfcWall, 104 were inspected over the two visits. Among those inspected, 31 were inspected only during
the first visit, 61 only during the second, and 12 of them during both visits. This structure was strategically chosen to
evaluate BAUS's robustness in real-life monitoring practices. Fig. 8 illustrates the spatio-temporal distribution of the
inspected loca- tions over the building floor. Using the quadcopter-captured videos, one frame was extracted per partition
resulting in 43 and 73 digital images for the first and second visits (a total of 116 images), respectively. The images are
1,920 1,080 pixels in size. The cap- tured images were manually assigned to the partitions to which they belong. Future
studies can exploit the quadcopter's sensor readings to automate the image-to-object assignment. However, the design of the
image processing method is outside the scope of this work. Because of the cyclic and continuous nature of progress
monitoring, an updated 4D model may replace the as-designed model for the next updating iteration. In addition, 4D BIMs
with various schedule hierarchy, task-object relationships, and with varying extent of schedule update may act as input to the
BAUS. Hence, whereas the first visit's update was carried out using the as-designed 4D BIM, two variations on the choice of
input models were considered for the second visit: (1) the as-designed 4D BIM, and (2) the first visit's three updated BIMs
(outputs of modules 1 to 3). It is noted that the module 3 output is the updated 4D BIM,
ASCE 04017012-10 J. Comput. Kebudayaan. Eng.
J. Comput. Kebudayaan. Eng., -1--1
whereas the other two modules' outputs are partially updated and were generated to evaluate BAUS's performance with
respect to the choice of input model. As a result, BAUS was evaluated in five testing scenarios (Table 2). In each test, the
input BIM passes through BAUS's three modules, and the output is an updated 4D BIM, in which 43 (first test) or 73 (other
four tests) partitions are updated. Accordingly, the same number of images as partitions are processed to acquire the
estimated percentage complete information for each updating test.
The user involvement includes data capture at the site and providing the input 4D BIM. Both the progress assessment and
4D BIM updating process are automated.
Results To achieve updates, BAUS not only modifies the existing data instances in the IFC data model, but it also populates
it with new entities and data structures that expand the input model. Hence, it is important to ensure the new entities are IFC-
conformant be- cause even a slight deviation from the schema may corrupt the entire data file. It is also important to ensure
that the modified and created object-task, task-task, and task-control relationships are correctly defined, and that each
object's incorporated progress and scheduling data match its input data (eg, percentage comple- tion extracted from an
image).
. devresersthgirlla; ylnoesulanosrepro F. ECSA thgirypo C. 7 1 / 4 2 / 2 0 no ELTSACWENFOYTISREVINU EHT ybgro .
yrarbilecsa
morfdedaolnwo D
Fig. 9. (Color) Results of the 4D BIM updating process: (a) the as-designed BIM and schedule; (b) the color-coded 4D BIM
for the first visit and its updated schedule; (c) the color-coded 4D BIM and updated schedule for the second visit using the
as-designed 4D BIM as input; (d) the second visit's updating results using as input the first visit's updated 4D BIM
Hence, multiple measures were taken to validate BAUS's outputs. First, updated BIMs were imported into Synchro and
Revit to visualize the changes and ensure that files are completely readable by BIM tools. Second, all the instances of
updated partitions were controlled one by one, and the modifications of all three modules were compared with the expected
results. Finally,
Table 2. 4D BIM Updating Tests Based on Their Choice of Input Model and Performance
Test Visit Input model Average run time per partition (s)
1 1st As-designed 43 (100) 0.15 2 2nd As-designed 73 (100) 0.2 3 2nd First visit's updated
4D BIM Number of correctly updated BIM partitions (updating accuracy) (%)
73 (100) 0.2
4 2nd Module 1's output (first visit's update)
73 (100) 0.2
5 2nd Module 2's output (first visit's update)
73 (100) 0.2
ASCE 04017012-11 J. Comput. Kebudayaan. Eng.
J. Comput. Kebudayaan. Eng., -1--1
the updated 4D BIMs were controlled for conformity to the IFC schema.
To control the updates for objects, the model updating perfor- mance was broken down into three parts: module 1
(hierarchical modification of the model), module 2 (progress updating and color coding), and module 3 (schedule updating).
For instance, an object is considered correctly updated by module 1 if its schedule hier- archy is correctly modified; the
object-task relationship and sched- uling information are correctly adjusted; and the model integrity is not jeopardized. Each
object's updated status, color-coding, scheduling data (such as dates), and incorporated progress ratios were manually
controlled with the expected values, calculated based on the input data.
Table 2 lists the model updating performance. The total number of correctly updated partitions are reported, along with
the updating accuracy measures [Eq. (4)]. The first test uses updating informa- tion collected on the first visit and tests 2 and
3 relate to the second visit but used different input models to demonstrate what impact the state of the input model has on the
results. Tests 2 used the original as-designed model, whereas test 3 used the updated BIM model that resulted from test 1.
Tests 4 and 5 further investigated the ro- bustness of BAUS by using partially updated BIM (output BIM
. devresersthgirlla ; ylnoesulanosrepro F . ECSA thgirypo C . 7 1 / 4 2 / 2 0 no ELTSACWENFOYTISREVINU EHT ybgro .
yrarbilecsa
morfdedaolnwo D
from Modules 1 and 2 of the first visit) as input for the second visit updates. The BAUS updates performed perfectly without
regard to the state of the input model. The average run times reported (Table 2) include the performance of BAUS's three
modules:
BIM updating accuracy(%) =
Number Total number of correctly of updated updated objects
objects
100 (4)
The results of the updating process are shown in Fig. 9, corresponding to test 1 in Fig. 9(b), test 2 in Fig. 9(c), and test 3
in Fig. 9(d). According to the results in the first visit, the work on the partitions is slightly ahead of schedule (shown in blue)
[Fig. 9(b)]. Hence, the parent tasks' durations appear to be longer than their subtasks for which partitions were inspected
[Fig. 9(b)]. In the second visit, the work appears on schedule (shown in green) for the majority of partitions and behind
schedule (shown in red) for a few [Figs. 9(c and d)]. The parent tasks' durations are lengthened by BAUS's third module
based on the delayed tasks; however, the task successors were not updated to provide the user with more flexibility in terms
of how they want the schedule updating to take place.
Fig. 9(d) depicts the result of updating the first visit's updated 4D BIM [Fig. 9(b)]. The partitions that were inspected in
the first but not the second visit are color coded in white. During the first visit, the duration of the tasks for which the work
was ahead of schedule was shortened. Hence, it is observed in the second visit's updates that a portion of the partitions
inspected during both visits, marked by circles in Fig. 9(d), appear slightly behind schedule (yellow) when using the
previously updated model [Fig. 9(d)], and on schedule (green) when using the as-designed 4D BIM as input [Fig. 9(c)].
To demonstrate the effect of input progress data for 4D model updating, the images used for progress assessment along
with their detected state and estimated percentage complete are shown in re- lation to their corresponding color coded
partitions in the second visit (Fig. 10). Here, the drywall sheets were scheduled to be hung. Therefore, the partitions with
hung drywall sheets are on schedule
(a) (b) (c) (d)
C-60% B-50% A-30% A-30%
(e) (f) (g) (h)
B-40% C-60% A-30% C-60%
Fig. 10. (Color) 2D images used for 4D model updating, their detected state of progress, and their estimated percentage
complete (images by Hesam Hamledari); (c), (d), and (g) partitions in framing state; (b) and (e) partitions in insulation state;
(a), (f), and (h) partitions with installed drywall
ASCE 04017012-12 J. Comput. Kebudayaan. Eng.
J. Comput. Kebudayaan. Eng., -1--1
(green); partitions with insulation are slightly behind schedule (orange); and partitions with framing and electrical in
progress are behind schedule (red).
The three modules' modifications to input BIMs were checked for conformity to the IFC EXPRESS schema using
IfcObject- Counter (IfcObjectCounter 2016). This IFC file validator analyzes the IFC file semantically and syntactically, and
generates a log for all the issues and nonconformant entities. After each updating scenario is complete, an updated 4D BIM
is generated, resulting in five BIMs. The models were passed to the validator, and all BAUS's modifications to the BIMs
successfully conformed to the schema specifications. For the list of other validator tools see (BuildingSMART 2016b).

Discussion
Considering BAUS's high accuracy rates (Table 2), its robustness to various real-life updating processes (eg, use of
previously updated, partially updated, or as-designed BIMs with various schedule hierarchies), and its low computation cost
(Table 2), it appears to offer a promising solution for the automated updating of 4D BIMs. Furthermore, BAUS operates
using low LoD 3D BIMs, requiring the least level of modeling effort. This is vital, be- cause object-based monitoring
solutions require the project-related components to be included in the model (Roh et al. 2011), a barrier to their practical
implementation because not all tasks of interest are explicitly modeled in a BIM (Akcamete et al. 2009).
The three BAUS modules do not require any user involvement because of their reliance on the automatically queried
object types and task-object relationships. This eliminates the dependency on naming conventions. The first module's
analysis of a model's hierarchy helps adapt BAUS to each input model. When a previ- ously updated BIM acts as input to the
system, it will automatically bypass module 1 because the model has already been prepared. The reported results (Table 2)
represent accuracy not only when oper- ating on the as-designed 4D BIMs, but also on its own previously updated 4D BIMs.
This ensures BAUS's reliable support for peri- odic BIM updating.
. devresersthgirlla; ylnoesulanosrepro F. ECSA thgirypo C. 7 1 / 4 2 / 2 0 no ELTSACWENFOYTISREVINU EHT ybgro .
yrarbilecsa
morfdedaolnwo D
The use of IFC data exchange format significantly aids BAUS's application and eliminates its reliance on proprietary
software packages. This file format has proved beneficial to other aspects of monitoring practices, such as automated indoor
path planning (Lin et al. 2013), and it fosters interoperable exchange of information, both during and after the construction
phase (Leite et al. 2011). According to the results of conformance testing, the model updating method does not jeopardize
the integrity of a model and its semantics. This is crucial because the proposed method creates IFC-conformant data
structures with which the BIM file is populated. In other words, the 4D BIM file is automati- cally expanded by the
algorithm, resulting in modifications to the machine-generated data model, which is significantly more challenging than
mere extraction of data from an IFC file.
Also challenging is dealing with the increasing size of the IFC file as information is updated and added. It would be
highly advisable to have the detailed progress information and highly detailed schedule semantics (eg, operation level work
and one- to-one object-task relationships) eliminated when the work on a floor, space, or a wall object is finished. To achieve
this, module 1's modifications should be reversed; an algorithm in BAUS iter- ates on tasks on each floor, eliminates object's
task and scheduling entities for the completed ones, and assigns their corresponding ob- jects to the parent tasks [moving
from Fig. 5(f) towards Fig. 5(a)]. As a result, the schedule appears more detailed when the work is underway on a floor, and
it becomes cleared of updating details when work is completed and moves to other areas. Schedule adjust- ments are
reflected in the parent tasks.
For the resulting 4D BIM to reflect the actual achieved schedule, it is highly recommended that BAUS be employed
regularly, eg, daily throughout the project. This allows the system to capture actual start dates for new activities. Once in
progress, percent complete and expected finish date can be regularly updated. When the task reaches 100% completion, the
expected finish date can be recoded to an actual finish date.
Updating a task's ScheduleFinish results in the update of the scheduled start and finish dates of its successor(s). When a
succeeding task is in progress, its actual start (previously expected start) has a different value than if BAUS had not updated
the sched- ule up to that point. In other words, updates for in-progress tasks result in the updating of the whole schedule. If
BAUS is only used intermittently, then the resulting update only reflects a point in time and may not provide reliable
information about projected comple- tion times. Supplemental information would be needed from the user, such as the start
dates for the tasks underway.
In an alternative approach to the hierarchical changes made to the input 4D BIM, a task can be assigned to several objects
contained within modeled spaces (IfcSpace) on each floor. This re- sults in the creation of as many wall-related tasks as the
number of spaces, which would provide an average assessment of walls in proximity to each other, especially a better
alternative to the one-to-one task-object relationship as a model increases in size. Also, the user can provide progress data
directly for parent tasks, in which case module 1 will be bypassed.
Whereas BAUS can take as input both an as-designed and an updated 4D BIMs, it is important that the model reflects the
site conditions as accurately as possible. This calls for more emphasis on regular progress inspections and continuous use of
such 4D model updating systems. However, if the model does not satisfy this condition (eg, project is delayed, or it has been
inactive), there are multiple measures that can be taken. First, if the user can provide BAUS with the amount of delay or
inactivity, it will iterate on all tasks for which the works in not complete, and the IfcSche- duleTimeControl attributes will be
automatically adjusted. This can
also be performed in the scheduling tools. Second, the user can set a limit on the extent to which a task's duration can be
shortened or lengthened. This can prevent updates for models that do not accurately reflect site conditions. For example, the
user will be notified and directed to a task, if its duration is shortened by more than a week, so the necessary corrective
measures can be taken in case there are significant discrepancies between schedule and actual site conditions.
Delays can affect the schedule's critical path. Hence, it is imperative for a model updating system to recalculate the
critical path each time the individual tasks are updated. Automated retrieval of the critical path can be achieved by
calculating the length of all the paths. For each task, the IsPredecessorTo attribute is used to acquire the list of all its
successors. This results in the calculation of all the existing paths. Some scheduling software also offers the same capability
because a model updating system (such as BAUS) updates individual tasks and subtasks.
Another area of research that needs further attention is the detection of progress for operation level tasks. This is
especially vital to model updating methods that develop and update highly detailed schedules. For example, in a few recent
research studies (Han and Golparvar-Fard 2014, 2015), the operation level works were automatically detected in digital
images using the frequency of the materials present in each frame. This provides model updating techniques, such as BAUS,
with more capabilities and supplementary semantic information.
Whereas BAUS continuously documents changes to a model, it is also important to design algorithms that verify the
input progress data and detect errors associated with progress detectors. One potential solution is utilizing multiple reality
capture technologies for multisensory data fusion techniques (Shahi et al. 2014) to help achieve more reliable input data.
Controlling the consistency in the input progress data over subsequent updates can also help identify the errors and ensure
more robust scheduling. Before a task is updated, a model updating system can query the status of its pred- ecessors and
successors to evaluate the logic behind the update.
IFC files are highly redundant (Jeong et al. 2009), and a 4D BIM updating system may employ an IFC optimization tool
while preserving the model integrity. Here, a series of experiments with IFCCompressor (Sun et al. 2015) resulted in an
average 35% size reduction for the 3D BIM and 60% size reduction for the as- designed 4D BIMs. The input 4D BIMs were
approximately 20 Mb for a single building floor and contained nearly 300,000 entities. Hence, it is vital that a 4D BIM
updating process populates the data model as efficiently as possible.
According to our experiments using Synchro, manual updating of the BIM resulted in a 10% model size increase on
average, whereas BAUS achieved the same updates with less than 1% increase. For instance, BAUS updates an as-designed
4D BIM with 263,879 entities by adding 866 new entities and modifying a portion of the existing ones. It also is noted that
updating a previously updated 4D BIM will result in a significantly lower size increase than the average 1%.
Limitations and Future Works
Even though BAUS's reliance on low LoD 3D BIMs facilitates its convenient use, future work should address the updating
of high LoD 3D BIMs. When a low LoD model [Fig. 6(a)] is used, regardless of manual or automated model update, the
schedule can be developed and updated in both low [Fig. 5(b)] or high SDL [Fig. 5(c)]. In both cases, BAUS updates the
schedule and progress for subtasks and parent tasks. However, when both the 3D model and the schedule are highly detailed,
this will be more challenging
ASCE 04017012-13 J. Comput. Kebudayaan. Eng.
J. Comput. Kebudayaan. Eng., -1--1
. devresersthgirlla; ylnoesulanosrepro F. ECSA thgirypo C. 7 1 / 4 2 / 2 0 no ELTSACWENFOYTISREVINU EHT ybgro .
yrarbilecsa morfdedaolnwo D
automated because of the many one-to-one relationships that need to updating and detection of progress for operation level
be formed between subtasks and highly detailed objects (eg, indi- works also require further research in coordination with
industry vidual drywall panels). Furthermore, attributable to the highly practices and trades associated with each object type.
cluttered and complicated nature of construction sites, the logical relationships between components, such as construction
sequenc- ing knowledge (Han et al. 2015), need to be exploited to infer the Acknowledgments progress of partially inspected
spaces.
The representation of both the as-designed and as-built progress data within the same 4D BIM should be investigated.
Currently, the as-designed and as-built information is explicitly available through the attributes of the
IfcScheduleTimeControl unless the schedule becomes updated, changing the as-designed information to projected ones.
Using completion ratios, the as-designed informa- tion is extractable from an updated schedule. However, this is not the case
when a model becomes repeatedly updated. It has been proposed (Akinci and Boukamp 2003) to add a new attribute for
holding context information (eg, as-built and as-designed) to the IfcRelationship class using the IfcRepresentationContext
concept.
The authors extend their gratitude to Adrienne De Francesco and Steve Miszuk from the University of Toronto; Tom Finan,
PMX Construction's principal; Emily C. Penn, PMX Construction's project manager; Steve Di Santo, Eastern Construction's
project coordinator; Jan-Willem Gritters, Perkins+Will Inc. Associate; and Evgueni Litvinenko, Perkins+Will Inc. Int.
Architectural Tech. for their tremendous help and support during the data collection stage, site visits, and granting access to
the project's model. The authors are also grateful for the financial support of the Natural Science and Engineering Research
Council Grant No. 203368- 2012. This issue needs to be studied in coordination with the IFC4 capabilities, such as the
IfcTaskType entity (Liebich 2013), that facilitate such a goal. References There are some challenges associated with the
generic use of BAUS for high detail schedule updating. First, operation level works differ for various object types. This
requires expert opinion on what the number of operation level subtasks should be, how they Akcamete, A., Akinci, B., and
Garrett, JH (2009). Motivation for computational support for updating building information models (BIMs). Proc., 2009
ASCE Int. Workshop on Computing in Civil Engineering, ASCE, Reston, VA, 523532. should be sequenced, and what
their approximate relative durations Akinci, B., and Boukamp, F. (2003). Representation and integration of are. The latter is
challenging because of the effect of various factors as-built information to IFC based product and process models for on the
relative durations of operation level works to the parent task. Furthermore, a single subtask configuration may not satisfy the
requirements of various trades. Hence, there is a need for further research on high detail level updating of 4D BIMs in
coordination with the industry practices related to each object type and involved trades.
This study validated BAUS's performance using an image
automated assessment of as-built conditions, NIST Special Publication SP, Gaithersburg, MD, 543550. Amor, R., and Ma,
H. (2006). Preservation of meaning in mapped IFCs.
Proc., EC-PPM, Taylor & Francis, Abingdon, VA, 233236. Belsky, M., Sacks, R., and Brilakis, I. (2015). Semantic
enrichment for building information modeling. Computer-Aided Civil and Infrastruc- ture Engineering, Malden, MA.
Bosche, F., and Haas, CT (2008). Automated retrieval of 3D CAD model processing-based method for generating progress
data as inputs objects in construction range images. Autom. Constr., 17(4), 499512. to the BAUS. Future works need to
investigate multisensory data
Brilakis, I., et al. (2010). Toward automated generation of parametric fusion (Shahi et al. 2014) using various reality
capture technolo- BIMs based on hybrid video and laser scanning data. Adv. Eng. Info., gies. This is vital because the
overall accuracy of a 4D BIM updat- ing process depends both on the updating algorithm (eg, BAUS) and the reality capture
technology. Integrating multiple sensor read- ings can provide more reliable progress results and reduce errors associated
with progress detectors. Future efforts should also be focused on the verification of progress data provided as input for
BAUS.

24(4), 456465. BuildingSMART. (2007). IFC2x edition 3. http://www.buildingsmart


-tech.org/ifc/IFC2x3/TC1/html/ (Feb. 25, 2016). BuildingSMART. (2016a). IFC GUID summary.
http://www.buildingsmart
-tech.org/implementation/get-started/ifc-guid (Feb. 25, 2016). BuildingSMART. (2016b). IFC file validators.
http://www .buildingsmart-tech.org/implementation/get-started/ifc-toolboxes/ifc -toolboxes-summary#IFC_file_validators
(Feb. 25, 2016). BuildingSMART. (2016c). Industry foundation classes (IFC). http://www .buildingsmart-
tech.org/specifications/ifc-overview/ (Feb. 25, 2016). Conclusion
Chen, K., Lu, W., Peng, Y., Rowlinson, S., and Huang, GQ (2015). Bridging BIM and building: From a literature review to
an integrated For 4D BIMs to be effective for progress tracking, they must be regularly updated to reflect the state of
progress at construction sites. In this work, a procedure was proposed to exploit the avail- able progress data for building
elements for automated updating of 4D BIMs. The BAUS method modifies the schedule hierarchy of
conceptual framework. Int. J. Project Manage., 33(6), 14051416. Chin, S., Yoon, S., Choi, C., and Cho, C. (2008).
RFID+ 4D CAD for progress management of structural steel works in high-rise buildings. J. Comput. Kebudayaan. Eng.,
22(2), 7489. Costin, AM, and Teizer, J. (2015). Fusing passive RFID and BIM for increased accuracy in indoor
localization. Visual. Eng., 3(1), 120. the 4D BIM, incorporates the progress data into it, color codes its Dankers, M., van
Geel, F., and Segers, NM (2014). A web-platform for objects based on their actual and expected progress, and updates the
linking IFC to external information during the entire lifecycle of a schedule accordingly. BAUS was tested in a real-life
indoor con- building. Proc. Mengepung. Sci., 22, 138147. struction site, taking into consideration various updating
scenarios and factors. The high accuracy rates, low computation cost, and the use of the open and neutral IFC data format are
indicative of its promise for the future of automated BIM updating. It is suggested that BAUS be used regularly, so that it
updates tasks for which the work is in progress. Future work needs to address the preservation of both as-is and as-designed
semantics throughout the construction
Ding, L., Zhou, Y., and Akinci, B. (2014). Building information modeling (BIM) application framework: The process of
expanding from 3D to computable nD. Autom. Constr., 46, 8293. El-Omari, S., and Moselhi, O. (2008). Integrating 3D
laser scanning and photogrammetry for progress measurement of construction work. Autom. Constr., 18(1), 19. Fathi, H.,
Dai, F., and Lourakis, M. (2015). Automated as-built 3D reconstruction of civil infrastructure using computer vision:
Achieve- phase; and the extension of the work to 5D/high LoD BIMs. Itu ments, opportunities, and challenges. Adv. Eng.
Inf., 29(2), 149161.
ASCE 04017012-14 J. Comput. Kebudayaan. Eng.
J. Comput. Kebudayaan. Eng., -1--1
. devresersthgirlla; ylnoesulanosrepro F. ECSA thgirypo C. 7 1 / 4 2 / 2 0 no ELTSACWENFOYTISREVINU EHT ybgro .
yrarbilecsa morfdedaolnwo D
Laakso, Froese, T., et al. (1999). Industry foundation classes for project manage-
M., and Kiviniemi, A. (2012). The IFC standard: A review of ment-a trial implementation. Electron. J. Inf. Technol.
Constr., 4, 5375. history, development, and standardization, information technology. Froese, T., and Yu, K. (1999).
Industry foundation class modeling for Electron. J. Inf. Technol. Constr., 17(9), 134161. estimating and scheduling.
Durability Build. Mater. Compon., 8(4), Lee, G., Won, J., Ham, S., and Shin, Y. (2011). Metrics for quantifying the 2825
2835.
similarities and differences between IFC files. J. Comput. Kebudayaan. Eng., 10 Gallaher, MP, O'Connor, AC, John, L.,
Dettbarn, J., and Gilday, LT .1061/(ASCE)CP.1943-5487.0000077, 172181. (2004). Cost analysis of inadequate
interoperability in the US capital Lee, Y.-C., Eastman, CM, and Lee, J.-K. (2015). Validations for facilities industry.
National Institute of Standards and Technology, ensuring the interoperability of data exchange of a building information
Gaithersburg, MD.
model. Autom. Constr., 58, 176195. Golparvar, M., Pe na-mora, F., and Savarese, S. (2009). D4AR A
Leite, F., et al. (2016). Visualization, information modeling, and 4-dimensional augmented reality model for automating
construction simulation: Grand challenges in the construction industry. J. Comput. progress monitoring data collection,
processing and communication.Civ. Eng., 10.1061/(ASCE)CP.1943-5487.0000604, 04016035. J. Inf. Technol. Constr., 14,
129153.
Leite, F., Akcamete, A., Akinci, B., Atasoy, G., and Kiziltas, S. (2011). Golparvar-Fard, M., Pe n a-Mora, F., and Savarese,
S. (2011). Integrated
Analysis of modeling effort and impact of different levels of detail sequential as-built and as-planned representation with D
4 AR toolsin building information models. Autom. Constr., 20(5), 601609. in support of decision-making tasks in the
AEC/FM industry. J. Constr Liebich, T. (2013). IFC4The new buildingSMART standard. IC Meet- Eng. Manage,
10.1061/(ASCE)CO.1943-7862.0000371, 10991116.
ing, bSI Publications, Helsinki, Finland. Hamledari, H., and McCabe, B. (2016). Automated visual recognition of
Lin, Y.-H., Liu, Y.-S., Gao, G., Han, X.-G., Lai, C.-Y., and Gu, M. (2013). indoor project-related objects: Challenges and
solutions. Proc., The IFC-based path planning for 3D indoor spaces. Adv. Eng. Inf., Construction Research Congress,
ASCE, Reston, VA, 25732582. 27(2), 189205. Hamledari, H., McCabe, B., and Davari, S. (2017). Automated computer
Lopez, R., Chong, H.-Y., Wang, X., and Graham, J. (2015). Technical re- vision-based detection of components of under-
construction indoor view: Analysis and appraisal of four-dimensional building information partitions. Autom. Constr.,
74(2017), 7894. modeling usability in construction and engineering projects. Han, KK, Cline, D., and Golparvar-Fard, M.
(2015). Formalized J. constr. Eng. Manage., 06015005. knowledge of construction sequencing for visual monitoring of
Meadati, P., Irizarry, J., and Akhnoukh, AK (2010). BIM and RFID work-in-progress via incomplete point clouds and low-
LoD 4D BIMs. integration: A pilot study. 2nd Int. Conf. Construction in Adv. Eng. Inf., 29(4), 889901.
Developing Countries, Florida International Univ., Miami, 570578. Han, KK, and Golparvar-Fard, M. (2014). Automated
monitoring of Navisworks Jetstream [Computer software]. Autodesk, Inc., San Rafael, operation-level construction progress
using 4D bim and daily site CA. photologs. Proc., Construction Research Congressm, ASCE, Reston,
Navon, R., and Sacks, R. (2007). Assessing research issues in automated VA, 10331042.
project performance control (APPC). Practice, 16(4), 474484. Han, KK, and Golparvar-Fard, M. (2015). Appearance-
based material Oh, M., Lee, J., Hong, SW, and Jeong, Y. (2015). Integrated system for classification for monitoring of
operation-level construction progress
BIM-based collaborative design. Autom. Constr., 58, 196206. using 4D BIM and site photologs. Autom. Constr., 53, 44
57.
Omar, T., and Nehdi, ML (2016). Data acquisition technologies for Hartmann, T., Gao, J., and Fischer, M. (2008). Areas
of application for 3D construction progress tracking. Autom. Constr., 70, 143155. and 4D models on construction
projects. J. Constr. Eng. Manage., 10 Park, J., Kim, B., Kim, C., and Kim, H. (2011). 3D/4D CAD applicability
.1061/(ASCE)0733-9364(2008)134:10(776), 776785.
for life-cycle facility management. J. Comput. Kebudayaan. Eng., 10.1061 IfcObjectCounter. (2016). Ifcobjectcounter.
http://www.iai.kit.edu/www /(ASCE)CP.1943-5487.0000067, 129138. -extern/index.php?id=1134&L=0 (Feb. 25,
2016).
Patraucean, V., Armeni, I., Nahangi, M., Yeung, J., Brilakis, I., and Haas, C. IfcOpenShell. (2015). Ifcopenshell open
source library. http://(2015). State of research in automatic as-built modelling. Adv. Eng. ifcopenshell.org/ (Feb. 25,
2016).
Inf., 29(2), 162171. Issa, R., Flood, I., and O'Brien, W. (2005). 4D CAD and visualization in Pazlar, T., and Turk, Z.
(2008). Interoperability in practice: Geometric data construction: Developments and applications, Taylor & Francis,
exchange using the IFC standard. http://www.itcon.org/paper/2008 Abingdon, VA.
/24, (Jan. 2017). Jeong, Y.-S., Eastman, C., Sacks, R., and Kaner, I. (2009). Benchmark Python 2.7 [Computer software].
Python Software Foundation, Wilmington, tests for BIM data exchanges of precast concrete. Autom. Constr.,Delaware.
18(4), 469484.
Randall, T. (2011). Construction engineering requirements for integrating Kim, C., Kim, B., and Kim, H. (2013a). 4D
CAD model updating using
laser scanning technology and building information modeling. J. Constr. image processing-based construction progress
monitoring. Autom.Eng. Manage., 10.1061/(ASCE)CO.1943-7862.0000322, 797805. Constr., 35, 4452.
Revit [Computer software]. Autodesk, Inc., San Rafael, CA. Kim, C., Kim, H., Park, T., and Kim, MK (2010).
Applicability of 4D Roh, S., Aziz, Z., and Pe n a-Mora, F. (2011). An object-based 3D CAD in civil engineering
construction: Case study of a cable-stayedwalk-through model for interior construction progress monitoring. bridge
project. J. Comput. Civ Eng., 10.1061/(ASCE)CP.1943-5487 Autom. Constr., 20(1), 6675. .0000074, 98107.
Sacks, R., Kaner, I., Eastman, CM, and Jeong, Y.-S. (2010). The rose- Kim, C., Son, H., and Kim, C. (2013d). Automated
construction progress wood experimentBuilding information modeling and interoperability measurement using a 4D
building information model and 3D data.for architectural precast facades. Autom. Constr., 19(4), 419432. Autom.
Constr., 31, 7582.
Shahi, A., Cardona, JM, Haas, CT, West, JS, and Caldwell, GL Koo, B., and Fischer, M. (2000). Feasibility study of 4D
CAD in (2012). Activity-based data fusion for the automated progress tracking commercial construction. J. Constr. Eng.
Manage., 126(4), 251260. of construction projects, Univ. of Waterloo, Waterloo, ON, Canada. Kopsida, M., Brilakis, I.,
and Vela, PA (2015). A review of automated Shahi, A., Safa, M., Haas, CT, and West, JS (2014). Data fusion construction
progress monitoring and inspection methods. Proc., 32nd process management for automated construction progress
estimation. CIB W78 Conf., Endhoven, Netherlands, 421431. J. Comput. Kebudayaan. Eng., 10.1061/(ASCE)CP.1943-
5487.0000436, Kropp, C., Koch, C., and Knig, M. (2014). Drywall state detection in im- 04014098. age data for automatic
indoor progress monitoring. Proc., Computing in Shen, W., et al. (2010). Systems integration and collaboration in Civil
and Building Engineering, ASCE, Reston, VA, 347354. architecture, engineering, construction, and facilities management:
Kropp, C., Koch, C., Knig, M., and Brilakis, I. (2012). A framework for A review. Adv. Eng. Inf., 24(2), 196207.
automated delay prediction of finishing works using video data Song, K., Pollalis, S., and Pe n a-Mora, F. (2005). Project
dashboard: and BIM-based construction simulation. Proc., 14th Int. Conf. on Concurrent visual representation method of
project metrics on 3D build- Computing in Civil and Building Engineering, Moscow State Univ.
ing models. Proc., Int. Conf. on Computing in Civil Engineering, of Civil Engineering, Moscow.
ASCE, Reston, VA.
ASCE 04017012-15 J. Comput. Kebudayaan. Eng.
J. Comput. Kebudayaan. Eng., -1--1
. devresersthgirlla; ylnoesulanosrepro F. ECSA thgirypo C. 7 1 / 4 2 / 2 0 no ELTSACWENFOYTISREVINU EHT ybgro .
yrarbilecsa morfdedaolnwo D
Weise, Sun, J., Liu, Y.-S., Gao, G., and Han, X.-G. (2015). IFCCompressor: A
M., Liebich, T., Tulke, J., and Bonsma, P. (2009). IFC support content-based compression algorithm for optimizing
industry founda-for model-based scheduling. CiBW78, International Council for Build- tion classes files. Autom. Constr.,
50, 115.
ings, Delft, Netherlands. Synchro [Computer software]. Tamagawa Seiki, Shin-Kamata, Tokyo.
Wu, Y., Kim, H., Kim, C., and Han, SH (2010). Object recognition Tang, P., Huber, D., Akinci, B., Lipman, R., and Lytle,
A. (2010). Automatic reconstruction of as-built building information models from laser-scanned point clouds: A review of
related techniques. Autom. Constr., 19(7), 829843. Tanyer, AM, and Aouad, G. (2005).Moving beyond the fourth
dimension with an IFC-based single project database.Autom. Constr., 14(1), 1532. Teizer, J. (2015). Status quo and open
challenges in vision-based sensing and tracking of temporary resources on infrastructure construction sites. Adv. Eng. Inf.,
29(2), 225238. Turkan, Y., Bosche, F., Haas, CT, and Haas, R. (2012a). Automated progress tracking using 4D schedule
and 3D sensing technologies. Autom. Constr., 22, 414421. Turkan, Y., Bosch, F., Haas, CT, and Haas, R. (2012b).
Toward automated earned value tracking using 3D imaging tools. J. Constr.
in construction-site images using 3D CAD-based filtering. J. Com- put. Kebudayaan. Eng., 10.1061/(ASCE)0887-
3801(2010)24:1(56), 5664. Xia, D., et al. (2014). Achievements and challenges in machine vision- based inspection of
large concrete structures. Adv. Struct. Eng., 17(3), 303318. Yang, J., Park, M.-W., Vela, PA, and Golparvar-Fard, M.
(2015). Con- struction performance monitoring via still images, time-lapse photos, and video streams: Now, tomorrow, and
the future. Adv. Eng. Inf., 29(2), 211224. Zhang, X., et al. (2009). Automating progress measurement of construc-
tion projects. Autom. Constr., 18(3), 294301. Zhu, Z., and Brilakis, I. (2010). Concrete column recognition in images and
videos. J. Comput. Kebudayaan. Eng., 10.1061/(ASCE)CP.1943-5487 Eng. Manage., 10.1061/(ASCE)CO.1943-
7862.0000629, 423433.
.0000053, 478487.
ASCE 04017012-16 J. Comput. Kebudayaan. Eng.
J. Comput. Kebudayaan. Eng., -1--1

You might also like