Workforce location tracking to model, visualize and analyze workspace requirements in building information models for construction safety planning

Automation in Construction 60 (2015) 74–86

Contents lists available at ScienceDirect

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

Workforce location tracking to model, visualize and analyze workspace requirements in building information models for construction safety planning Sijie Zhang a, Jochen Teizer b,⁎, Nipesh Pradhananga c, Charles M. Eastman d a

Chevron Energy Technology Company, USA RAPIDS Construction Safety and Technology Laboratory, Germany c College of Engineering and Computing, Florida International University, USA d School of Architecture, Georgia Institute of Technology, USA b

a r t i c l e

i n f o

Article history: Received 2 May 2015 Received in revised form 3 August 2015 Accepted 19 September 2015

Keywords: Building information modeling Construction safety Global positioning system Lean construction Resource location tracking Workspace modeling and visualization

a b s t r a c t Safety as well as productivity performance in construction is often poor due to congested site conditions. We lack a formalized approach in effective activity-level construction planning to avoid workspace congestion. The purpose of this research is to investigate and prototype a new Building Information Modeling (BIM) enabled approach for activity-level construction site planning that can pro-actively improve construction safety. The presented method establishes automated workspace visualization in BIM, using remote sensing and workspace modeling technologies as an integral part of construction safety planning. Global Positioning System (GPS) data loggers were attached to the hardhats of a work crew constructing cast-in-place concrete columns. Novel algorithms were developed for extracting activity-specific workspace parameters from the recorded workforce location tracking data. Workspaces were finally visualized on a BIM platform for detecting potential workspace conflicts among the other competing work crews or between material lifting equipment. The developed method can support project stakeholders, such as engineers, planners, construction managers, foremen and site supervisors and workers with the identification and visualization of the required or potentially congested workspaces. Therefore, it improves the foundation on how decisions are made related to construction site safety as well as its potential impact on a productive and unobstructed work environment. © 2015 Elsevier B.V. All rights reserved.

1. Introduction Traditional safety planning mainly relies on manual observation, which is labor-intensive, time-consuming, and potentially highly inefficient. The link between planning for safety and work-task execution is often weak: for example, many contractors use two-dimensional drawings or field observations to determine hazard-prevention techniques [1,2]. The resulting safety plans are often error-prone due to subjective judgments of the available decision makers. Currently, historical workspace information for an activity and the corresponding contextual information depicting the condition under which the activity is accomplished are not stored. Hence, workspace planning for work activities in construction planning is often overlooked. Although knowledge is typically transferred from one project to the next, this important task could be optimized, especially when experienced field staff is hired elsewhere. These circumstances lead to workspace congestion often at the beginning of projects. This may then largely impede worker safety, health, and productivity on a ⁎ Corresponding author. E-mail address: [email protected] (J. Teizer).

http://dx.doi.org/10.1016/j.autcon.2015.09.009 0926-5805/© 2015 Elsevier B.V. All rights reserved.

construction project. There is a need for an approach to collect, formalize, and reuse historical activity-specific workspace information. The selected approach, other than previous approaches [1], describes an empirical study method that collects the active workspace, obtains its geometric parameters, visualizes the workspace, and detects workspace conflicts in building information models (BIM). A BIM-based application prototype for workspace visualization is eventually presented which demonstrates how this approach can assist activity-level construction planning. This paper is structured as follows: Section 2 reviews workspace representation techniques and existing studies on the application of advanced location tracking technology in the construction industry. Section 3 presents the objective and scope of this study. In Section 4, workspace conflict taxonomy and representations are presented. The computation of workspace parameters based on location tracking data and visualization of workspace in BIM are explained in Section 5. Workspace conflict detections are discussed in Section 6. Section 7 presents a case study to the proposed workspace conflict detection using the developed prototype system. A summary of the contributions and discussions about future research needs is presented in the final section of this paper.

S. Zhang et al. / Automation in Construction 60 (2015) 74–86

2. Background 2.1. Construction workspace representation Experts witness another phenomenon on construction sites: fast tracking and schedule compression create task overlaps on construction sites due to pressure for completion on time. This conflict occurs often, because multiple concurrent tasks compete for the limited workspace on site. Dealing with the planning and execution of simultaneous tasks and their workspaces is a main challenge that has been addressed in multiple workspace planning studies. For this reason, 4D (3D and time) representation and analysis of workspaces for construction activities during the planning, scheduling, and eventually already at the design phase, are encouraged, since they can minimize workspace congestion and conflicts which frequently exist at construction sites. It also keeps the construction personnel working safely and productively. Thabet and Beliveau [3] and Riley and Sanvido [4] presented a scheduling model that incorporates workspace constraints in the scheduling of repetitive work in multistory buildings. Their model proposed a method to define and quantify several workspace parameters (space demand b physical space demand and surrounding space demand N and space availability). Akbaş [5] described a geometry-based modeling and simulation approach called geometry-based process model (GPM) for modeling and simulation of construction processes based on geometric models and techniques, which provides improved modeling and simulation techniques for construction operations and more effective use of geometry for construction practice and research. However, GPM relies on the user to define the crew parameters and sequences to generate the activities and simulate the process given these parameters. Akinci et al. [6] firstly developed space templates linked to construction method templates to enable users to define the space requirements of different construction methods; secondly, they developed a prototype system called the 4D WorkPlanner Space Generator (4D SpaceGen) [7]. It uses the spatial requirement knowledge, captured generically in the space templates, to automatically generate the project-specific instances of spaces; thirdly, they formalized time–space conflict analysis as a classification task and addressed these challenges by automatically (1) detecting space conflicts, (2) categorizing the conflicts, and (3) prioritizing the multiple types of conflicts between conflicting activities [8]. However, their work did not consider the material travel path nor defined the required workspace. Choi et al. [9] classified workspace by its function and its relocatability to further represent different characteristics of a workspace. The latter one enables better integration of the workspace requirement and their planning processes. One limitation of this work is that enormous efforts are required to prepare the input data such as detailed construction schedules. Dawood and Mallasi [10] applied entity-based 4D CAD technology for detecting workspace congestion to help identify potential safety hazards on-site using critical space-time analysis (CSA) in 4D visualization. The proposed CSA associates certain visual features for workspace planning with the workspace competition. The PECASO (Patterns Execution and Critical Analysis of Site-space Organization) prototype was developed to encapsulate and evaluate the outcome of the CSA. Kassem et al. [11] created an Industry Foundation Class (IFC) compliant 4D tool for workspace management. Haque and Rahman [12] linked a 3D BIM model with the schedule and construction space requirements, and simulated the 4D model to detect whether there is any space conflict during the activities. Jongeling et al. [13] used distance between the different types of work as an important factor in safe and productive work execution, by manually extracting 4D spatial content from 4D CAD models. Zhang and Hu [14] proposed an integrated solution of analysis and management for conflict and structural safety problems during construction. Moon et al. [15] generated workspaces using a bounding box model and an algorithm in order to identify schedule and workspace conflict. Moon et al. [16] realized a BIM-based active simulation system

75

using genetic algorithm (GA) process for an alternative schedule to minimize the simultaneous interference level of the schedule-workspace. Su and Cai [17] presented a life-cycle approach to workspace modeling and planning. However, no scientific method is provided for generating the space requirement more rapidly, or eventually automatically. A summary and their shortcomings of some of the applied methods in recent 3D workspace management studies are listed in Table 1. Many existing and well-executed studies focused on critical space analysis and space planning which use workspace as input criteria in their developed systems. However, neither one of these approaches has provided reliable spatial information since their workspace input are either estimated based on the authors' background or experience, or it requires a user to determine their own input values. Riley and Sanvido [18] concluded therefore, that different materials and activities have repeating (predictable) space needs from one project to the next. The challenge is to find more appropriate ways to represent the workspace and to suggest acceptable workspace parameters. 2.2. Resource location tracking in construction Safety risks on construction sites are often closely related to the proximity of construction materials, equipment, and workers to nearby hazards [19,20]. Some of these are explicit, for example, the risk of falling from the leading edge of a concrete slab floor [2]. Some of the risks have also been defined and quantified in Hallowell and Gambatese [21] and Rozenfeld et al. [22]. Some researchers recommended using positioning devices to locate construction resources and deliver pro-active information in real-time to mitigate a worker from entering a hazardous area or performing unsafe or unhealthy work activities [23–27]. While some research investigated the error performance of positioning and path planning technology [28–30], Maalek and Sadeghpour [31] studied the performance of an Ultra Wideband (UWB) tracking system in static mode under conditions that commonly occur on construction sites. They proved that the accuracy of commercially available real-time location tracking technology could be used to display resource location in information models. They further indicated that “the accuracy of the system could be used in the definition of the size of buffer zones in construction site safety applications”. Many technologies exist today that might offer a solution for pro-active real-time hazard detection and warning based on pre-defined and geo-referenced hazard zones. Examples of research using location tracking technology for safety purposes are small GPS data loggers [32] and UWB [29,30]. Although each of the proposed technology has shortcomings and may only work under certain conditions in the harsh construction environment, they can gather valuable activity-based location data from resource (worker, equipment, and materials) movements. Once data is gathered and processed using computer algorithms, the resulting information has the potential to support workspace modeling and visualization. A construction site is a very dynamic environment in which workspace related to construction activities changes continuously. The locations and volumes of these spaces change in three dimensions and over time, according to project-specific design data. Unless advanced automation or lean approaches are applied, congestion among various work activities can often not be eliminated, which can lead to additional safety or health hazards [33]. Hence, there is a need for more effective activity-level construction safety planning. 3. Motivation to use novel technology and methods According to US Occupational Safety and Health Administration (OSHA), “routes for the suspended loads should be pre-planned in order to ensure that no worker has to work directly below a suspended load (except for those workers who must hook up or unhook the load, or work on the initial connection of the steel members).” [34]. From 1992 to 2006, 307 crane accidents in the private U.S. construction industry sector killed 323 workers [35]. In 2006, cranes contributed both as

76

S. Zhang et al. / Automation in Construction 60 (2015) 74–86

Table 1 Summary of the applied methods in recent 3D workspace management studies. Study

Feature Workspace computation

Workspace generation and representation

Workspace visualization and conflict detection

Workspace conflict rationalization and resolution

Akinci et al. [6–8]

User input

4D CAD prototype

Dawood and Mallasi [10]

Unit-based space quantification

4D CAD prototype

Prioritization of the time–space conflicts category –

Jongeling et al. [13]

–

–

Zhang and Hu [14]

Manually extracting the 4D content with assumptions on how to measure –

Parametric workspace representation Approximation envelope model of 3D element 2D

–

–

Moon et al. [15] Moon et al. [16] Su and Cai [17]

User input User input User input

4D CAD prototype 4D CAD prototype 4D CAD prototype

– Genetic algorithm based optimization –

Choi et al. [9]

User input

4D BIM

Pertinent strategies

Kassem et al. [11]

User input

Hierarchical bounding box model Bounding box model Assisted by User through GUI Product model with workspace adjustment Parametric workspace representation Assisted by user through GUI

Proposed approach

Historical location tracking data

4D IFC compliant prototype BIM authoring platform using API

Rule-based heuristic strategies using a centralized database Ontology-based reasoning

Parametric workspace representation

primary and secondary source of injuries to 72 of the fatal occupational injuries in the United States. This number is slightly lower than the average number of 78 fatalities per year between 2003 and 2005. 61% of these fatalities were categorized as “contact with objects or equipment” [36]. In 2012, Engineering News Record (ENR) published results to a case study stating that ‘worker contact’ was the cause of accidents in 46.7% of over 700 investigated crane-related accidents [37]. As many of these statistics indicate, safe crane operation requires well-coordinated activity planning, including all related processes and resources, such as involving the workers that rig material and the equipment [38]. In view of these statistics, detecting struck-by falling objects hazard is the focus in this research. This research aims to develop a general approach that collects, formalizes, and reuses historical activity-specific workspace information for automated activity-based workspace visualization and congestion identification in BIM. To limit the scope, this study focused on a highrisk activity common on high-rise construction sites: concrete column construction. According to a report of the Bureau of Labor Statistics (BLS) [34], (sub-) contractors pouring concrete for foundations and structures have been recognized as one of the specialty trades whose work is at very high risk of getting someone severely hurt, injured, or even killed. The GPS devices used in this research are commercially available Wintec G-Ray 2 data logger (see Fig. 1). The error analysis of this device can be found in Pradhananga and Teizer [32]. The research methodology includes the following steps. These are explained in detail in Sections 4 to 6:

based workspace modeling with BIM. The workspace sets considered in this study include: • Building component space: the space building component itself occupies; typically it is shown in BIM as a final product. • Worker space: the required space for a crew to perform its work; • Space for material handling path: the handling path required for the material movement, for example, the space required for moving a rebar cage from its staging area to the installation location using a crane. • Equipment space/temporary structure space: the space occupied by equipment, such as crane and scaffolding. • Protective space: the space needed to protect workers from safety

1) Define the workspace conflict taxonomy and workspace representation (Section 4); 2) Compute the workspace parameters based on worker location tracking data collected from construction site experiments (Sections 5.1 to 5.3); 3) Integrate the computed workspace parameters into the Construction Safety Ontology [39,47] as properties for the workspace visualization (Section 5.4); 4) Visualize the workspaces in BIM based on the identified workspace parameters and construction schedule (Section 5.5); and 5) Detect the workspace conflicts in BIM and analyze the results based on the workspace conflict taxonomy (Section 6). 4. Workspace modeling The goal and intention is to develop an activity-based workspace modeling method, and to create a framework to integrate activity-

Fig. 1. Monitoring confined workspaces using GPS data loggers mounted on construction workers' helmets.

S. Zhang et al. / Automation in Construction 60 (2015) 74–86

77

Table 2 Workspace conflict taxonomy (adopted and modified from [8]). Activity 1

Activity 2

Building component Worker space Material handling path Equipment space Protective space

Building component

Worker space

Material handling path

Equipment space

Protective space

Design clash

Congestion Congestion

Congestion Congestion Congestion

Congestion Congestion Congestion Congestion

No impact Safety hazard Safety hazard No impact No impact

The definition for the significance of using italics is that it highlights the scope of the developed approach.

Table 3 Examples of each space conflict leads to safety hazard.

authors [39]. Table 3 shows examples for two types of space interferences, which can lead to safety hazard.

Type of space interference Example Protective space × Worker space Protective space × Material handing path

Worker works under overhead loads (e.g. rebar cage, formwork, concrete bucket, precast concrete elements) Unauthorized worker moves material through post-tensioning area during tensioning operations

In this study, workspace is generated corresponding to the reference object (see Table 4). Reference surfaces are illustrated in gray color, required workspace for workers is shown with yellow dashed lines, and protective space is shown using red dashed lines. 5. Workspace parameter computation and visualization

hazards, such as a post-tensioning zone during rebar installations and space underneath crane loads during lifts.

Intersecting workspaces cause hazards that should be avoided. Table 2 shows the workspace conflict taxonomy: • Design clashes caused by two building components are outside the scope of this research since existing commercial available applications (e.g., clash detection and coordination) already solve this issue. • Congestion can be caused by several reasons; for example, worker space clashing with building component space reduces the space available for the workers, and material handling space clashing with another material handing space can cause a crew/equipment to wait for another crew/equipment to release the workspace. Workspace congestion usually results in the disruption of the workflow, which often leads to lower productivity [13]. However, it needs to be noted that congestion may also lead to more serious safety hazards in certain situations. For instance, two crane booms may collide due to poor construction planning. Therefore, congestion generally requires further case-by-case analysis to understand its potential impact. • Safety hazards can be caused by the conflict of either protective space and worker space or protective space and space for material handling path. It needs to be noted that safety hazards posted by the activity itself has been considered in previous JHA analysis research of the

It is assumed that analyzing worker location data can result in the approximate workspace that was used to complete a work task. The geometric data then generates the workspace parameters for a type of work activity. An occupancy grid model [40–43] is used for calculating the frequency of the visits of a worker to a predefined virtual cube, which represents part of the entire work area. After creating the occupancy grid map following Cheng et al. [29], algorithms were developed for generating and retrieving workspace parameters based on workforce location data densities. These parameters were used to represent distance offsets with reference to a building object. Finally, the parameters are used to generate the required workspace for each activity in BIM. It allows for safer work activity planning, if the same construction activity and method are used again. 5.1. Description of the experimental setting A three-day long experiment was conducted which included the observation of a multistory concrete structure (see Fig. 2). Data were collected on concrete column construction activities on the fifth floor, and included (a) frame the column with formwork, (b) erect bracing for the column formwork, (c) pour concrete in the column formwork, and (d) strip the formwork off the column. Two GPS tags were tagged to each of the hardhats of the three volunteering workers. They and a tower crane were involved in the activity (see Fig. 1). The crane lifted

Table 4 Workspace representations. Reference position

Above

Around

In front of

Below

Worker space for pouring a concrete slab h: worker height

Worker space for erecting or tying rebar of a column d: depth of the worker space

Worker space for installing or setting pins for a formwork wall element w: width of the worker space

Protective space preventing falling object hazards below a crane load H: distance between the load and the ground

Diagram

Example Parameter

78

S. Zhang et al. / Automation in Construction 60 (2015) 74–86

Fig. 3. Airborne photo from an unmanned aerial vehicle (UAV) generates (a) an orthophoto and (b) a manually annotated photo of the workspace and obstructions in plan view (marked with a red cross if the workspace is limited for a column).

and moved the formwork elements, rebar cages, and the concrete bucket associated to this work task. A commercial time-lapse and video camera according to Bohn and Teizer [44] was set up at a nearby structure to record the progress of the experiment over time. Both generated data that helped in post-analyzing the GPS data after the data collection effort. In addition, the accurate geometric information of the experimental environment of the complete structure was acquired using photogrammetry-based data from an Unmanned Aerial Vehicle (UAV) [45]. The generated imagery was used to get a better understanding of Fig. 2. Monitoring the experimental test bed on the construction site using time-lapse and video cameras.

S. Zhang et al. / Automation in Construction 60 (2015) 74–86

79

Fig. 4. Example data plot of the occupancy grid model.

the workspace parameters. It further helped establishing a correspondence between the GPS data and the BIM of the structure. Out of 51 pictures the UAV had taken, an orthophoto (plan view) of the as-built the construction site status was generated (see Fig. 3a). The highresolution image was manually annotated to analyze how many columns had severe space restrictions. According to the opinion of a professional safety engineer, out of 17 columns in the image under or soon under construction, 6 had severe spatial constraints. Equipment and material carts blocking parts of the workspace and general housekeeping issue, for example, had high likelihood to impact safety of the construction personnel (see Fig. 3b).

5.2. Data processing The GPS data were first transformed from the world coordinate system to the local coordinate system of the building model. Since the average distance between the concrete columns was 6.4 m, the value 6.4 m was used as a threshold to remove any GPS data outliers. The data were then filtered using a Robust Kalman Filter [46]. Kalman filtering, as it has been historically used for filtering and smoothing positioning or signal data, helped remove outlier data and error reads from the same type of GPS logger [32,47].

5.3. Workspace parameter computation A 2D occupancy grid model [40] is applied to visualize the different activity levels of the workers. Based on the previous research findings [32,47] and the knowledge of site dimension and error rate of the available GPS data logging devices, the construction space was divided into virtual spaces of identical dimensions: 0.5 × 0.5. Fig. 4 shows a gridbased map in plan view for erecting and stripping the formwork of a column. We explain further how the workers utilized the available workspace over time. The position of the column is shown as a small white rectangular box. The average location of the workers during the completion of all work tasks associated to that one column is illustrated by a yellow dot in Fig. 4. These areas are computed following a spiral pattern illustrated in Fig. 5. The spiral pattern was used to calculate the areas that were occupied by the workers 50%, 75%, and 100% of the time (e.g., the time it took them to erect and strip the formwork) from the average point on the 2D grid map. The areas are marked with red, green, and blue bounding boxes in the occupancy grid map, respectively. In addition, they are marked as S50, S75, and S100 (for later use in calculating workspace congestion and conflicts). Based on data from 9 columns (see Table 5), the average required workspace parameters for the activity stripping formwork from concrete columns, are 2 m to account for 50% of the time the workers spent near the column, 2.4 m for 75%, and 3.2 m for 100%, respectively. Based on data collected from observing all of the columns and all activities, the average workspace parameters for each activity are computed as well. Results are shown in Table 6 by activity type and by time classification (50%, 75%, and 100%). Note here that the task of erecting the rebar had been completed already. Table 6 denotes diagrams and the number of columns (#) that were observed. For each of the activities, mean, median, and standard deviation (SD) values are calculated based on the average of the data from all of the columns. The mean value is used as the workspace parameter. A further investigation was made with respect to the location and other spatial constraints of a column on the building floor, for example,

Table 5 Average distances between column edge and bounding boxes (in meters). Time

Fig. 5. Spiral pattern for calculating three activity levels in the occupancy grid model.

50% 75% 100%

Column# 1

2

3

4

5

6

7

8

9

Mean

2.15 2.47 3.47

2.33 2.21 2.35

2.47 2.85 3.60

1.85 2.23 3.23

1.97 2.60 3.60

1.99 2.85 3.73

1.47 2.10 3.10

1.72 2.22 3.10

1.59 2.09 3.00

1.95 2.40 3.24

80

S. Zhang et al. / Automation in Construction 60 (2015) 74–86

Table 6 Workspace parameters of the construction activities for a concrete column (in meters). Activity

Parameter

Column #

Time

Mean

Median

SD

Center shift

Erect formwork elements for the column

Diagram

d1

18

50% 75% 100%

1.23 1.69 2.45

1.3 1.7 2.45

0.19 0.15 0.31

0.42

Stabilize formwork by installing bracing support stands

d21

18

18

2.72 3.23 3.92 0.39 0.53 1.01

2.65 3.07 4.04 0.32 0.5 0.94

0.77 0.65 0.65 0.52 0.36 0.54

–

d22

50% 75% 100% 50% 75% 100%

Pour concrete in the formwork to build the column

d3

19

50% 75% 100%

1.04 1.37 1.89

0.95 1.32 1.95

0.17 0.21 0.26

0.2

Strip the formwork from the column

d4

16

50% 75% 100%

1.16 1.59 2.22

1.11 1.54 2.13

0.26 0.29 0.49

0.67

working near a leading edge or opening. Since workers intentionally or unintentionally avoid falling to a lower level (only one type of a severe hazard), the center of the utilized workspace usually shifts away from hazards. Such hazards can be geometrically marked, e.g. finding the nearest slab edge or opening to the column. In order to quantify how much the location of a column influences the workspace occupancy, another space parameter called center shift is introduced (see Fig. 6). For any of the columns that are observed, if one or two faces of a column

Fig. 6. Shift of the occupied workspace away from a potential fall hazard.

are close to the slab edge or a corner, the distance between the center of the occupied workspace and the center of the column is computed. As indicated in Table 6, center shifts are calculated based on the average value of all the columns. They are positive, thus meaning away from a leading edge or opening. As a result, further workspace is required at the opposite sides of the workspace. It needs to be considered in the work task planning phase. According to the average distances between the column edge and the bounding boxes (see Table 5), planners should keep in general an area of about 3.5 m to each side of a column free from obstructions. Such space will enable workers to work safely and effectively. Once collected and made available, such information becomes useful for foreman and supervisors for pre-planning the tasks of the next work day. Eventually additional workspace can be provided adjusted by the skill level of work crew. Although the length of the experiment and size of the construction site did not allow such a detailed study, the availability of such information may yield optimization of near real-time planning of workspaces, including material handling (e.g., efficient delivery routes and positioning) and crew and equipment management (e.g., crews overlapping). Since good resource leveling in construction is done on many jobsites already, a concrete crew precisely starts on building new columns after they have striped the formwork off columns they had built the days before. Otherwise they would lack the required material components. The developed algorithms calculate the corresponding times, including break and other delays, accordingly. Processing the data further, the time each activity requires can also be calculated (see Table 7). Even though productivity is not the focus of this study, it is used to automatically populate the start and end

Table 7 Aggregated times spent per sub-activity for building concrete columns. Activity

Place_Rebar_Cage

Frame_Column

Column_Bracing

Pour_Column

Strip_Column

Sum

Time

17%

19%

18%

35%

11%

100%

S. Zhang et al. / Automation in Construction 60 (2015) 74–86

81

Table 8 More realistic values used in 4D simulation add observed break times and other delays. Activity

Place_Rebar_Cage

Time interval

Frame_Column

Time Interval

Column_Bracing

Time Interval

Pour_Column

Strip_Column

Time interval

Sum

Time

13.6%

4%

15.2%

8%

14.4%

4%

28%

8.8%

4%

100%

time of each activity. This is particular of interest to many practitioners that intend to link 4D simulations in BIM to realistic data collected in the field. As such, this study has taken a first step towards solving such demand. Data to all major activities in building a concrete column is collected (place pre-fabricated rebar cage, erect formwork, install bracing, pour concrete, strip formwork). Considering some breaks that occur between the sub-activities (e.g., a delay occurs between placing a rebar cage and getting ready to install the formwork), standard time intervals that were observed via observation through a video camera are added (see Table 8). 5.4. Construction Safety Ontology extension with workspace parameters A Construction Safety Ontology is proposed to formalize the safety management knowledge in [39,48]. Based on the developed Construction Safety Ontology, Fig. 7 shows an extension of that with workspace information included. Each Job Step is linked with a set of workspace including worker space, equipment space and etc. As an example, Stand_Forms_Into_Place (see Fig. 8) requires a concrete crew as a resource, which occupies StandForms_WorkerSpace. The computed workspace parameters are stored in the workspace class as properties. In addition, the reference position of the workspace to a building element is also specified according to Table 3. 5.5. BIM-based workspace visualization and JHA communication Based on the master schedule made in a commercially available BIM platform, e.g., the Task Manager of Tekla Structures, the detailed activity-level construction schedule needs to be populated for

generating the accurate geometric workspace information. Instead of building one column after another, it is found that workers on a construction site tend to work on one type of activity after the previous one has been completed, i.e., place the rebar cage for all eight columns and then install all eight formwork elements belonging to these rebar cages. The pseudo-code for generating a detailed schedule for each activity is shown in pseudo code of the algorithm in Table 9. Fig. 9 shows the feasibility of generating and visualizing an activitybased workspace in BIM. Along with a 4D simulation of the expected construction progress, workspace sets can be visualized by referencing the building elements in BIM that are under construction at every time stamp. As shown in an example in Fig. 9, the workspace used for installing the formwork for one concrete column (in orange) is illustrated using pink (50%), green (75%), and light blue (100%) cubes. By default, the height of the space cubes is set to be equal to the height of the column. The percentage indicates the spatial-temporal relationship of the occupied workspace and the time required for the construction workers to complete the work task. The location of the column in regards to the slab is also computed to determine whether a space center shift is needed. As explained earlier, the workspace shifts once work occurs near a leading edge or an opening in a slab. Illustrated in an example in Fig. 9, the column is very close to a corner of the slab, thus potentially causing a fall hazard. The novel design of the workspace is set automatically using the ontology by shifting it according to the center shift s. The application of the author's job hazard analysis (JHA) tool [39] also considers the geometric conditions of the column and inserts a Fall_To_Lower_Level potential hazard as shown in Fig. 10. Precise instructions to field supervisors and workers can be provided in written and visual format before work starts.

Fig. 7. Extended Construction Safety Ontology to include workspace information [39].

82

S. Zhang et al. / Automation in Construction 60 (2015) 74–86

S75, and S50 are considered to be minor, moderate, and severe degree of congestion, respectively. Safety hazards due to use of or intrusion into a reserved workspace cause similar conflicts accordingly. Based on Table 2, SWRL (Semantic Web Rule of Language) rules were developed to check against the multiple-conflict potential between the occupancy of workspaces. The SWRL rules for (1) the conflict between two worker space and (2) the conflict between worker space and protective space are shown as below: Worker_Space(?ws1) ∧ Worker_Space(?ws2) ∧ conflictWith(?ws1, ?ws2) → hasConflictResult(?ws1, “congestion”) ∧ hasConflictResult (?ws2, “congestion”) (SWRL Rule — WorkerSpace&WorkerSpace) Worker_Space (?ws1) ∧ Protective_Space(?ps2) ∧ conflictWith(?ws1, ?ps2) → hasConflictResult(?ws1, “safety hazard”) ∧ hasConflictResult (?ps2, “safety hazard”) (SWRL Rule — WorkerSpace&ProtectiveSpace) Fig. 8. Workspace parameters for Stand_Forms_Into_Place (Job Step).

6. BIM-based workspace conflict detection Workspace conflicts are detected based on the geometric conditions of different settings in the workspace. According to Table 2, this research intends to mainly detect two types of major workspace conflicts: (1) workspace congestion and (2) safety hazards. In terms of congestion, the degree of workspace congestion can be determined for each of the activities based on their conflict volume, and it is defined after the conflict ratio in [8] and the space capacity factor in [3]. In terms of the space available to the workers or equipment (Sa), the degree of workspace congestion further considers different, overlapping occupancy levels of the workspace. The cases of Sa conflicting with S100, Table 9 Pseudocode for detailed schedule calculation for each activity.

The SWRL Rule WorkerSpace&WorkerSpace checks for the workspace conflict between two Worker_Space (?w1 and ?w2). If a conflict exists, i.e., ?w1 and ?w2 overlap in geometry, both ?w1 and ?w2 will have the hasConflictResult attribute filled with “congestion”. Similarly, the SWRL Rule WorkerSpace&ProtectiveSpace checks for the workspace conflict between Worker_Space (?w1) and Protective_Space (?ps2). Once a conflict exists, both ?w1 and ?ps2 will have the hasConflictResult attribute filled with “safety hazard”.

7. Case study A case study was conducted to test the workspace conflict detection method. Three major concerns when optimizing workspaces for safer and more productive work are addressed: 1. Workspace congestion detection: Aside interior finishing, one of the most frequent workspace congestions observed at tight construction sites is during formwork and concrete pouring of column construction. Projects facing ever tighter finishing schedules through lean management approaches make accomplishing these tasks more important [1]. For the purpose of testing the proposed method, a compressed, fictive schedule was created so that the shoring activity for the upper level slab and stripping column activity need to be executed at the same time. The space conflict detection uses the same approach as the clash detection in BIM. The system checks for overlapping geometry among the workspace boxes. The workspaces are visualized using colorful boxes. As shown in Fig. 11, a space

Fig. 9. Workspace visualization for the activity “erecting framework for a concrete column” in BIM (left. 3D view, right. top plan view).

S. Zhang et al. / Automation in Construction 60 (2015) 74–86

Fig. 10. Job hazard analysis considering geometric condition.

conflict is detected, and the user-interface of the developed prototype displays the type and severity of the conflict. 2. Fall and struck-by safety hazards detection: The safety hazards that were focused on in this case study are: (a) fall to lower levels and (b) struck-by an overhead crane load. The first was solved by

83

automatically detecting and identifying the fall hazards using the author's BIM safety rule checker (explained in detail in [2]). The second was solved using the developed prototype that aims to detect workspace conflicts between worker and equipment. In this case, a protective space underneath the crane load (crane lifting path) is created. The crane lifting path is simplified using a rectangle with four meter width (See Fig. 12). The material path originates in the material laydown area next to the building under construction and leads to the location of the construction activity currently underway (e.g. material drop zone). Based on Table 4, a protective space is automatically designed in BIM to box-in the space of the crane lifting path on the top level of the concrete slab. This has to be maintained free of other work crews until the material lift is completed. Fig. 12 illustrates the identified potential struck-by hazard caused by overlapping the zone of the crane lifting path and space the workers require for the columns. 3. Safest construction sequence: One additional application of the developed prototype is to compare two construction sequences in terms of their safety performance. Two simulated construction sequences, using the empirical workspace parameters collected from data in the field, evaluate the safest operation of the construction project. ‘Safe’ in this case means minimal overhang loads over work crews on the ground. One simulation to erect the building starts from building section A first, then B, and finally C. The other simulation starts from section C to B, and B to A, instead. Both sequences are assumed to require the same time and productivity, while the temporary laydown yard position was given to the left of building section

Fig. 11. Fall protection guardrail automatically modeled and workspace congestion identified in BIM between the activities of erecting and stripping formwork of two nearby columns (left: isometric view, top right: plan view, bottom right: text output of user interface).

Fig. 12. Potential struck-by or falling load hazards automatically identified in building section A when using a tower crane and a concrete bucket to erect columns in building section B (main image: isometric view in BIM, inlet image: plan view).

84

S. Zhang et al. / Automation in Construction 60 (2015) 74–86

Fig. 13. Automated 4D simulation in BIM determines unsafe work sequence from ‘A–B–C’ and safer work practices in sequence ‘C–B–A’.

A. Two 4D simulations in BIM were run to identify potential struckby and loads over work crew hazards. Comparing the results allows determining the level of safety performance expected in the field. Such understanding may assist the project and the safety manager in making concise decision on safety and productivity. Their choice, in theory and in practice, should be the selection of the safest and most productive construction sequence. 4D simulations were run with a 5-minute interval for both construction sequences (‘A–B–C’ and ‘C–B–A’). Since the material laydown area is located next to section A, 25 potential struck-by or loads falling hazards were embedded within sequence ‘A–B–C’. Building the upper building floors according to sequence ‘C–B–A’ would have zero hazards. Fig. 13 illustrates the results of both work sequences. Objects in blue denote work progress, objects in orange are under construction, spaces in semi-transparent orange represent space reserved for formwork construction, spaces in semi-transparent yellow represent protected workspace for the crane lifting path, and spaces set in semi-transparent pink–green–blue colors are for workers-on-foot performing construction work at the concrete columns. Fig. 13 shows that work crews in sequence ‘C–B–A’ are protected by the already-built floor slab of sections C, B, and lastly A that has already been poured. 8. Conclusions and outlook This paper developed a novel framework and prototype methods that collect, formalize, and reuse historical activity-specific workspace information for automated activity-based workspace modeling and visualization of work congestion identification and safety analysis in BIM. Global Positioning Systems (GPS) data to worker location tracking

were collected to compute accurate workspace occupation parameters based on the occupancy level for all of the work activities involved in building concrete columns using a critical construction resources (a work crew, pre-fabricated rebar cages, formwork and bracing, concrete material, and a tower crane). The demonstrated prototype shows the feasibility of computing workspace parameters from location tracking data. The integration of the generated workspace parameters and the Construction Safety Ontology was explained for workspace visualization in BIM. Two types of mitigation of workspace conflicts, namely workspace congestion and safety hazards, were successfully tested. The developed prototype showed in case studies further the capability to visualize activity-based space needs, detect space conflicts, and infer conflict resolution potential by measuring consequences and severities. The study identified several limitations that future research may address: (1) The GPS data loggers used in the experimental phase of the study have a comparatively low accuracy and only work well in outdoor environments. High-end and more accurate GPS data loggers can improve the accuracy measuring the workspace parameters. For indoor construction activities, other sensing technology such as UWB could be deployed instead [49]. (2) The workspace for material movement needs to be represented with various densities in the same way as workers' space was considered in this study [50]. For instance, the movement of formwork during its set up and removal could be tracked to visualize temporary laydown areas on the building floors. Thus, highly detailed, automated data gathering can serve a more detailed 3D workspace congestion identification. The assessment of the severity of the space overlaps, which indicate the resulting potential in safety risks, deserves also further attention, as well as considering highly detailed planning of site layout and project specific site layout constraints.

S. Zhang et al. / Automation in Construction 60 (2015) 74–86

The current work assumes that the use space is independent of the number of crew members involved. More crew members or different construction methods, however, take more or less space. Further work in this area is warranted. Future research can build upon this initial work and explore the opportunities: (1) calibration of a use area based on an occupancy grid: to collect location tracking data from other work activities and type of projects in order to explore more accurate workspace representations for better numeric and shape illustration [51], (2) integration of workspace parameters into site layout planning or schedule optimization [52]: parameters should address these variations and generate a parametric form and should be able to adapt to edge of slab, holes and other occlusions, and (3) to conduct advanced field trials that explore its application to traditional construction safety risk analysis [53–57].

Acknowledgments The authors would like to thank the McCarthy Building Companies, Inc. for the generous access to their construction site. Any opinions, findings, or conclusions included in this paper are those of the authors and do not necessarily reflect the views of anyone else.

References [1] O. Rozenfeld, R. Sacks, Y. Rosenfeld, ‘CHASTE’: construction hazard assessment with spatial and temporal exposure, Constr. Manag. Econ. 27 (7) (2009) 625–638. [2] S. Zhang, J. Teizer, J.-K. Lee, C.M. Eastman, M. Venugopal, Building information modeling (BIM) and safety: automatic safety checking of construction models and schedules, Autom. Constr. 29 (2013) 183–195. [3] W.Y. Thabet, Y.J. Beliveau, Modeling work space to schedule repetitive floors in multistory buildings, J. Constr. Eng. Manag. 120 (1) (1994) 96–116. [4] D.R. Riley, V.E. Sanvido, Space planning method for multistory building construction, J. Constr. Eng. Manag. 123 (2) (1997) 171–180. [5] R. Akbaş, Geometry-based Modeling and Simulation of Construction Processes(Ph.D. thesis, Dissertation) Department of Civil and Environmental Engineering, Stanford University, Stanford, California, 2003. [6] B. Akinci, M. Fischer, J. Kunz, R. Levitt, Representing work spaces generically in construction method models, J. Constr. Eng. Manag. 128 (4) (2002) 296–305. [7] B. Akinci, M. Fischer, J. Kunz, Automated generation of work spaces required by construction activities, J. Constr. Eng. Manag. 128 (4) (2002) 306–315. [8] B. Akinci, M. Fischer, R. Levitt, R. Carlson, Formalization and automation of time– space conflict analysis, J. Comput. Civ. Eng. 16 (2) (2002) 124–134. [9] B. Choi, H. Lee, M. Park, Y. Cho, H. Kim, Framework for work-space planning using four-dimensional BIM in construction projects, J. Constr. Eng. Manag. 140 (9) (2014) 04014041. [10] N. Dawood, Z. Mallasi, Construction workspace planning: assignment and analysis utilizing 4D visualization technologies, Comput. Aided Civ. Infrastruct. Eng. 21 (7) (2006) 498–513. [11] M. Kassem, N. Dawood, R. Chavada, Construction workspace management within an Industry Foundation Class-Compliant 4D tool, Autom. Constr. 52 (2015) 42–58. [12] M.E. Haque, M. Rahman, Time–space-activity conflict detection using 4D visualization in multi-storied construction project, Visual Informatics: Bridging Research and Practice, Springer 2009, pp. 266–278. [13] R. Jongeling, J. Kim, M. Fischer, C. Mourgues, T. Olofsson, Quantitative analysis of workflow, temporary structure usage, and productivity using 4d models, Autom. Constr. 17 (6) (2008) 780–791. [14] J.P. Zhang, Z.Z. Hu, BIM and 4D based integrated solution of analysis and management for conflicts and structural safety problems during construction: principles and methodologies, Autom. Constr. 20 (2011) 155–166. [15] H. Moon, N. Dawood, L. Kang, Development of workspace conflict visualization system using 4D object of work schedule, Adv. Eng. Inform. 28 (1) (2014) 50–65. [16] H. Moon, H. Kim, C. Kim, L. Kang, Development of a schedule-workspace interference management system simultaneously considering the overlap level of parallel schedules and workspaces, Autom. Constr. 39 (2014) 93–105. [17] X. Su, H. Cai, Life cycle approach to construction workspace modeling and planning, J. Constr. Eng. Manag. 7 (7) (2014) 04014019. [18] D.R. Riley, V.E. Sanvido, Patterns of construction-space use in multistory buildings, J. Constr. Eng. Manag. 121 (4) (1995) 464–473. [19] J.W. Hinze, J. Teizer, Visibility-related fatalities related to construction equipment, J. Saf. Sci., Elsevier 49 (5) (2011) 709–718. [20] J. Teizer, B.S. Allread, C.E. Fullerton, J. Hinze, Autonomous pro-active real-time construction worker and equipment operator proximity safety alert system, Autom. Constr., Elsevier 19 (5) (2010) 630–640. [21] M.R. Hallowell, J.A. Gambatese, Activity-based safety risk quantification for concrete formwork construction, J. Constr. Eng. Manag. 135 (10) (2009) 990–998. [22] O. Rozenfeld, R. Sacks, Y. Rosenfeld, H. Baum, Construction job safety analysis, Saf. Sci. 48 (4) (2010) 491–498.

85

[23] J. Teizer, D. Lao, M. Sofer, Rapid Automated Monitoring of Construction Site Activities using Ultra-Wideband, Proceedings of the 24th International Symposium on Automation and Robotics in Construction, Cochin, Kerala, India 2007, pp. 23–28 (September 19-21). [24] C.-S. Park, H.-J. Kim, A framework for construction safety management and visualization system, Autom. Constr. 33 (2013) 95–103. [25] T. Cheng, J. Teizer, G.C. Migliaccio, U.S. Gatti, Automated task-level productivity analysis through fusion of real time location sensors and worker's thoracic posture data, Autom. Constr., Elsevier 29 (2013) 24–39. [26] T. Cheng, G.C. Migliaccio, J. Teizer, U.C. Gatti, Data fusion of real-time location sensing (RTLS) and physiological status monitoring (PSM) for ergonomics analysis of construction workers, J. Comput. Civ. Eng. 27 (3) (2013) 320–335. [27] A. Sivanathan, F.N. Bosche, W. Abdel, S. Mohamed, T. Lim, Towards a Cyber-Physical Gaming System for Training in the Construction and Engineering Industry, Proceedings of the ASME 2014 International Design Engineering Technical Conferences and Computers and Information in Engineering Conference IDETC/CIE 2014 August 17-20, 2014, Buffalo, New York, USA, 2014. [28] K.S. Saidi, J. Teizer, M. Franaszek, A.M. Lytle, Static and dynamic performance evaluation of a commercially-available ultra wideband tracking system, Autom. Constr., Elsevier 20 (5) (2011) 519–530. [29] T. Cheng, U. Mantripragada, J. Teizer, P.A. Vela, Automated trajectory and path planning analysis based on ultra wideband data, J. Comput. Civ. Eng. 26 (2) (2011) 151–160. [30] T. Cheng, M. Venugopal, J. Teizer, P.A. Vela, Performance evaluation of ultra wideband technology for construction resource location tracking in harsh environments, Autom. Constr., Elsevier 20 (8) (2011) 1173–1184. [31] R. Maalek, F. Sadeghpour, Accuracy assessment of ultra-wide band technology in tracking static resources in indoor construction scenarios, Autom. Constr. 30 (2013) 170–183. [32] N. Pradhananga, J. Teizer, Automatic spatio-temporal analysis of construction site equipment operations using GPS data, Autom. Constr. 29 (2013) 107–122. [33] J. Melzner, S. Zhang, J. Teizer, H.-J. Bargstädt, A case study on automated safety compliance checking to assist fall protection design and planning in building information models, Constr. Manag. Econ. 31 (6) (2013) 661–674. [34] U.S. Bureau of Labor Statistics, Incidence rates of nonfatal occupational injuries and illnesses by industry and case types, http://www.bls.gov/iif/oshwc/osh/os/ ostb3191.pdf 2011 (online, accessed Sep 12, 2013). [35] CPWR, Crane-related deaths in construction and recommendations for their prevention, http://www.cpwr.com/sites/default/files/CPWR%20Crane%20Rept% 20Recmmdtns%20Nov-2009-BLS%20UPDATED.pdf2009 (online, accessed April 4, 2014). [36] U.S. Bureau of Labor Statistics, Crane-related occupational fatalities, http://stats.bls. gov/iif/oshwc/osh/os/osh_crane_2006.pdf2008 (online, accessed Sep 12, 2013). [37] T. Hampton, Why Every Pick Needs a Plan, Engineering News Record, McGraw-Hill, 2012 8–9. [38] A. Shapira, M. Simcha, Measurement and risk scales of crane-related safety factors on construction sites, ASCE, J. Constr. Eng. Manag. 135 (10) (2009) 979–989. [39] S. Zhang, F. Boukamp, J. Teizer, Ontology-based semantic modeling of construction safety knowledge: towards automated safety planning for job hazard analysis (JHA), Autom. Constr., Elsevier 52 (2015) 29–41. [40] J. Teizer, C.H. Caldas, C.T. Haas, Real-time three-dimensional occupancy grid modeling for the detection and tracking of construction resources, ASCE J. Constr. Eng. Manag. 133 (11) (2007) 880–888 (Reston, Virginia). [41] J. Teizer, C.T. Haas, C.H. Caldas, F. Bosche, Applications for real-time 3D modeling in transportation construction, 9th International Conference on Applications of Advanced Technology in Transportation, Chicago, Illinois 2006, pp. 123–128. [42] J. Teizer, C. Kim, F. Bosche, C.H. Caldas, C.T. Haas, Real-time 3D modeling for accelerated and safer construction using emerging technology, Proceedings of the 1st International Conference on Construction Engineering and Management, Seoul, South Korea 2005, pp. 539–543. [43] J. Teizer, F. Bosche, C.T. Haas, C.H. Caldas, Real-time, 3D object detection and modeling in construction, Proceedings of the 22nd International Symposium on Automation and Robotics in Construction, Ferrara, Italy 2005, pp. 49–53. [44] J.S. Bohn, J. Teizer, Benefits and barriers of construction project monitoring using hiresolution automated cameras, ASCE J. Constr. Eng. Manag. 136 (6) (2010) 632–640 (Reston, Virginia). [45] S. Siebert, J. Teizer, Mobile 3D mapping for surveying earthwork projects using an Unmanned Aerial Vehicle (UAV) system, Autom. Constr. 41 (2014) 1–14. [46] Z.M. Durovic, B.D. Kovacevic, Robust estimation with unknown noise statistics, IEEE Trans. Autom. Control 44 (6) (1999) 1292–1296. [47] A. Vasenev, N. Pradhananga, F. Bijleveld, D. Ionita, T. Hartmann, J. Teizer, A. Dorée, Information fusion approach to increase the quality of GNSS data sets in construction equipment operations, Adv. Eng. Inform., Elsevier 28 (4) (2014) 297–310. [48] S. Zhang, F. Boukamp, J. Teizer, Ontology-based semantic modeling of safety management knowledge, Proceedings of Computing in Civil and Building Engineering 2014, pp. 2254–2262. [49] T. Cheng, J. Teizer, Real-time resource location data collection and visualization technology for construction safety and activity monitoring applications, automation in construction, Autom. Constr., Elsevier (2013) 3–15. [50] D. Grau, C.H. Caldas, C.T. Haas, P.M. Goodrum, J. Gong, Assessing the impact of materials tracking technologies on construction craft productivity, Autom. Constr., Elsevier 18 (7) (2009) 903–911. [51] J. Yang, P.A. Vela, J. Teizer, Z.K. Shi, Vision-based crane tracking for understanding construction activity, ASCE J. Comput. Civ. Eng. 28 (1) (2014) 103–112 (Reston, Virginia). [52] H. Said, K. El-Rayes, Performance of global optimization models for dynamic site layout planning of construction projects, Autom. Constr. 36 (2013) 71–78.

86

S. Zhang et al. / Automation in Construction 60 (2015) 74–86

[53] S. Zhang, K. Sulankivi, M. Kiviniemi, I. Romo, C.M. Eastman, J. Teizer, BIM-based fall hazard identification and prevention in construction planning, safety science, Saf. Sci., Elsevier 72 (2015) 31–45. [54] J. Teizer, Status quo and open challenges in vision-based sensing and tracking of temporary resources on infrastructure construction sites, Adv. Eng. Inform. 29 (2) (2015) 225–238. [55] N. Pradhananga, J. Teizer, Cell-based Construction Site Simulation Model for Earthmoving Operations using Real-time Equipment Location Data, Visualization in Engineering, 3:12, Springer Verlag, 2015. http://dx.doi.org/10.1186/s40327-015-0025-3.

[56] J. Teizer, Wearable, wireless identification sensing platform: self-monitoring alert and reporting technology for hazard avoidance and training (SmartHat), J. Inf. Technol. Constr. (ITcon) 20 (2015) 295–312. [57] A. Marx, M. König, Modeling and simulating spatial requirements of construction activities, Proceedings of the 2013 Winter Simulation Conference, 20 (2013) 3294–3305.