Automation in Construction 43 (2014) 59–72

Contents lists available at ScienceDirect

Automation in Construction

j ourna l homepage: www.e lsev ie r .com/ locate /autcon

Evolution of the crane selection and on-site utilization process formodular construction multilifts

Jacek Olearczyk a,⁎, Mohamed Al-Hussein a,1, Ahmed Bouferguène b,2

a University of Alberta, Department of Civil & Environmental Engineering, Hole School of Construction, 4-110H Markin/CNRL Natural Resources Engineering Facility, Edmonton,Alberta T6G 2W2, Canadab Campus Saint Jean, University of Alberta, 8406-91 Street, Edmonton, Alberta, Canada

⁎ Corresponding author. Tel.: +1 780 492 1637; fax: +E-mail addresses: jaceko@ualberta.ca (J. Olearczyk), m

(M. Al-Hussein), ahmed.bouferguene@ualberta.ca (A. Bou1 Tel.: +1 780 492 0599; fax: +1 780 492 0249.2 Tel.: +1 780 465 8719.

http://dx.doi.org/10.1016/j.autcon.2014.03.0150926-5805/© 2014 Elsevier B.V. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Recieved 5 June 2012Recieved in revised form 4 March 2014Accepted 8 March 2014Available online 27 March 2014

Keywords:Mobile crane liftLift logisticsSite layoutMathematical algorithmModular construction

Modular construction technology has beenused in building construction for decades, having beenwidely utilizedon sites with high traffic volume, restricted accessibility/storage, or business operations which cannot beinterrupted by long construction time. A key challenge in this method of construction lies in planning andexecuting lifts within a short timeframe. In this regard, proper crane selection and site layout optimization cansignificantly increase productivity and shorten the lifting schedule. This paper thus proposes a methodologyfor the crane selection process and introduces mathematical algorithms to assess the construction of multi-liftoperations. The modular lift process is divided into three stages: crane load and capacity check, crane location,and boom and superlift clearance. Each stage's parameters are introduced, analyzed, and graphically explained.The methodology logic is supported by a generalized mathematical algorithm and is applied and tested on acase study involving the construction of five three-storey dormitories in 10working days forMuhlenberg Collegein Allentown, Pennsylvania.

© 2014 Elsevier B.V. All rights reserved.

1. Introduction

Modular construction refers to the process of dividing a house orapartment building into smaller units (module or panels), to bemanufactured on a production line at a fabrication facility and thenmoved to the construction site for assembly. However, modularconstruction projects are often incorrectly associated with individualmodular units, which have historically provided shelters for a varietyof outdoor projects and activities at remote and urban locations. Overthe last decade themodularmethod of constructing offices, dormitories,hotels, and government facilities has warranted a second look. Due totime savings and high production quality at factory assembly lines,more individual businesses and organizations are recognizing theconvenience of having a building erectedwithminimal onsite construc-tion time. The decreased construction time compared to conventionalstick-built construction benefits all parties involved in a buildingconstruction project.

Modular construction is not new to the construction industry; how-ever, this term is primarily associated with single-family homes, or atmost low-rise, multi-family housing. There are few studies available

1 780 492 0249.ohameda@ualberta.caferguène).

describing a means of automating the CADmodel development processor using robotics inmodular construction, although several articles havepraised the concept ofmodular constructionwith respect to its ability toadd modular units to existing structures or significantly reduce onsitelabor [4,14,24]. Modular construction also plays an important role onsites where a large number of units need to be assembled within ashort construction time for an international event. However, the mostsignificant niches for this type of building are school facilities, campus/dormitory living, and affordable housing [2,5,9]. These units are notlimited to low-rise accommodations; in fact, the construction of high-rise facilities using a modular approach has also been considered [6].Other examples of successful implementation have been airport roofs;the NASA spacecraft building; and health care units, from singlecheck-up rooms to operating theaters or pharmacy centers [10,11,15,17]. These facilities or units can be customized over the Internet similarto the manner in which an automobile may be customized by thepurchaser [3]. Ready modules are delivered to the site for assembly,and in this respect a heuristic algorithm can be effectively utilized [7].

Cranes are the most critical equipment for handling materials on aconstruction site. Many factors contribute to the selection of the type,number, and location of cranes, and expert judgment is essential inthis process. Due to the increasing complexity of construction sitelayouts, a number of computer applications have been developed toassist practitioners in the selection and use of cranes [13,23]. Computermodels have been developed for crane lift planning. The approachdescribed by Tam [18] as well as by Tam and Tang [19] has analyzed a

60 J. Olearczyk et al. / Automation in Construction 43 (2014) 59–72

crane lift area and used a genetic algorithm (GA) to optimize towercrane operations. Matsuo [20] and Sivakumar [25] have concentratedon developing a path planner for two crane lifts. Ali Deen [8] andMashood [12] have continued with this research topic, employing GAto address the challenge of effectively utilizing equipment on congestedconstruction sites. Al-Hussein [1] and Moselhi [16] have introduced analgorithmbywhich to select an optimal cranewith respect to lift capac-ity, using 3D animation for visualization.

The present contribution is based on a case study in which five,three-storey dorms for Muhlenberg College in Allentown, Pennsylvaniawere assembled on site in just 10 working days. The extremely tightschedule combinedwith the substantial number of lifts to be performednecessitated minute analysis of the cost and constraints of each opera-tional step. In this context, and since the crane was the master elementof the project, its positioning and interaction with surrounding objectswere crucial for maintaining high efficiency of equipment operationsand meeting the planned schedule.

Addressing problems as they are encountered on site in specific sit-uations is a reactive approach typically leading to “quick fixes” thatcomprise efficiency and productivity. With regard to crane utilizationfor major projects, the challenge facing planners is to envision idealcrane lift operations without any interruption, where each movementof the lifted object is predefined. Planners must also seek to minimizeunpredicted crane and lifted object movements. The aim of this paperis to support decision making with respect to these challenges byaddressing some of the unknowns involved when crane lift planningissues are encountered. The main objective of the research is to breakdown the crane operation process into smaller components that canbe captured in the software language for the purpose of optimizationand ultimately for full crane lift automation operations.

2. Methodology

The crane selection and utilization methodology encompasses fourstages, as illustrated in Fig. 1. The input section contains available liftedobject parameters and crane/rigging configuration arrangements. Themain process includes a crane load capacity check, crane location

INPUTCrane data

Object data

Obstructions

Site layout

Rigging data

Crane Lo(fit on the g

Crane LCapacity

OUTPUTObject paths

databaseObstrucconflic

Boom & councleara

Object TraOptimiz

Fig. 1.Methodology of

assessment, craneboomclearance arrangement, and lifted object trajec-tory optimization, the latter of which is outside the scope of this paper.The main process is subject to restrictions including lifted object size,crane rental cost, specific ground preparation for crane placement,schedule, and weather conditions. As an output the user may extractinformation from each stage or analyze the final results, such asobject path, obstruction conflicts, or lifted object sequence prioritysuggestions.

The methodological concepts presented below were developed inconnection with a modular construction project in which five three-storey buildings were erected on a hill with a slope exceeding 7%.Each building required six modules per floor, plus the roof and thebridge connecting the ground level to the second-floor main entrance.For the purpose of the project described in Section 3, themost appropri-ate crane was selected following the three steps presented in Fig. 1: (i)crane capacity check; (ii) crane reach based on the selected optimal lo-cation; and (iii) boom and counterweight clearances.

2.1. Crane load capacity check

The crane lifting capacity must exceed the weight of the lifts andtheir associated rigging, which is calculated satisfying Eq. (1).

p OCð Þ≥Tw ¼ Lw þ Hw þ SLw þ SBw: ð1Þ

Where:

OC Lifted object gross capacityTw Total lift weightLw Lift weightHw Hook weightSLw Total weight of slingsSBw Total weight of spreader bar.

The term, p, introduced in Eq. (1), is a factor which is generallyselected to be lower than 1 (usually 0.85). In practice, it increases thesafety of the lifting operation by setting a limit on the object's weight

cationround)

oadCheck

tiont ID

Sequence priority

terweightnce

jectoryation

CRITERIAObject size

Lift procedure

Renting cost

Ground prep.

Area regulations

Standards

Weather restict.

Schedule

the main process.

Fig. 2. Site crane position.

61J. Olearczyk et al. / Automation in Construction 43 (2014) 59–72

at p% of the maximum allowable load for the specific crane selected forthe job.

2.2. Crane placement location

This is a far more complex operation than a simple load capacitycheck. Extensive research has been conducted on single-lift crane anal-ysis [1,13,16] andmulti-lift crane position placement [21,22]. For our re-search, which involves multi-lifts, the crane location input comprisesthree data segments corresponding to the site, crane, and lifting objectparameters. Site data provides information about available layoutareas, surveying and ground density, and obstructions under andabove ground level. Crane-related data includes outriggers, configura-tion, counterweight, and superlift space requirements, as well asspreader bar availability or boom attachment additions. Lifted objectdata includes weight, geometric center (GC), lifting lug dimensions,and information about additional supports for temporary storage.

In the context of the case study uponwhich the present contributionis based, the time factor related to the swing angle has been ignored,since the most critical issue is positioning the crane such that it is ableto lift objects from pick points to destinations. In this respect, thecrane location is expressed as an optimization problem described bythe following objective function:

minXnk¼1

mkdk þXpp¼1

mmax;pδp

" #ð2Þ

Where:

mk Mass of object, kdk Distance of object, kmmax,p Maximum weight from pick pointδp Pick point distance from the crane.

It is important to note here that the masses are included as part ofthe objective function since it is themomentswhichdeterminewhetheror not a lift is possible. There are three types of constraints, as follows:

1. The location of the crane determined using Eq. (2) leaves sufficientclearance for maneuverability, satisfying Eq. (3):

d Qk;Cð ÞNRCW þ CCW ð3Þ

Where:

d(Qk, C) Counterweight of closest distance to object kRCW Counterweight swing radiusCCW Counterweight-object clearance.

In contrast to a tower crane, the counterweights of which are at-tached to the boom,which swings high above any obstruction, amo-bile crane has its counterweights located at the base. As a result, thepositioning of a mobile crane not only needs to satisfy capacity con-straints, but also needs to allow the counterweight to swing freely,even when surrounded by multiple obstacles. Although Fig. 2 is spe-cific to the case study addressed later in this paper, it nonetheless il-lustrates the issue of counterweight clearance, a common problemwith mobile cranes on congested construction sites.To determine whether a candidate crane location, C : (xc, yc), isacceptable, we need to ensure that the assembled objects in the sur-rounding area do not interfere with the swinging of the counter-weight. This issue is depicted in Fig. 3.An algorithm is devised to test each candidate crane location C : (xc, yc)for sufficient clearance between the counterweight and surroundingobjects. This algorithm is described by a pseudo-code, as presentedin Fig. 4.

To prepare the above pseudo-code for implementation, the coordi-nates Ω : (Ωx, Ωy) are provided satisfying Eq. (4):

Ωx ¼ xA þxB−xAð Þ xC−xAð Þ þ yB−yAð Þ yC−yAð Þ

xB−xAð Þ2 þ yB−yAð Þ2 xB−xAð Þ

Ωy ¼ yA þxB−xAð Þ xC−xAð Þ þ yB−yAð Þ yC−yAð Þ

xB−xAð Þ2 þ yB−yAð Þ2 yB−yAð Þ:ð4Þ

The points (xA, yA), (xB, yB) and (xC, yC) are the Cartesian coordinates ofthe pair of consecutive corners (A, B) and corner C (the current candi-date crane location being tested by the procedure in Fig. 4), respective-ly.

2. Boom clearance analysis, especially with respect to static obstacles, iscritical for congested sites. As a result, it is critical to determinewhichboom angle will allow the crane to avoid obstacles while satisfyingthe clearance constraint. In this section, we provide a formula to cal-culate the main-boom angle in the most general case, namely, whenan extension is used. This case is depicted in Fig. 5.Based on Fig. 5, the relationship linking the clearance Δ to the boomangle α and the angle between the extension and themain boomφ isobtained using Eq. (5):

L sin αð Þ−H½ � cos α−φð Þ þ sin α−φð Þ d−L cos αð Þ½ � ¼ Δ: ð5Þ

The above equation is non-linear in the variable α and can only besolved numerically. The solution is subject to two constraints reachand lifting capacity which are addressed in another paper by thesame authors. At this juncture it is important to note that when theboom is not extended a corresponding formula for the boom anglecan be directly derived from Eq. (5) by setting φ = 0, thus leadingto Eq. (6):

d sin αð Þ−H cos αð Þ ¼ Δ: ð6Þ

Eq. (6) can then be transformed into a quadratic equation, (d2 + H2)x2 − 2dΔx + (Δ2 − H2) = 0, where x = sin(α) and for whichsolutions can be obtained manually.

3. The crane is able to reach the pick and set points,

dkbRmax mkð Þ≤L cos αð Þ þ Lext cos α−φð Þ: ð7Þ

Where:Rmax(mk) is themaximumreach of the crane liftingmk. In thisrespect, a manufacturer generally provides information about the

Fig. 3. Obstruction clearance.

62 J. Olearczyk et al. / Automation in Construction 43 (2014) 59–72

relationship between the crane's boom length and the maximumlifting capacity at that length. However, if such data is not available,basic statics principles can be used to express the maximum reachas a function of the lifted mass. According to Fig. 6, the crane is ableto lift the mass,mk, only if the following relationship is satisfied:

ndk2þmk � dkbp M � Dð Þ: ð8Þ

Where:

M CounterweightD Distance from the hinge point of the boom

Fig. 4. Testing of counterweight clearance

n Weight of the boom (assumed to be uniform)mk Mass of the object to be lifteddk Distance from the hinge pointp Percentage of total capacity (usually 85%).

As a result, the maximum reach allowed for the crane is;

Rmax mkð Þ ¼ p M � Dð Þn2

� �þmk

: ð9Þ

Since the distances occurring in the above equations involve Car-tesian coordinates, the leading optimization problem is non-

for a given candidate crane location.

63J. Olearczyk et al. / Automation in Construction 43 (2014) 59–72

linear. Consequently, a simpler approach is investigated in which

Fig. 5. General case (main boom & extension) when booming is required to avoid anobstruction.

Fig. 6. Typical crane mass distribution diagram.

Fig. 7. Hypothetical destination points corresponding to four objects weighing M, M, M,and 5M.

the crane location is chosen as the object's center of mass (cen-troid). In this respect, the crane coordinates are calculated accord-ing to the following equations:

x ¼

Xni¼1

mixi

Xni¼1

mi

; y ¼

Xni¼1

miyi

Xni¼1

mi

; z ¼

Xni¼1

mizi

Xni¼1

mi

ð10Þ

where n is the number of objects to be lifted, whilemi and (xi, yi, zi)are the mass and geometric center coordinates of object i, respec-tively. When the crane location is calculated using Eq. (10), thepick point information is not considered. In practice, to ignore pickpoint information is acceptable, provided there is at least one areawhich satisfies the lifting constraint, namely:ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

xp−x� �2 þ yp−y

� �2 þ zp−z� �2

r≤Rmax mmax:p

� �; p ¼ 1;2;…:; P :

ð11Þ

Here (xs, ys, zp) gives the coordinates of the pick-point andRmax(mmax,p)expresses the farthest crane reach calculated for the heaviest loadto be lifted from pick point p. However, in addition to the aboveconstraint, an even greater issue may be encountered when thecoordinates x; y; zð Þ are used as the location of the crane. Indeed,since the centroid is essentially the weighted average of the dis-tances between n objects and the crane, the crane location willbe attracted to the heaviest object. As a result, a crane may belocated in a position that cannot reach some destinations. Toillustrate this case, let us assume that a 16-m reach crane is usedto serve four destination points as illustrated in Fig. 7.Using the data in Fig. 7, the crane is located at the coordinates (5 m,5 m). Accordingly, the upper right corner in the above figure will beout of reach since its distance to the crane is 21.21 m. Fortunately,for the purpose of our case study, the masses of the objects aresufficiently close, thus allowing the crane to reach every destinationpoint. To complete the algorithm, Eq. (10) needs to be supplemented

by a test similar to that defined in Eq. (11), to ensure that a crane locat-ed at the centroid can reach every destination point. Formally, this testcan be written as:ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

xk−xð Þ2 þ yk−yð Þ2 þ zk−zð Þ2q

≤Rmax mmax:p

� �; p ¼ 1;2;…:; P

ð12Þ

in which (xk, yk, zk) are the destination point coordinates of object, k,with mass,mk.Fig. 8 shows the site layout of the Muhlenberg College project, andFig. 9 shows the project's geometric center in reference to one building.

Fig. 8. Site top layout view.

64 J. Olearczyk et al. / Automation in Construction 43 (2014) 59–72

2.3. Outrigger clearance

To finalize the crane placement location analysis, the outrigger orcrawler layout is evaluated accordingly. It is important to mention thateven though the formuladescribedbelow is specific to the case study con-figuration described by Fig. 10, it can easily be modified to fit other sitelayouts. Fig. 10 shows the clearance outrigger layout (Clol) dimension,which is calculated satisfying Eq. (13):

Clol ¼ yD−yc−Wol

2ð13Þ

where yc and yD are the coordinates of the crane set point and one of thecorners of the module referred to as “3 × 2”, respectively. In the figure, inthe ellipsoidal code “3 × 2”, it should be noted “3” represents the buildingnumber, “x” represents the elevation of the module from 1 to 3, and “2”represents module type from 1 to 6. (Since the example illustrated inFig. 10 is a top-view, the elevation is represented by the variable “x”.) Asobtained from the manufacturer, wol represents the outrigger's (orcrawler's) width. The calculated crane outrigger clearance (Col) must beat least the predefined minimum object distance (MOD) value to ensurethat the evaluated crane configuration will not be rejected from the list.

2.4. Boom clearance to slope surface analysis

During the course of theMuhlenberg College project whichmotivat-ed this research, clearance between crane components and surroundingobjects was a major challenge which needed to be accurately evaluatedin order to determine whether or not the required lifts could be safelyperformed. The derived formulas to calculate outrigger, counterweight,and boom clearances that would ensure crane maneuverability andusability in the presence of obstacles have been presented above. Inaddition to deriving these formulas, as the buildings were assembledit was necessary to investigate the clearance between the main boomthe sloped roof, which is a special type of obstruction. Although specificto the case study, this interaction, which is not addressed in the currentliterature, constitutes an extension of the case of a boom–flat roof inter-action previously developed by Shapiro. Since the boom angle canimpact the clearance, we propose the establishment of a relationshipbetween these two quantities using a limited set of parameters, essen-tially the coordinates of the edge of the roof (assumed to be a line)and the crane location, which is assumed to be centered on the origin(without any loss of generality), as shown in Fig. 11. Since the aim isto linkα to the clearanceΔ, the equation of the boomaxis is convenient-ly defined by spherical coordinates.

Fig. 9. Site GC to one building.

Fig. 10. Crane outrigger layout clearance.

65J. Olearczyk et al. / Automation in Construction 43 (2014) 59–72

Fig. 11. Clearance between main boom and sloped roof.

66 J. Olearczyk et al. / Automation in Construction 43 (2014) 59–72

The starting point for the equations below is the parametric equationsof the boom axis and roof edge, which are defined according to Eq. (14).

Craneaxis :x ¼ λuxy ¼ λuyz ¼ λuz

Roof edge :x ¼ x1 þ μ vxy ¼ y1 þ μ vyz ¼ z1 þ μ vz

8<:

8<: ð14Þ

where λ and μ are real numbers. The direction vectors u! and v! aredefined according to Eq. (15),

u! cos αð Þ cos φ0ð Þ; cos αð Þ sin φ0ð Þ; sin αð Þð Þ v! x1−x2; y1−y2; z1−z2ð Þ :ð15Þ

Note that for the case study, the value of the angleφ0 is arbitrary inso-far as it is constant. In order to calculate the smallest distance between theboom and the roof edge, we start by defining a line orthogonal to theboomand intersecting the edge of the roof. Since such a line is orthogonalto the direction vector u!, its dot product with P4P2

��!satisfies Eq. (16):

Xr¼x;y;z

ur r4−r3ð Þ ¼ 0⇒X

r¼x;y;z

ur μ vr−λur þ r2−r1ð Þ ¼ 0: ð16Þ

Using Eq. (16), the objective is to determine the constants λ and μwhich will lead to the smallest distance P4P3, which also representsthe clearance between the boom and the edge of the roof. FromEq. (16), the value λ satisfies Eq. (17),

λ ¼ Aμ þ B where A ¼

Xr¼x;y;z

urvrXr¼x;y;z

u2r

and B ¼

Xr¼x;y;z

ur r1Xr¼x;y;z

u2r

: ð17Þ

At this point the square of the clearance Δ2 that is, min[(P4P3)2] canbe expressed according to Eq. (18):

Δ2 ¼ min P4P3ð Þ2h i

¼ minX

r¼x;y;zvr−Aurð Þμ þ r1−Burð Þ½ �2

" #: ð18Þ

Setting the derivative of Eq. (18) with respect to μ equal to 0, oneobtains the value μ leading to the following clearance value:

μ ¼

Xr¼x;y;z

Bur þ r1½ � vr−Aurð ÞX

r¼x;y;zvr−Aurð Þ2

: ð19Þ

Inserting the value of μ into Eq. (18) provides the clearance calculat-ed from the axis of themain boom to the edge of the roof. As a result, it isimportant to subtract the radius of the boom (assumed to be cylindri-cal), since in practice clearance is calculated from the boom's envelopeto the edge of the roof. Since the relationship between the clearanceand the boom angle is non-linear, a trial-and-error procedure isemployed which consists of increasing iteratively the angle by smallincrements until appropriate clearance is reached.

3. Case study

The proposed methodology is best described through a case studywhich involved the construction of five three-storey dormitory buildingsfor Muhlenberg College in Allentown, Pennsylvania. The task was to re-place out-dated (1981) and inadequate single-level dormitory spaceunits, which accommodated only 56 students. The new three-storey,771 m2 (8300 sq ft) buildings, which accommodate 145 students,were designed by local architects to esthetically complement thesurrounding neighborhood buildings with their brick exteriors.

Each of the buildings has six apartments, most comprising onedouble- and three single-bedroom units, complete with full kitchens.

These well-built attractive buildings were assembled from mod-ules manufactured in Lebanon, N.J. by Kullman Building Corporation.

67J. Olearczyk et al. / Automation in Construction 43 (2014) 59–72

Each building consists of 18 modules which had to meet specific sizeand weight requirements. These modules not only had to belimited n height and weight to be able to pass under highway over-passes and satisfy crane capacity limitations—the largest moduleswere 4 × 17.3 m (13 × 57 ft) and weighed 3266 kg (72,000 lb), butalso had to withstand the rigors of being transported to the site.For the latter requirement, one advantage of modular which cameinto play in this project is that fabricated units are structurally supe-rior to those which are stick-built. Figs. 12 and 13 present photo-graphs of the existing single-level dormitory and a CAD model ofthe new dormitory, respectively. These new dormitories were ar-ranged using a new layout which differed from the existing one.The site elevation scenario (over 7% slope) contributed to the com-plexity of the assembly process, and the elevation differences be-tween buildings were at least 3 m (10 ft). Once the old units hadbeen demolished, the subcontractor prepared the site to accommo-date the new foundation layout and modified the access road forthe lifting crane set point area. These activities were planned basedon a specific analysis of the crane access path to the final workingspot. A temporary storage area near the construction site acted as abuffer for constant crane feed in order to lift modules in accordancewith the assigned schedule.

The second access routewas prepared to transport themodules thatwere to be lifted by the selected crane. Due to the project's rigid timeconstraints, detailed and thorough investigation was required prior todemolishing the existing units. The first investigative task focused onselecting an appropriate crane using a database originally developedby Al Hussein [1]. This database allows researchers to consider possibleoptions based on a number of parameters, including capacity, geometry,crane lift radius, and obstruction proximity. This task led to the selectionof the hydraulic mobile crane Demag AC 500-1, with a telescopic boomand a lifting capacity of 500,000 kg (600 tons). Figs. 14 and 15 providetechnical details pertaining to the Demag AC 500-1.

TheDemagAC500-1 has several possible configurations. The chosenconfiguration has a full extended and pinned boom 56 m (183.7 ft), asuperlift attached to the boom, and a superstructure with a maximumcounterweight balance of 180,000 kg (396,900 lb).

For the next task, efforts shifted to examining the project dynam-ics. As a result, it was discovered that assembling the roofs anderecting the buildings concurrently on site would cut down the

Fig. 12. Photograph o

initial duration by more than 50% (10 days instead of 21). The roofswere to be built on the ground and lifted to the top of each buildingusing the same crane that lifted the modules. This alternative wasidentified after simulating a variety of scenarios for which appropri-ate triangular distributions were used to estimate project duration.For a project that involved significant lifting, visualization was bene-ficial since it identified potential challenges (especially interferencebetween lifted modules and existing obstacles) early in the designstage. Use of the CAD model ensured accurate results and proper re-source management so as to circumvent delays in the final productdelivery.

Fig. 16 shows the generated CADmodel, while Fig. 17 gives the actu-al aerial view. Roofs, visible in Fig. 17, were assembled on the nearbytennis court and temporarily placed on the building foundations. Thisoperation eliminated the roof assembly from the scheduled criticalpath. The final roof assembly was directly lifted from the tennis courtwith an added boom extension as shown in the CAD model snapshotin Fig. 16.

Following the crane location optimization simulation, construc-tion operation optimization simulation, and final project schedulecreation, virtual 4D construction was executed for visualizationchecking purposes. Intelligent digital objects were assigned propermaterial properties and texture mapping. Concatenation sets ofseparate hoist, swing, placing, and returning operations allowedplanners to create optical representations of 4D construction opera-tions. Another advantage of running the animation several times, es-pecially in the presence of the rigging crew, was that it served as ateaching tool which allowed the crew to envision and familiarizethemselves with planned activities. The rigging crew demonstratednearly optimal productivity at the beginning of assembly, in partdue to the decreased learning curve resulting from the crew havingviewed the animations.

3.1. Methodology implementation

3.1.1. Capacity checkCapacity checking plays a significant role in selecting the optimal

crane to lift modules and place them at their final locations. Whilecapacity must be considered, in practice it is insufficient to considercapacity only, since one must also factor in assembly flexibility, and

f existing dorms.

Fig. 13. New dorm CAD model exploded view.

Fig 14. Demag AC 500-1 mobile crane body.

68 J. Olearczyk et al. / Automation in Construction 43 (2014) 59–72

Fig. 15. Demag AC 500-1 CAD model.

69J. Olearczyk et al. / Automation in Construction 43 (2014) 59–72

site accessibility, especially in relation to the ground configuration.From Eq. (1) the crane must accommodate the heaviest load lifted. Inthe case of this project, the heaviest load is calculated as follows:

max TWð Þ ¼ max LW1ð Þ þ SLW þ SBW ¼ 72;080þ 500þ 5600¼ 35;461 kg 78;180 lbð Þ:

And without spreader bar lift weight:

TW ¼ LW2 þ SLW ¼ 42;460þ 500 ¼ 19;486 kg 42;960 lbð Þ

Where:

Tw Total lift weightLw1 Lift weight1 max 32,694 kg (72,080 lb)Lw2v Lift weight2 max 19,259 kg (42,460 lb)SLw Total weight of slings 227 kg (500 lb)SBw Total weight of spreader bar 2540 kg (5600 lb).

Fig. 16. CAD mode

Figs. 18 and 19 show the typical under-hook configuration withspreader bar and load without spreader bar, respectively.

3.1.2. Crane placement and clearanceTo define the crane position at the construction site, it is necessary to

determine the crane set point coordinates as an optimization problemaccording to Eq. 2, subjected to constraints (3) and (4). Given eachmodule's corner coordinates, the corresponding set point is calculatedas the geometric center of the corners (assuming that each modulehas a right parallelepiped). In such a case, the geometric center is de-fined as the average of the eight corners' coordinates:

xk ¼ 1=8ð ÞX8i¼1

xi;k; yk ¼ 1=8ð ÞX8i¼1

yi;k; zk ¼ 1=8ð ÞX8i¼1

zi;kk ¼ 1;2;…;n:

ð20Þ

The coordinates of themodule set points are summarized in Table 1,alongwith the correspondingweights (module and spreader bar and/orslings). Note that only the x and y coordinates are shown, since themo-ments depend only on the distance of the object from the crane.

In addition, for the current project only one pick point, with co-ordinates (xp = 22.35 m (880.125 in.), 60.33 m (2375.33 in.)),was chosen for the modules. After running the optimization pro-cedure in which the counterweight swing radius was RCW =6.14 m (241.75 in.), the crane set point, which minimizes the sumof the moments while allowing all lifts to be possible, had the coordi-nates (xc = 32.39 m (1275 in.), yc = 33.76 m (1329 in.)) roundedto the nearest cm. Based on these coordinates, the clearance betweenthe counterweight and the module referred to as “3 × 2” was foundto be 4.8 cm (1.89 in.). A more practical value for the counterweightclearance may be obtained by re-optimizing the same problem afteradding to RCW the targeted clearance. As for the capacity constraint,cf., Eq. (3), a linear relationship was demonstrated between thecrane's capacity and radius (as obtained by inputting the manu-facturer's data),

Capacity in kgð Þ ¼ 0:453592 ð−1220:193−0:3048 � R metersð Þþ 218;033:537:

At this point, the results of the centroid method, the simplicity ofwhich motivated its use for this project, should be highlighted. In thecontext of this case study, the crane set point was defined by the coor-dinates xc ¼ 33:43 m 1316 in:ð Þ; yc ¼ 32:11 m 1264 in:ð Þð Þ for whichthe smallest counterweight clearance was found to be 0.81 m

l aerial view.

Fig. 17. Aerial photograph of actual construction.

70 J. Olearczyk et al. / Automation in Construction 43 (2014) 59–72

(32 in.), corresponding to module “3 × 2”, as was the case in the op-timization procedure. The shift between the optimal location and itscentroid counterpart was approximately (1.83 m (6 ft)). In order toprovide a complete overview of the optimization procedure, inTable 2 the optimal crane location for selected values of the counter-weight clearance is presented. The results in the last row of the tablecorrespond to the centroid method.

Fig. 18. Load with spreader bar. Fig. 19. Load without spreader bar.

Table 1Geometric center calculation data.

Module Spreader bar/slings Reference pointcoordinate

Module Spreader bar/slings Reference pointcoordinate

Module Weight Weight x y Module Weight Weight x y

[kg] [kg] [m] [m] [kg] [kg] [m] [m]

1 × 1 17,817 227 1.29 42.21 3 × 1 17,817 227 28.94 46.291 × 2 19,260 227 7.11 38.14 3 × 2 19,260 227 34.76 42.221 × 3 32,695 2540 6.60 43.66 3 × 3 32,695 2540 34.26 47.741 × 4 32,695 2540 8.96 47.02 3 × 4 32,695 2540 36.61 51.101 × 5 17,817 227 8.50 52.51 3 × 5 17,817 227 36.15 56.591 × 6 19,260 227 14.32 48.43 3 × 6 19,260 227 41.97 52.512 × 1 17,817 227 20.77 3.92 4 × 1 17,817 227 38.68 6.232 × 2 19,260 227 24.85 9.74 4 × 2 19,260 227 44.50 2.162 × 3 32695 2540 19.33 9.24 4 × 3 32,695 2540 43.99 7.682 × 4 32,695 2540 15.97 11.59 4 × 4 32,695 2540 46.35 11.042 × 5 17,817 227 10.48 11.13 4 × 5 17,817 227 45.89 16.532 × 6 19,260 227 14.55 16.95 4 × 6 19,260 227 51.71 12.455 × 1 17,817 227 57.85 52.71 5 × 4 32,695 2540 62.66 45.045 × 2 19,260 227 53.78 46.89 5 × 5 17,817 227 68.15 45.505 × 3 32,695 2540 59.30 47.39 5 × 6 19,260 227 64.07 39.68

Table 2Crane set point as the counterweight increases.

Counterweight clearance (xc, yc) Closest module Distance to closest module

R = RCW + 1 m (ft) 32 m, 32.87 m (1262 in., 1294 in.) “3 × 2” 0.94 m (37 m)R = RCW + 1.5 m (5 ft) 31.65 m, 32.41 m (1246 in., 1276 in.) “3 × 2” 1.55 m (61 m)R = RCW + 2.5 m (8 ft) 31.39 m, 31.47 m (1236 in., 1239 in.) “3 × 2” 2.44 m (96 m)Not applicable xc ¼ 33:43 m 1316 in:ð Þ; y ¼ 32:11 m 1264 in:ð Þð Þ “3 × 2” 0.81 m (32 in.)

71J. Olearczyk et al. / Automation in Construction 43 (2014) 59–72

4. Conclusion

The presented methodology has been evaluated based on imple-mentation on a case project. Over 90 modular units were delivered toMuhlenberg College in Allentown, Pennsylvania for the assembly offive new dormitory buildings. Three types of modules were fabricatedto be “ready-to-use” at Kullman Building Corporation's fabricationshop in Lebanon, New Jersey. On the construction site, modules wereassembled in much the same manner as a puzzle into three-storeyunits with minor hook-ups between compartments. Assembling theentire fleet of modules in a record time of 10 days required conciseand precise logistical preparation. The main aspect of this rapid con-struction relied on optimization of the crane position to ensureminimalconfiguration changes for the selected hydraulic mobile crane. Thedevelopedmethodology calculated the crane rotation point and assistedin modifying the boom configuration rather than relocating the unit.The effectiveness of the algorithms used was contingent on the collec-tion of detailed data about the construction site, selected crane, andlifted objects. Each operation was classified based on its resourcerequirements and time constraints. Three different values associatedwith this operation (pessimistic, most likely, optimistic) were devel-oped to satisfy triangular distribution of the simulation model. Theselected crane boom configuration was evaluated for all possible clear-ance scenarios involving nearby obstructions, and proper data was laterfed into the simulation engine. In particular, the engine considered thenew clearance relationship which had been identified between thecrane boom envelope and the sloped roof surface. This operation wasbased on boom clearance analysis of a flat roof obstruction surface asdescribed in the literature. Complicated mathematical relationshipshave been evaluated and analyzed, creating a new point of referencefor interested professionals. Actual construction assembly validatedthe preparation and algorithm results, and ultimately justified thisresearch as a basis for managerial decision making.

References

[1] M. Al-Hussein, S. Alkass, O. Moselhi, Optimization algorithm for selection and on sitelocation of mobile cranes, J. Constr. Eng. Manag. ASCE 131 (5) (2005) 579–590.

[2] A. Atkinson, J. Payling, B. Lucas, K. Gray, Affordable Housing Adaptation Using Mod-ular Construction— A Case Study in Applied Project Management, CIB World Build-ing Congress, Wellington, New Zealand, 2001.

[3] W. Booth, House Proud, Smithsonian magazine, Sci. Technol. 37 (10) (2007) 74–76.[4] T. Bock, Construction robotics, Auton. Robot. 22 (2007) 201–209.[5] A. Cardenas, F. Domenech, Assembly required, Am. Sch. Univ. 74 (4) (2005) 39–41.[6] J.P. Cartz, M. Crosby, Building high-rise modular homes, Struct. Eng. 9 (2007) 20–21.[7] C. Da Cunha, B. Agard, A. Kusiak, Design for cost: module-basedmass customization,

IEEE Trans. Autom. Sci. Eng. 4 (3) (2007) 107–121.[8] M.S.A. Ali Deen, N.R. Babu, K. Varghese, Collision free path planning of cooperative

crane manipulators using genetic algorithm, J. Comput. Civ. Eng. ASCE 19 (2)(April 2005) 182–193.

[9] T.C. Dolan, Instant classrooms, Constr. Sch. Plan. Manag. 4 (1) (2006).[10] Editorial Hospital, Modular Technique Builds New Hospital Wing, News, Hospital

Development (HD), Wilmington Media Ltd, July 20 2006. 3 (Published by).[11] Editorial Health, Building Systems: Modular mission, Hospital Development (HD),

April 2007. 29.[12] P.K. Mashood, C.S. Krishnamoorthy, K. Ramamurthy, KB GA based hybrid system for

layout planning of multistory buildings, J. Comput. Civ. Eng. ASCE 21 (4) (2007)229–237.

[13] J.D. Manrique, M. Al-Hussein, A. Telyas, G. Funston, Constructing a complex precasttilt-up-panel structure utilizing an optimization model, 3D CAD and animation,J. Constr. Eng. Manag. 133 (3) (2007) 199–207.

[14] Editorial Leonard, RAIC award— innovation in architecture, Leonard Avenue Modu-lar Housing, Can. Archit. 05 (2007) 28–29.

[15] Editorial Operating, Modular Operating Theatres for Cannock Chase Hospital, News,Hospital Development (HD), WilmingtonMedia Ltd, May 03 2007. 3 (Published by).

[16] O. Moselhi, S. Alkass, M. Al-Hussein, Innovative 3D-modeling for selecting and locat-ing mobile cranes, Eng. Constr. Archit. Manag. 11 (5) (2004) 373–380.

[17] Editorial Pharmacy, Modular Build for Hospital Pharmacy, News, Hospital Develop-ment (HD), Wilmington Media Ltd, August 28 2007. 4 (Published by).

[18] C.M. Tam, T.K.L. Tong, W.K.W. Chan, Genetic algorithm for optimizing supply loca-tions around tower crane, J. Constr. Eng. Manag. 127 (4) (2001) 315–321.

[19] C.M. Tam, T.K.L. Tong, GA-ANNmodel for optimizing the location of tower crane andsupply points for high-rise public housing construction, Constr. Manag. Econ. 21(2003) 257–266.

[20] H. Matsuo, J.S. Shang, R.S. Sullivan, A crane scheduling problem in computer inte-grated manufacturing environment, Manag. Sci. 37 (5) (1991) 587–605.

72 J. Olearczyk et al. / Automation in Construction 43 (2014) 59–72

[21] J. Olearczyk, M. Al-Hussein, A. Bouferguène, A. Telyas, Virtual construction automa-tion for modular assembly operation, 2009 Construction Research Congress, Seattle,WA, 2009, pp. 406–415.

[22] J. Olearczyk, M. Al-Hussein, A. Bouferguène, A. Telyas, Crane layout optimization forfive new student residence buildings at Muhlenberg, MacGregor Village in Allen-town, Pennsylvania, Muhlenberg College, Final report, August 2007.

[23] M. Al-Hussein, S. Alkass, O. Moselhi, An algorithm for mobile crane selection and lo-cation on construction sites, Constr. Innov. J. UK 1 (2001) 91–105.

[24] M. Nasereddin, M.A. Mullens, D. Cope, Automated Simulator Development: A Strat-egy for Modeling Modular Housing Production, Autom. Constr. 16 (2007) 212–223.

[25] P.L. Sivakumar, K. Varghese, N.R. Babu, Automated path planning of cooperativecrane lifts using heuristic search, J. Comput. Civ. Eng. 17 (3) (2003) 197–207