2016_City of Dreams Hotel, Macau

�

56

TheStructuralEngineer March 2016

Project focus City of Dreams

City of Dreams, Macau – making the vision viable Emidio Piermarini EI, BEng, MEng, Engineer, BuroHappold Hong Kong Hayden Nuttall MSc, DIC, BEng, CEng, FIStructE, MHKIE, Director, BuroHappold Hong Kong Rob May CEng, MIStructE, PE, MHKIE, MHKIBIM, Associate Director, BuroHappold Bath Victoria M. Janssens PhD, PEng, Senior Structural Engineer, BuroHappold Hong Kong

Synopsis

Introduction

This article describes how cuttingedge parametric-based engineering techniques have been used to achieve the detailed design of 2500 complex steelwork connections for the exoskeleton of the new City of Dreams hotel in Macau, China. It discusses the tools, methodology and strategy employed by the engineering team to automate autom ate the difficult and time-consuming time-consu ming process of creating, verifyin ver ifying g and and docum documenti enting ng the the geometrically challenging, large-scale steel connections using finiteelement methods within an ambitious timescale of just 12 months.

An extraordinary building is taking shape in the City of Dreams entertainment resort in Macau (a Special Administrative Region of the People’s Republic of China). The 42-floor twin-tower construction incorporates an irregular-form, aluminiumclad structural exoskeleton with connections of such scale and complexity that they are possibly the most analytically and geometrically challenging large-scale steelwork connections ever to be built (Figure 1). 1) . The project for Melco Crown Entertainment by Zaha Hadid Architects and BuroHappold is under construction (Figure 2). 2). When it opens in 2017 it will provide the City of Dreams development with a dramatic landmark building to complement the existing complex of hotels

1 �Figure City of Dreams hotel – architect’s rendering

S T C E T I H C R A D I D A H A H A Z

www.thestructuralengineer.org 57

�

58

TheStructuralEngineer

Project focus

March 2016

City of Dreams 3 �Figure Structural system

2 Figure City of Dreams hotel –

current progress (January 2016)

a) Concrete cores

V O L I G A M S I M I D A V

and entertainment facilities on the “Cotai Strip”. Housed within its 150 000m2 of floor space will be a seven-storey atrium, 780 hotel rooms, suites and villas, various restaurants, luxury retail outlets, gaming and event facilities, and a sky pool. BuroHappold carried out the general structural design of the building, together with the detailed design and construction documentation of all the steelwork connections. The structural design work faced engineering challenges arising from the typhoon wind climate, seismic design requirements, complex load paths and highly irregular geometry of the building, but it is the uniquely complex problem of the detailed design and documentation of the thousands of dissimilar and irregular steelwork connections of t he exoskeleton – and the innovative methodology used to solve it – that are the subject of this article.

In structural terms, t he steel exoskeleton and the two internal concrete cores act together to provide lateral load resistance, sharing wind and seismic loads in proportion to stiffness. The gravity system comprises composite beams and slabs that span between the exoskeleton and the cores with minimal internal columns (Figure 3). There are approximately 2500 connections in the exoskeleton. The members and connections are fabricated from steel plate up to 150mm thick using grades up to S460. Many of the connections incorporate “offshore-quality” plate to BS EN 102251 in order to ensure adequate ductility and strength in the through-thickness direction. Members are generally bolted together at connections in the flat regions and site-welded in the free-form central zone and the corner fillets (Figure 4).

Structure

Methodology

The design concept for the City of Dreams hotel is of a striking exoskeleton which wraps around the two concrete cores, bringing them together with a flowing midsection featuring three irregular-shaped curved openings. Inside the building, the free-form steel framework continues, curving high above a huge atrium space that is echoed by that of the sky pool above.

With such complex and irregular geometry it was clear from the outset that traditional code-based methods and standard drawing software would not be suffi cient to design and document the exoskeleton connections. Instead, the BuroHappold team decided that the complex stress states that exist where members merge into the connections meant that finiteelement (FE) analysis was the only viable

b) Exoskeleton

c) Total system

option to verify their structural adequacy. It was also clear that standard software packages would not have the functionality required to create the construction documentation, especially for the free-


Input Parameters (variables)

Output Model Viewed In Rhino 3D

Grasshopper Definition/Script

4 �Figure Zones of exoskeleton

5 �Figure Parametric definitions using Grasshopper visual programming for Rhino 3D

form central region. To complicate matters further, since the exoskeleton would be clad in aluminium, all connections and associated plates and bolts would have to be located within the cladding zone defined by the architect. This would inevitably constrain and limit options for the geometry of the connections and necessitate non-planar solutions. Finally, the timescale for detailed design and documentation of all 2500 exoskeleton connections was just 12 months. Put another way, the team would need to complete an average of 50 connections per week. In response to this seemingly impossible task, BuroHappold drew on its expertise in parametric engineering and structural optimisation developed on previous projects, including the Aviva Stadium in Dublin, Ireland2, and the Louvre Museum in Abu Dhabi, United Arab Emirates3. Essentially, the team’s solution was to create a unique, bespoke computational approach using application programming interface (API) techniques to allow efficient pro cessi ng of the hug e numbe r of FE models required and, critically, corresponding three-dimensional

(3D) visualisation of every connection throughout. The approach allowed the engineering team to focus on the quality of the engineering solution, rather than on cumbersome data handling and repetitive number-crunching tasks, resulting in significantly faster and reliable output. As a result, the entire detailed design and documentation process was completed on schedule, in a fraction of the time that the team estimated would have been required using a more conventional methodology. Tools

Parametric design is a process in which problem parameters are defined as variables and a series of functions applied in order to find the solution(s). By varying these parameters, many variations of the same problem can be solved. In this case, the problem was FE analysis of the many and various steel connections in the exoskeleton. The modelling software Rhinoceros 3D (Rhino 3D)4 and its plug-in module Grasshopper 5 were chosen as parametric design tools for the speed and accuracy they would bring to the task. The combination allowed the team to create the geometry for a large number of

6 �Figure Design process

complex 3D forms quickly and accurately using visual programming techniques and, crucially, to make changes to the geometry by changing the parameters (Figure 5). They could literally “see what they were doing” in each step of the programming logic and in the corresponding geometry as it was being created, making the code debugging process much easier and quicker than it would have been using traditional practices. Rhino 3D was also used to model the outer surface geometry as a clashdetection study to show that the connections were within the cladding zone. Autodesk’s Robot Structural Analysis (RSA)6 software was used to create the local FE models for each unique connection type. In this context, it is worth noting that the size and complexity of the structure meant that the global analysis model for the building, which was created using MIDAS structural analysis software7, took over 12 hours to run. Hence, it was not viable to create and insert FE models of all the connections into the global model, as this would have increased the analysis time even further, possibly by three or four times. Similarly, if the models were inserted

�

60


Project focus

March 2016

City of Dreams

7 Figure Grasshopper script to identify similar connections

one at a time, there would be at least a 12-hour wait each time the team wanted to investigate alternative arrangements for a connection. The only practical alternative was to create separate “local” FE models of the connections and transfer, or map, onto them the corresponding moments and forces from the global model results file, for all 105 load combinations. To maintain the t ight programme, almost every aspect of the local model generation and analysis was automated using bespoke Visual Basic scripts that linked MIDAS, RSA and Excel with Grasshopper via their APIs. In achieving the solution to this ambitious project, the BuroHappold engineering team found themselves at the cutting edge, using the software in ways that had not been done before, sometimes working at the limits of the products’ capabilities. The team maintained frequent dialogue with all the software companies’ technical support teams throughout, which proved to be highly productive for both parties.

Strategy The engineering team’s strategy was to break this immense problem into five key steps (Figure 6). The first four months of the project were spent developing bespoke Grasshopper scripts for every step. This significant time investment was justified many times over by the huge time saving made in the subsequent analyses and generation of documentation. It is important to understand that bespoke programming, however skilled it might be, does not replace engineering expertise. Rather, it augments it by handling large amounts of data efficiently and releasing engineers to focus on optimising the design. Accordingly, visualisation, manual verification and

8 Figure Visualisation of data for similar connections

a)

�

Figure 9 Developing connection arrangement

b)

N A M N I T R E P U R

c)


acceptance were considered essential and built into the process throughout. The five steps were:

a total number of over 2500 connections, this reduced the number of unique types to about 400.

1. Identify similar connections 2. Develop the connection arrangement 3. Find and map forces from global analysis model 4. Analyse the connection 5. Generate analysis reports and construction documents

Step 2: Develop the connection arrangement

Step 1: Identify similar connections

The first task was to confirm the number of unique connection types required by identifying those that were similar, in order to reduce fabrication and erection time. With up to nine elements connecting at each node, and each element potentially having a different section shape, section size and/or curvature, this was not an easy task. To identify unique connection types, a Grasshopper script was written to interrogate the exoskeleton member geometry Rhino 3D file created previously to help build the global structural analysis model. It contained the member centreline geometry and the associated section shapes and sizes. The script was used to search this file for all intersections of centrelines (to locate the connections) and to collect and organise the relevant geometric data, such as the number of intersecting members, whether members are straight or curved, the member shapes and sizes and the angles between adjacent members. Thus organised into a programming library, the data could be easily and accurately compared to determine similarity of the intersections, allowing for cases where the geometry is handed (Figure 7). Once the unique connection types had been identified, the script visualised the geometric data from the Rhino 3D file, allowing the team to verify the similar connection information easily (Figure 8). From

Next, the principles of the connection were developed through engineering judgement based on the load paths (Figure 9a) and a Grasshopper script was created to allow the designer to rapidly conceive a connection’s geometry to meet architectural and fabrication constraints before sending the connection to be analysed. Mindful of the fabrication and erection challenges that such massive connection nodes would present, 3D study models of each connection were created using Rhino 3D to ensure that the connections could be readily fabricated. The models show the “plateby-plate” fabrication sequence for appropriate clearance at every stage, including edge distance tolerances and room for site welding, testing and bolt tightening (Fig. 9b).

The connection designs also had to accommodate extraordinary architectural constraints. Zaha Hadid Architects had provided a Rhino 3D model of the inner surface of the cladding zone that all the steel elements and connections had to fit inside (Figure 10a). For simpler connections, with little or no curvature, the connections and associated plates and bolts were similar in size to the steel elements; therefore, clashes were relatively easy to manage. However, in the freeform central zone, multiple members typically meet with high curvature, leading to complex intersections (Fig. 10b). For this reason, the connections are necessarily significantly larger than the individual members. Given the architectural envelope was not only tight but also varied in depth where it was in double curvature or warped, clashes were a real possibility (Fig. 10c). Since the constraints of fabrication would often oppose those presented by the architecture, the team realised that

a)

b)

c)

10 Figure Working

with architectural envelope

�

62

TheStructuralEngineer March 2016

Project focus City of Dreams

N A M N I T R E P U R

11 �Figure Parametric connection definition and fabrication connection

finding an optimal solution meant being able to explore the design space for each connection rapidly. To address this, BuroHappold engineers programmed the geometry of each connection using a parametric script with variables defined for all dimensions that were likely to need further study to meet architectural, fabrication and construction constraints (Figure 11). The more complex the geometry of the connection, the more complex the parametric script became, but some guiding principles were common to 13 Figure Global analysis model

multiple connections. These allowed parts of the parametric scripts for one unique connection to be copied or developed for application to others. For example, as a general design principle, a 25mm edge distance tolerance was allowed for members being sitewelded to the connections, to account for erection tolerances. However, increasing the thickness of a connection node in order to maximise edge distance for site welds would make it more likely that the connection would clash with the architectural envelope. In order to explore this, the thickness was defined as a parameter within the Grasshopper script. The value could then be adjusted until the edge distance tolerance of 25mm was achieved. Thanks to Grasshopper’s powerful visualisation, all these changes occurred graphically and in real time as the designer moved the slider value up and down (Figure 12). If the 25mm tolerance could not be achieved because of a clash with the architectural envelope (as in the example shown), the designer could rapidly determine what value would optimise the edge distance while remaining within the architectural envelope. Step 3. Find and map forces from global analysis model

With 105 load combinations and up to nine members in a single connection, the process of finding the correct forces/moments in the global model and correctly

applying them to the local connection models was significant. Once again, Grasshopper’s capability as a tool for creating and visualising geometry offered a number of benefits in terms of speed and reliability. With a script similar to that used in Step 1 to identify unique connections, data including connection geometry, bar/node numbering, section sizes and member orientations were t ransferred to Grasshopper from the global analysis model and mapped for each connection under consideration (Figure 13). The corresponding forces/moments were then also extracted. The volume of data this created was so large that it was split into 55 separate files, each containing up to five million sets of bar forces/moments. The bespoke scripts allowed designers to search for any set of forces/moments from the entire data set and instantly visualise them on screen. In-built vector transformation tools could then be used to map the forces/moments onto the local model. The task would have been much more difficult and time-consuming wit hout the powerful visualisation functionality that Grasshopper provides, allowing as it did for “visual debugging” of the script. Even with Grasshopper’s power vector tools, mapping and translation of the forces/moments from multiple files was susceptible to error, so the team used a two-step verification process comprising visual and numerical checks to ensure the extracted data were correct (Figure 14). For the visual check, the connection was displayed in 3D together with

www.thestructuralengineer.org 63 12 �Figure Using parametric definition to meet multiple constraints

vectors showing the magnitude and direction of the applied forces/moments. This quickly displayed any missing data and verified that the forces were acting in the correct direction. Additional analytical information from the global model, such as bar/node numbers, section properties, gamma angles and local axes, could be displayed as well to ensure proper mapping of bar information. For numerical verification, an equilibrium check was performed for all load combinations to ensure no “out-ofbalance” forces/moments existed. Any questionable load combinations or nodes were then displayed graphically and further interrogated. Step 4: Analyse the connection

The accurate prediction of the resultant stresses where multiple members intersect was a major concern. Consideration of even a simple cruciform example illustrates the importance of accurately predicting stresses where members merge (Figure 15). At the start of the project, the BuroHappold team had determined that neither established code-based methods nor bespoke first-principles methods would readily capture the complex stress states that exist in the many and varied connections of the exoskeleton where individual plates intersect and overlap, especially in locations where multiple plates up to 750mm wide merged into a single plate. Given the geometric complexity and sheer size of the connection nodes, an FE approach was the only viable method for verifying the adequacy of the connections. Almost every step of the connection analysis process was semi-automated

14 �Figure Visualisation and numerical check of mapped forces

�

64


Project focus

March 2016

City of Dreams

15 �Figure Interaction of in-plane principle stresses and theoretical

a) In Rhino 3D

von Mises envelope for simple cruciform connection

b) Connected live to RSA

σ1 and σ2 for the incoming members of the cruciform are set at the yield stress of the material (py). When σ1 and σ2 are both positive or negative (right-hand case), the maximum stress in the overlapping region does not significantly increase. However, when σ1 and σ2 have opposite signs (left-hand case), the maximum stress in the overlapping region reaches √3 × p . This phenomenon is predicted by inspection of the well-known “von Mises failure envelope”. y NB In both cases, the stress levels

to reduce set-up and processing time, using bespoke scripts to link the various software programs to Grasshopper though their APIs. The scripts were used to • generate the local FE model • add extension bars and apply the forces • apply analytical links and boundary conditions • run the analysis and extract results At every stage, the engineer could employ visual checks to ensure the correct data were being used. Once the scripts had been created, these local analysis models took just a few minutes to run (compared to 12 hours for the global analysis using MIDAS), allowing the team to run them as many times as they needed to, in order to match plates’ thicknesses to stress levels and optimise the connections. The FE models were based on 2D shell elements that incorporated all plates in the connection together with an appropriate portion of the incoming members. Beyond this, bar elements were added to match those in the global model and the mapped forces/moments from the global model were applied to these. Since the geometry and the forces/moments in the local and global models should match, it was easy to check these visually and numerically against one a nother. The first step was to generate the local analytical model in RSA (Figure 16). The script first created a Rhino 3D model of 2D

16 �Figure Connection model

surfaces at the centre of the plates, which could be planar or curved, and converted these surfaces into RSA objects. It then asked RSA to create the FE mesh of 2D shell elements from these objects. Since the FE mesh would be generated inside RSA, the geometry of the surfaces created in Rhino 3D needed to be of suffi cient accuracy to avoid meshing problems, which can occur when the meshing algorithms cannot determine the intended common boundary between adjacent surfaces. Since the Rhino 3D geometry was defined parametrically, the overall geometry could be altered as necessary until the connection was optimised and the various fabric ation/architectural constraints had been met. Once the 2D shell elements were generated, the script automatically added the bar elements to the model. The bar geometry was extracted directly from the global analysis model and placed in the same virtual position in the local model. As the bar forces had been mapped inside Grasshopper, and the bar numbers generated in the local model matched the global model, the load combinations and bar forces/moments could be automatically applied using Grasshopper.

This again mitigated errors associated with manual processes such as copy and pasting tabulated data. Under a conventional approach, the definition of the analytical links between the bars and the shell elements in RSA, and the definition of boundary conditions (analytical supports) would both have been time-consuming manual operations. Here, they were both scripted to happen automatically, saving considerable time for the project. The nodes of the FE mesh were imported into Grasshopper, which applied a script that used geometric search algorithms to find the appropriate nodes to which the bar elements should be connected. This information was then sent back to RSA and used to create the analytical connections. The script also automatically applied the required boundary conditions to the local RSA model in predetermined locations. After the forces/moments for all load cases had been applied, the models were batch-processed. Finally, the sum of each reaction was checked to ensure they all equalled zero before the results were prepared for extraction. To avoid unnecessary handling of large and cumbersome data files, and to speed


17 Figure RSA von Mises stress plot and fabricated connection

18 �Figure Example of calculation output

19 Figure 3D documentation for largest free-form connection

up the process, a script was developed to extract stresses in batches to determi ne the governing load cases. Stress maps of these connections were interrogated using a scale based on the maximum plate thickness for a given selection of plates (Figure 17). The stress maps were then visually inspected to establish whether the stresses in any areas were unacceptable. If necessary, the plates’ thicknesses, arrangements or grades were changed and the whole process re-run until satisfactory

results were achieved. Finally, the results were all individually reviewed by BuroHappold engineers as part of the verification and acceptance process. Step 5: Generate analysis reports and construction documents

It was recognised early in the project that, given the large number of unique nodes, the generation of engineering documentation for each connection

could be a laborious task. Since all the visual data available to the designers during the design process were created in Grasshopper, the logical solution was to transfer this to an Excel template after the analysis was complete. By creating a tool to automate this task, the team made considerable time savings and provided a comprehensive visual record of all steps of the design process, ensuring that any independent party could easily follow the assumptions made and data used for the design of each connection (Figure 18). While documentation was not a primary objective of the process, the Grasshopper scripts generated rich and coordinated data that could be easily extracted to provide accurate and relevant information for the fabricator. After careful consideration of the options, it was agreed with the contractor that the construction information would be issued in the form of 2D drawings for the connections in the flat-sided areas and curved corners, where the geometry could be readily defined using conventional drawing software, and as 3D digital models for the free-form areas to assist the fabricator in understanding the connection geometry (Figures 19 and 20). This was because the design intent for the connections in the free-form area was more diffi cult to communicate using 2D drawings. Since the 3D information was readily available, it seemed illogical to convert this to conventional 2D drawings that would have required multiple views, sections and coordinates to define the shape, position and orientation relative to the finished structure. Rather, using the Rhino 3D surface models that had already been created for the clash-detection studies, 3D models that were geometrically

�

66


Project focus

March 2016

City of Dreams

20 �Figure Example of complex 3D documentation

a) Assembly details

b) 3D setting-out

21 �Figure 3D documentation via digital model and construction

accurate in every sense (plate thickness, plate geometry, plate hierarchy at plate intersections, actual position/orientation in the building) were provided for the fabricator, who simply transferred them into their own 3D construction model. The approach was mutually beneficial as it saved time for all parties in what was already an aggressive schedule and helped to minimise fabrication errors (Figure 21). Conclusion

To meet the aggressive construction programme for the City of Dreams hotel project, BuroHappold needed to develop a state-of-the-art approach to the complex design and documentation of the exoskeleton connections. This involved full FE analysis of more than 2500

connections and 105 load cases. The whole process was run using bespoke parametric Grasshopper scripts, which successfully integrated MIDAS, RSA, Rhino 3D and Excel. Due to the number of unique arrangements, their highly irregular shapes and the co mplex stress states that exist where the members merge, the exoskeleton connections are possibly the most analytically and geometrically challenging large-scale connections of any building constructed to date (Figure 22). The Grasshopper scripts not only allowed the engineering team to process vast amounts of data quickly; importantly, they also incorp orated “on-screen” visual checks at all stages of the process to help eliminate errors. The scripts were

carefully designed to avoid being a socalled “black box” set of tools, but rather an extension of the engineer’s hand; cutting out mundane tasks and allowing more time to focus on problem-solving. The initial decision to spend the first four months of the 1 2-month programme developing the process and writing/testing the parametric scripts was a bold one, but one which paid off later when some of the connections were being created, analysed and documented in less than one hour. There was inevitably periodic updating of the scripts throughout the project, but the majority of the development was completed in this early stage. Once set up, this innovative design approach achieved huge savings in man-hours and allowed BuroHappold to consistently deliver ahead


22 �Figure Node fabrication in Guangzhou, China

of schedule. Structural engineering in the modern era is challenged by projects of increasing complexity, falling fees and faster construction programmes. The profession will not meet these competing challenges successfully without harnessing the best available technology. The construction industry is now largely a “digital” industry, with the leading design teams, contractors and manufacturers increasingly creating and sharing digital information. For structural engineers, parametric and computational design are the to ols that will enable them to embrace this complexity, avoid getting bogged down in ever-increasing amounts of data and devote more valuable time to what t hey do best – engineering.

References

�1

�2

�3

�4

3D [Online] Available at: w ww.rhino3d.com (Accessed: January 2016)

British Standards Institution (2009) BS EN 10225:2009 Weldable structural steel for fixed offshore structures. Technical delivery conditions, London, UK: BSI

�5

Shepherd P. (2011) ‘Aviva Stadiu m – the use of parametric modelling in structural design’, The Structural Engineer, 89 (3), pp. 28–34

Robert McNeel & Associates (2016) Grasshopper [Online] Available at: www.grasshopper3d.com (Accessed: January 2016)

�6

Autodesk (2016) Robot Structural Analysis Professional [Online] Available at: w ww.autodesk. co.uk/products/robot-structural-analysisoverview (Accessed: January 2016)

Shrubshall C. and Fisher A. (2011) ‘The practical application of structural optimisation in the design of the Louvre Abu Dhabi’, Taller, Longer, Lighter: Proc. IABSE–IASS Symposium, London, UK, 20–23 September

�7

Robert McNeel & Associates (2016) Rhinoceros

MIDAS Engineering Software (2016) midas Gen [Online] Available at: http://en.midasuser.com/ product/gen_overview.asp (Accessed: January 2016)

A concrete shield A serious �re can damage business. Concrete’s inherent properties mean it doesn’t burn or give off noxious fumes. Its slow rate of ADVERT thermal conductivity also provides an effective �re shield, protecting lives, livelihoods and the environment. Choose concrete for built-in �re resistance. Visit www.concretecentre.com to �nd out more.

fact

2016_City of Dreams Hotel, Macau

Recommend Documents