Simulation Analysis in Architecture

Model making has always been a significant tool for architects in decision making supplementing bare imagination and experience. Modelling itself has evolved through a long way. Schematic presentations such as perspective drawings and physical models have been used for a long time. Techniques for proportioning such as the modular man and golden section rules were developed with advent of modern architecture. Virtual and animated models using software such as 3DStudio are recent variants, enabling analysis of prototype with more attention to details.



A new family of modeling applications called Monte Carlo models is emerging to make possible analysis of dynamic movement of men and materials within the built environment.


Monte Carlo Simulation Models
Schematic and physical models, serve in analysis of static aspects of the building such as its form, scale and proportions. Seldom do they lend useful information on the more so dynamic parameters of design of the built environment such as the circulation of occupants and actual functions of the building. Proportions of parking lots, elevators and escalators cannot be analyzed based on the actual utility and function in hand using these techniques.


Consider for instance the task of fixing the number, capacity, operation and layout of elevators in commercial complexes. Golden rules for proportioning elevators are indeed available for simple buildings with limited and uniform traffic on centrally located elevator systems. However, there are larger projects with lifts located at different positions in the building, such that some of them operate at express speeds and some are larger in capacity than others are. Normally architects rely on experience to select a suitable mix of elevator capacity and speed. Severe drawbacks in this approach are evident from real experience of users waiting long for elevator service in some buildings and lifts that remain idle for long durations in some.


Monte Carlo family of modelling applications is now becoming increasingly applicable to such analysis. This consists of virtually modelling the proposed circulation system on a computer and allowing a randomly generated oncoming traffic to flow through the system. This model is statistically analyzed for the average waiting time of the traffic and the idle time of the system. The system is then, continually modified to strike a balance between the two.


Software for Monte Carlo Simulation
Monte Carlo modelling has been used for some time in industrial and manufacturing engineering fields. A score of general simulation software such as GPSS (General Purpose Simulation Software), SimCad, FlexSim, Arena, Promodel, Simfactory, Siman and Simula have thence been developed for the purpose. As architectural applications are relatively new to this field, virtual modeling software used in architecture, are yet to include modules that permit movement simulation within ordinary animated models. An active drive may be necessary to facilitate the same within commonplace software.


However, most general-purpose simulation packages can currently be customized and programmed for the architectural application in hand. For example, the GPSS-World software contains an example application for car parking lot at a supermarket, in its manual. Information and free download of student version of the software is available on http://www.minutemansoftware.com/. The airport lobby model is detailed in FlexSim enterprise simulation software; more details about this software are available on http://www.taylorii.com/. SimCad from Create-a-Soft Company works on combining CAD with simulation (Refer http://www.simulationsoftware.com/). More information regarding simulation works and engineering applications are available on http://www.tcdc.com/.


Technology is indeed taking us places: we are driving nearer to designs that proportion the form closely following the functions.

Application to Lift System
Carrying the foregoing involvement slightly further, an examination of the difficulties in making rational decisions concerning the lift system example reveals the following:

1. People arrive from outside the complex and wait for lifts to upper floors and people within the building wait to come to lower floors in different manners.


2. The arrival rate vary from time to time with peak values during particular timings


3. People using the lift may come in groups and may prefer to move in such groups


4. The building may comprise of different segments, say a few floors for shopping, few for office spaces and few for residential occupation. Of the customers who arrive some may have to go to shopping floors, some to offices or residences


5. Lifts may be located at various places in the building so that a person may select a lift based on his proximity to the lift when he wants to travel upwards/downwards


6. The lifts may be of various speeds, some of them not stopping at certain floors so that the traffic is split among the lifts; not all passengers wait for the same lift

The situation is indeed complicated that it is apparently senseless to pursue to find a rational solution to the problem. Randomn simulation comes to help in such conditions.

In order to make an analysis, the modeler proposes a suitable elevator system comprising of certain number of lifts, some of them with a particular capacity/speed located at various places in the building. The number, capacity, speed and layout of the lifts become the parameters of the problem. The arrival of the passengers (outgoing and incoming), the passengers selection of lift location and speed, and the destination floor are simulated by randomn number generation. (A randomn number is a number generated by a wild guess; for any number guessed, it could have been some other number as well. Then, a wild guess between 0 and 10, for example, can be used to decide which floor a particular passenger is heading to, in a ten-story building. Guesses can also be made educated using actual statistics; if it is known that around thirty percent of traffic moves to first floor only, randomn number can be so generated such that thirty percent of all the randomn number is 1). Thus as passengers arrive, if the lifts are busy, they wait and the waiting time of each passenger is recorded, while passengers keep further coming according to randomn number generation. On the other hand, if no passenger is using the lifts for some time and the idle time of each lift is also recorded. Any discrepancy in the guesses evens out as the process is carried out for a full-simulated day.

At the end of this experiment, the modeler is informed about the limitations (average waiting time of the passengers and idle time of lifts) of the initially proposed system. Based on the results of this analysis, the modeler can make changes in the initial propositions, such as changing the position, number, capacity and speed of the lifts. The parameters of the problem are hence varied and the experiment is carried out again for a fresh simulated traffic. By repetition of this routine, the lift system can be optimized. The wild guesses can be made more educated if actual statistics of arrival time of passengers, service time of lifts and time spent by customers in the building are known, to yield more accurate and realistic results.

A similar analysis can also be made for car parking lots in commercial complexes for number of parking spaces such that space is neither wasted nor insufficient. It is also useful for deciding on circulation spaces such as airport lobbies and taxiways, production and storage space in laying out factories with line/batch production and so on.