Myriad and Simulex

(updated  31st August 2007)

Keywords: simulating people, simulating crowds, simulating crowd dynamics, workshops

Simulex is a software tool capable of modelling the evacuation of large populations through multi-storey buildings. Crowd Dynamics Limited have integrated the Simulex system with the Myriad crowd simulation and analysis suite. We have also deployed a flagship evacuation system at the UK Financial District - Canary Wharf.

If you would like to produce density, flow, speed, delay maps from Simulex we now offer a "Simulex Analysis Suite" as a stand-alone tool.

Simply save the Simulex data files and replay in our software. Details available on request.

Myriad is a multi-purpose spatial analysis tool.

We designed a Simulex interface to read data and perform density, exposure, delay and spatial analysis maps.

Click here for further information.

Some background to Simulex.

It has been extensively reviewed and validated by experiments by several international groups. Developed by Dr. Peter Thompson, Simulex accurately models the physical presence of each person as they move through complex spaces. Simulex within Myriad produces more comprehensive analysis of emergency evacuation times and highlights potential problem areas of the evacuation strategy.

In the diagrams (below) outputs from the Simulex system, the congestion is observed at the convergence points adding to the emergency egress times and the possibility of panic is increased. During emergency egress this can lead to raised anxiety, pushing, panic and increases the potential for falling or tripping.

Screen shots from Simulex. When the Simulex system is used in conjunction with Myriad the total picture of an emergency evacuation can be investigated. We use the combination of both of these tools for education and training, for risk  assessment and designing for optimal egress strategies. The outputs from the simulation (above) indicate the typical congestion observed during emergency egress.

The key elements to the Myriad Suite for office egress is the speed at which the operator can produce results, test different scenarios and evaluate compliance with the relevant building codes for normal and emergency evacuation. When used with the Simulex software the complete picture of building performance during normal and emergency egress is available.

The results from the simulation show that it takes more than 1 minute 30 seconds to clear this floor - but the people around exit 3 are all clear within 45 seconds. This disproportionate evacuation time is critical in understanding the loading on staircases where the bulk of the evacuation time will be encountered. Queueing on stairs is another aspect of the Myriad system and, in conjunction with the Simulex software, we can assess the delays, exposure, time and total egress times for complex spaces, offer alternative layouts/modifications and quantify risks associated with an emergency evacuation.

Output from the Myriad System.

It should be noted that the different types of threat (biochemical, bomb threat, phased and directed evacuation) require different strategies. Complex floor layouts all add to the complexity of achieving successful plans and complying with the building codes and regulations. Myriad cuts through all of the complications and produces simple to understand, risk based results and pictorial analysis.

Below we have a Simulex model (left) and the Myriad prediction of the Level of Service in this specific geometry.

Simulex validation

The evacuation software Simulex was originally developed at Edinburgh University. A significant proportion of the development time was spent on collecting data for individual person movement. Human movement parameters such as rates of body twist, acceleration/deceleration, speed fluctuations dependent on inter-person distances etc. have all been collated during extensive tests at different sights in Edinburgh. Data collection and testing exercises have also been carried out at Lund University, Sweden.

Flow rate tests have been carried out for different doorway geometries and sizes, and the results have been compared to data collected from 6 different research sources, including the ‘Building Regulations’ – approved document B.

Validation tests have been carried out by staff at Edinburgh University, Lund University, Ove Arup (Australia) and University of Ulster. Tests have been carried out on the following  types of buildings:

• Department stores

• Office buildings

• Lecture theatres

• Sports stadia egress areas

• University buildings

• Mock-up building geometries for student tests

The tests demonstrate that Simulex accurately models individual movement, and hence produces realistic results when the performance of ‘group’ tests are analysed. The simulated flow rates correspond well with real-life evacuation flow rates. Reference 2 is contained later within this document.

Simulex publications:

1.   Homberg, P., “Study of Evacuation Movement through Different Building Components”, internal document - Department of Fire Safety Engineering, Lund University, Lund, Sweden, March 1997.

2.   Olsson, P.A. & Regan, M.A., “A Comparison between actual and predicted evacuation times”, Proc. of 1st International Symposium on Human Behaviour in Fire, University of Ulster, August 1998.

3.   Thompson, P.A. and Marchant, E.W., Computer Models for Escape Movement, in the Proc. of the Conference “Fire Safety Modelling and Building Design”, University of Salford, Manchester, UK, 29 March, 1994. Proceedings avail from Dr. J. Hinks, Dept. Surveying, University of Salford.

4.   Thompson, P.A. and Marchant, E.W., Simulex: Developing New Techniques for Modelling Evacuation, in Proc. of the “Fourth International Symposium on Fire Safety Science”, National Congress Centre, Ottawa, Canada, 13-17 June, 1994, Elsevier Science Publishers, London. (Refereed conference paper).

5.   Thompson, P.A. and Marchant, E.W., Computer and Fluid Modelling of Evacuation, Journal of Safety Science, 18 (1995), pp 277-289.

6.   Thompson, P.A. and Marchant, E.W., A Computer Model for the Evacuation of Large Building Populations, Fire Safety Journal 24 (1995), pp 131-148.

7.   Thompson, P.A. and Marchant, E.W. Testing and Application of the Computer Model ‘Simulex’, Fire Safety Journal 24 (1995), pp 149-166.

8.   Thompson, P.A., Wu, J., and Marchant, E.W. Modelling Evacuation in Multi-storey Buildings with Simulex, Fire Engineers Journal (vol. 56, no. 185), November 1996, pp 6-11

9.   Thompson P, Wu, J., Marchant, E.W., "Simulex 3.0: Modelling Evacuation in Multi-Storey Buildings" in proc. of 5th Int. Sym. on Fire Safety Science, IAFSS, Melbourne, March, 1997.

10. Weckman, H., Lehtimaki, S., Mannikko, S., “Evacuation of a Theatre: Exercise vs. Calculations”, Proc. of 1st International Symposium on Human Behaviour in Fire, University of Ulster, August 1998.

A COMPARISON BETWEEN ACTUAL AND PREDICTED EVACUATION TIMES

Per Åke Olsson, Ove Arup and Partners, Currie and Richards Building, 79-82 Franklin Street, Melbourne VIC 3000, Australia

Mark A Regan, Department of Psychology, University of Canterbury, Private Bag 4800, Christchurch, New Zealand

ABSTRACT

Evacuation times and occupancy movement were observed in three university buildings during a simulated fire emergency. Two of the buildings were tall buildings, which contained offices, computer rooms, libraries, study rooms, and lecture theatres. The other building was a one-storey building, which was constructed of three large lecture theatres. All buildings were relatively new and were equipped with emergency lighting, illuminated exit signs, and evacuation alarms that varied between a siren type, and a pre-recorded PA message.

Human behaviour and movement were studied visually and recorded with video cameras. The total evacuation time, pre-movement time-lags, and the non-direct evacuation behaviour were analysed. The building occupant loads were recorded, and this, together with CAD drawings, served as indata to the evacuation model Simulex. In addition to the measured data, occupant load factors recommended in the literature were used to derive input data to a second set of simulations. The theoretical evacuation times were calculated and compared with the actual recorded escape times.

It was found the Simulex can be used with confidence to simulate travel times for the buildings previously described. The pre-movement times presented in the literature for office buildings and places of assembly seemed to be very conservative in comparison with the measured time-lags. It was also found that individuals with pre-recorded PA information were faster in the completion of pre-movement activities, than those in siren alarm evacuations, though the time difference were small (<32 s). A difference in pre-movement times was discovered between the dissimilar activities performed in the enclosures.

1 INTRODUCTION

A fundamental aspect in performance based fire safety design is to ensure that the life safety of the occupants in a building is maintained during a fire. The sound engineering approach to assess this traditionally involves a comparison of the total evacuation time for and enclosure with the time for untenable conditions to exits within an enclosure. While the design calculations that are used to derive the available safe egress time are treated with more fundamental principles, it seems that the predictions of the total evacuation time comprehend uncertainty and subjective assumptions. However, occupancy evacuation models are being developed1, 2, and could be speculated to be essential tools in the state-of-art fire engineering design in the near future. 

It can be argued that evacuation models describe the travel time relatively well, but to assess the total evacuation time it is important to include pre-movement time-lags and behaviour. This may present difficulties since there are commonly no valid data available for the particular situation, and the general knowledge about pre-movement evacuation behaviour is relatively poor amongst the fire engineering community. The intentions with the research conducted and concluded in this paper  have been to examine evacuation models as and engineering design tool, and to discuss the pre-movement time-lags effect on the evacuation time. The authors have summarised the findings of three case studies performed in New Zealand and simulated the evacuation with the evacuation model Simulex, to make an comparison between the actual and predicted evacuation times.

The geometry, content, and occupancy characteristics for the buildings that were used for the evacuation experiments are briefly described. The method used to measure and study pre-movement time-lags and non-direct evacuation behaviour is presented in this paper together with a concise discussion for the findings of the evacuation experiments, which are more comprehensibly given by Regan3. The assumptions and methodology behind the evacuation model Simulex are explained as well as the presuppositions made by the authors. The findings form the comparison between the actual and the predicted evacuation times concludes the paper.

2 BUILDING DESCRIPTIONS

The central lecture theatre is a single-storey building with and intermediate floor situated on the campus of the University of Canterbury, Christchurch, New Zealand. It consists of three large lecture theatres with escape routes to final exits at the ground level and intermediate level. The lecture theatres are equipped with a siren type evacuation alarm, emergency lighting, and illuminated exit signs.

The commerce building is an eight storey building containing lecture theatres, seminar rooms, computer labs, tutorial rooms, and offices. Each floor had access to three main stairs, which are fire isolated. There are three final exits form each stair at ground level. An evacuation alarm with a pre-recorded PA, emergency lighting, and illuminated exits signs are installed in the premise.

The law building is a five-storey structure, which enclose lecture theatres, library, tutorial rooms, and offices. The fire safety measures includes a live directives PA evacuation alarm, emergency lighting, and illuminated exit signs in addition to the fire isolated stairs and final exits at ground level of the complex.

All buildings are relatively new, and are built in accordance with the current performance based building regulations of New Zealand. The occupancy load was estimated using the concept of the occupancy load factor, as presented in the Life Safety Code Handbook4, and by using visual observations prior the evacuation drill. The table below provides a summary of the approximated occupant loads of the building.

3 EVACUATION EXPERIMENTS

Table 2.1: Estimated Occupant Loads.

Note 1: This has been derived using the Life Safety Code Handbook.

Note 2: The occupant loads are based on visual observations.

3.1       METHOD

The occupants characteristics varied depending upon the building and the time of day of the evacuation experiments. The majority of all evacuees were students in their late teens to early twenties. Some of evacuees were University of Canterbury staff or contractors of varying age.

Prior to starting the evacuations each building was assessed for the positioning of the two cameras and observers. They were place in positions that allowed a good view of the occupants. Each observer was provided with a video camera and asked to record the building occupants pre-movement behaviour. Most of the evacuations involved recording behaviour through a window as this reduced the effect of the camera presence in the rooms under observation. It was important that the observer was positioned so as not to interfere with the flow of evacuees, and to avoid suggesting to building occupants that other were not evacuating, thereby interfering with the results. Each of the observed evacuations occurred between the hours of 10 am and 2 pm, when most of the building occupants were present.

3.2        RESULTS

The buildings were equipped with pre-recorded PA, live directives PA, or a siren alarm system, and therefore presented different levels of information to the occupants. The mean time for completion of pre-movement activities is presented in table 3.2.1.

Table 3.2.1: Mean pre-movement times for different buildings and areas.

Overall it was found that individuals in the evacuation with the pre-recorded PA alarm were faster in completing the pre-movement activities than those in the evacuation with the siren type alarm. The table above also demonstrates the differences in the mean pre-movement time lags between the dissimilar activities performed in each enclosure showing library having the longest time, followed by computer laboratories and classrooms.

The total evacuation times were measured for the three buildings of concern. The total evacuation time was measured from the initiation of the alarm until no occupants could be detected in the buildings. Table 3.2.2 presents the results.

Table 3.2.2 Measured evacuation times.

4 EVACUATION MODELING WITH SIMULEX

4.1 DESCRIPTION OF THE MODEL

A detailed description of the model is out of the scope of this paper, and can be found in Thompson2, 5. However, a brief outline of the model is provided here.

The occupants characteristics are defined by the user and will control body size, unimpeded walking speed, response time and the distance map to be used by the individuals or groups of people. The floor plans are segmented into 0.2x0.2 m grids, which are used to calculate travel distances, and to serve as a basis for the route finding functions in the model. The distance-maps can be specified by the user by designating the available exits for each distance-map and allocating different distance maps to individuals or groups of people.

4.2 ASSUMPTIONS

In addition to the preceding assumptions in Simulex, the following have been presupposed in the computer simulations:

• The occupants have been illustrated by 40% males, 30% females, and 30% average body types (office type). Hence, walking speed and body sizes in the simulation have been those allocated to these body types in the computer model.

• The default distance map has been used. Thus the occupants have been assumed to chose the shortest way to the nearest final exits. This may be the case, since the occupants are familiar with the building and its egress design.

• The pre-movement time has not been simulated in Simulex. The pre-movement time lags have been dealt with separately to the computer modelling.
   

4.3        PRE-MOVEMENT TIME-LAG

The pre-movement time-lag has been assessed using the available information displayed in the Fire Safety Engineering in Buildings6, and also by incorporating the measured mean time-lags. Table 4.3.1 shows the different pre-movement time-lags used. It should be addressed that the pre-movement time-lags used and exposed below neglect non-direct evacuation routes and activities.

Table 4.3.2: Pre-movement time-lags used.

Note *: The value for places of assembly has been used.

Note **: The value for offices has been used.

4.4 RESULTS

Two different simulations were performed for each building. The first set of simulations used the occupant loads that were derived form the Life Safety Code (refer table 2.1). The latter modelling was conducted with the visually determined occupant loads. Table 4..4.1 displays the travel times for each simulation.

Table 4.4.1: Simulated travel times.

Table 4.4.2 shows the total evacuation times including the Simulex results (refer table 4.4.1) and the different pre-movement time-lags discussed in paragraph 4.3.

Table 4.4.2: Total evacuation time.

5 DISCUSSION

The simulated evacuation times were similar to the measured total evacuation times when the experimental evaluated pre-movement time-lags and occupant loads were used as input for the Simulex modelling. The calculated total evacuation time was slightly more conservative (18-41 s) for the lecture theatre and the law building, but to some extent underestimated (18 s) for the commerce building. When glancing over the research result for the discussed buildings and situations, it can be argued that Simulex is valid of those areas. The evacuation times were overestimated when the occupant loads were derived from the recommended occupant load factors in the Life Safety Code Handbook with 3-100 s, though the difference was insignificant and imperceptibly on the conservative side.

There were considerable differences between the modelled evacuation times and the on site estimated escape times when the pre-movement time-lags recommended in the Fire Safety Engineering in Buildings were used. The total evacuation times were over-estimated with 159-363 s. Clearly, the pre-movement time-lags recommended in the literature comprehend quite conservative estimations. This indicates a need for further research into the psychological aspects of egress.

6 CONCLUSIONS

The following conclusions have been derived from the research conducted and concluded in this paper:

• Simulex can be used with confidence to simulate travel times for the buildings previously described.

• The pre-movement times presented in the literature for office buildings and places of assembly seemed to be very conservative in comparison with the measure time-lags. Psychological research needs to be emphased in this area to enable a more realistic egress design.

• It was also found that individuals subjected to pre-recorded PA alarm were faster in their completions of their pre-movement activities, than those in siren type alarm environments, though the time difference was relatively small (<32 s).

• A difference in pre-movement times was discovered between the dissimilar activities performed in the enclosures. Indicating that different tasks, and possibly different environments, influence the pre-movement activities of individuals.
   

7 REFERENCES

1 Thompson, P. and Marchant, E. (1994) Simulex; Developing New Computer Modelling Techniques for Evaluation. Fire Safety Since - Proceedings from the fourth international symposium, pp 613-625.

2 Galea. E., and Lawrence, P. (1997) Advanced Occupant Behavioural Features of the building-EXODUS  Evacuation Model. Fire Safety Since - Proceedings from the fifth international symposium, pp 795-806.

3 Regan, M.A. (1998) Fire alarm information and building occupant pre-movement times. MSc Thesis, University of Canterbury, Christchurch, New Zealand.

4 Life Safety Code Handbook (1997), NFPA No. 101HB97, ISBN 0-87765-425-5, National Fire Protection Association Inc, USA.

5 Thompson, P. and Marchant, E. (1995) A computer model for the Evacuation of Large Building Populations Fire Safety Journal, vol 24, pp 131-1448.

6 Fire safety engineering in buildings (1997) Part 1. Guide to the application of fire safety engineering principles, Draft for development, BSI-DD240, ISBN 0-580-27952-9, UK.