Using Quantified Risk Assessment to Inform Ventilation System Responses

author list icon

A signalling upgrade now allows 3 trains to simultaneously occupy a tunnel ventilation section. What is the risk impact in the event of a tunnel fire?


The tunnel ventilation system for the metro line described in this article was designed in the late 20th century. It provides comfort cooling and smoke control and was based on a fixed block signalling system that allowed only a single train in any ventilation section. Commercial pressures to enhance timetable capacity resulted in a signalling upgrade to train-based control (“moving block” signalling), which permits up to three trains to simultaneously occupy a ventilation section. The client wished to understand the risk impact of this change and in particular how the ventilation system should now be operated to best effect in the unlikely event of a tunnel fire. 

Parsons Brinckerhoff (now part of WSP | Parsons Brinckerhoff) performed a comparative quantitative risk assessment (QRA), using available fire frequency data, to understand the impact of the ventilation system operation on the level of risk. This article describes the work and presents our findings. 

Review of available fire frequency and consequence data 

The client had comprehensive data covering fire events on its network over the past 20 years. Of the 7,291 records reviewed, 384 related to the line on which we were working and only 18 related to the area of interest. 

Electrical arcing initiated the majority of the relevant fire events, at 175 (45.6 percent); arson accounted for 32 fires (8.3 percent); overheating equipment a further 22 (5.7 percent); and the remainder had a variety of causes, or were listed as “other/unknown”. 

The data demonstrated that the operator experiences a modest number of fire events, the vast majority of which are small events that are managed by day-to-day operational staff with minor to insignificant consequences for passenger and staff safety. 

We concluded that fires could be categorised broadly as: 

  • “Small” in-car fires (up to around 200kW) – “common arson events” using readily available materials such as newspapers and unlikely to cause a major fire; 
  • “Small” undercar/track/tunnel fires; 
  • “Large” in-car fires (greater than 1MW) – “determined arson events” involving a quantity of accelerant and sufficient to cause a major conflagration (thankfully, to date no such event has occurred on the network); and 
  • “Large” undercar/track/tunnel fires. 

The QRA analysis 

A set of event trees was developed, using known initiating events and with various possible outcomes shown on different branches. The significant inputs were as follows: 

  • Frequency of initiating fire event (small or large fire, in-car or undercar/track/tunnel); 
  • Number of trains in section (1, 2, or 3); 
  • Train reaches next station (yes/no); 
  • Ventilation mode selected (remain in comfort cooling, switch off, select optimum smoke control mode, select sub-optimal smoke control mode); 
  • Smoke control achieved (yes/no); 
  • Smoke ingress into passenger compartment (yes/no); 
  • Driver controls passengers (yes/no); 
  • Passengers remain on train (yes/no); and 
  • Protection implemented for evacuating passengers (yes/no). 

The probability of each outcome was determined in consultation with the client. Some were easy to define, such as the number of trains in a single ventilation section (33 percent probability of each possibility under new signalling system), while others required more detailed consideration, for example the probability of smoke being drawn into the passenger compartment. 

probablilities consequeces used QRA event trees, roads, tunnels, risk, fire life safetyFigure 1 – Probabilities and consequences used in the QRA event trees

The client’s own modelling team undertook computer analyses to determine whether smoke control would be achieved with multiple trains in a ventilation section. 

These analyses suggested that critical velocity1 would be met with two trains in section but if three trains were present, critical velocity would be lost at the incident train but achieved at the non-incident trains due to cooling of the smoke along the tunnel length. The probabilities agreed are shown in Figure 1 above. Four event trees were then constructed and the resulting relative risk levels were reviewed. 

Results of the QRA 

Figure 2 shows the impact on each scenario of leaving the ventilation in a comfort cooling mode (no change), switching it off, setting it to a non-optimal mode, and setting it to the optimal smoke control mode. Note that with a moving block signalling system, the following trains could be close to the train in front (around 25 metres apart). The train positions shown in Figure 2 are not intended to convey an accurate location for each train. 

ventilation responses various scenarios risk roads tunnels fire life safetyFigure 2 – Impact of differing ventilation responses to various scenarios

Shaded results show an appreciable increase in risk due to the ventilation configuration selected. The worst case outcome for a small fire was essentially the same for all ventilation configurations: operating the ventilation system gave no material benefit, regardless of the number of trains in the ventilation section. However, there was no disadvantage in using it. Therefore, since staff may not know whether a fire is “small” or “large”, the ventilation response derived for large fires was considered acceptable for small fires as well. 

For large fires, the presence of additional trains has a marked effect on likely risk level. For a single train event, there is a modest benefit in operating the ventilation system in the optimal mode (although for a fire near the centre of the train, even the optimal mode may incur a large loss of life). When there are multiple trains in a section, however, the impact of using the optimal ventilation mode offers a substantial benefit for a large fire incident, even if critical velocity is lost over the incident train. 

Conclusions and recommendations 

The comparative QRA proved an important tool for decision making. The structured event trees allowed various ventilation options to be tested and the clear outcome guided changes to maximise safety on this railway. It showed that the optimal smoke control mode gave a significant benefit for large in-car and undercar fires, with the greatest benefit when there are multiple trains in the ventilation section. For small undercar fires, using the optimal smoke control mode also gave a fractional benefit, since it reduced the tendency for smoke ingress into the incident train. 

recommended actions risk assessment ventilation system, roadsFigure 3 – Table of recommended actions

When there is one train in a ventilation section, the optimal smoke control mode should be determined based on the fire location along the train and the driver’s intended direction of evacuation. When there is more than one train in the ventilation section, trains in front of the incident train should be driven forward at low speed, out of the ventilation section. The preferred direction of ventilation should then be forward, to avoid passing smoke over the trains that follow. Figure 3 summarises the recommended actions. 

1Critical velocity – the air flow required to prevent smoke from moving upstream of the fire location.

back to top