Catastrofic Fire In Tunnels

Handling Exceptions in Structural Engineering:

Robustezza Strutturale, Scenari Accidentali, Complessità di Progetto Roma, 13-14 Novembre 2008 – www.francobontempi.org/handling.php

Catastrophic fire in Tunnels A. Sacripanti

Dipartimento Tecnologie della Fisica e Nuovi Materiali (FIM), ENEA, Italy

ABSTRACT: To overcome the very difficult problems set by catastrophic fire in Tunnels in 2000 under the Prof Rubbia Patronage the ENEA built up a strategic project called FIT (Fire In Tunnel) flowed in one bigger meta-project called SITI (Safety In Tunnel Intelligence), the author, builder and general project manager of both projects, starting by a deep revision of the risk analysis philosophy, obtained a very flexible “systemic” safety approach useful for highway, railway and underground tunnels. The meta-project starting by the new concept of “comparative time dependent risk analysis”, on the basis of the experience gained in ITS and Nuclear Power Safety fields, developed through the application of integrated and modified, innovative technologies, the praxis of the “Effective Safety” new concept, useful for all kind of specific Tunnels. At the end, a concise description of main results obtained by the SITI meta-project built up by 15 work packages representing 35 interconnected sub-projects, and submitted for partial funds, to the Italian Research Ministry. The overall cost of research streams has been about 15 million Euros, assigned form the 2004 till to the end of 2007. 1 INTRODUCTION The tunnels, normally used in transportation systems, are engineering infrastructures utilized to overcome natural obstacles disconnecting territories, like rivers, mountains or little sea surface, to make easier or shorter the way of communication. The simpler way to classify the tunnel is founded on the basis of the kind of transportation system to which they are destined like: railway tunnel, roadway tunnel and underground tunnel. Connection was in all time an important function, also today in our globalized environment, the best connection makes easier the exchange of information, makes cheaper the consumer goods, makes more wide-spread the welfare. Normally from the philosophical point of view, the tunnel safety can be or prescriptive or performance based. The first is easier for national regulation administrations, and for planning management, but obviously not corresponds to the real safety for all tunnel, the second is nearer to the local tunnel safety but is more difficult to achieve. The safety in tunnels depends from the connected application of many integrated safety systems. No taking into account the very important issue: safety against the accidents during construction, (more or less one death for each tunnel km).

Different kind of accidents are possible during tunnel exercise, but accidents involving fire are the most dangerous as it is shown by the last terrible incidents involving human deaths for remember in only two years: Monte Bianco 24/3/1999 (Italy and France) 39 deaths, Tauern Tunnel 26/5/1999 (Austria) 13 deaths, Gleinalm 6/8/2001 (Austria) 5 deaths, Gottardo 24/10/2001 (Switzerland) 11 deaths. Tunnels fires have been rare occurrences, however the potential for entrapment and injury of large numbers of people who routinely use highway tunnels, or underground massive transport facilities, or mainline railroad tunnels, warrants special consideration. Normally safety systems into tunnels are, in function of their operability, and usually can be defined active or passive ones. Normally, tacking in account the specific differences, the safety strategy for major fire accidents into conventional tunnels, highway, underground or railroad is basically the same. In term of flow of time the basic steps are: 1. Detection. 2. Alarm. 3. Incident location. 4. Communications. 5. Planned responses. 6. Evacuation. 7. Smoke control. 8. Fire suppression. If we think to fire detection as wide application of thermal detectors or smoke detectors in the tunnels, also if there are thinking to the last generation of detectors, they are very limited in their capabilities.

For example smoke detectors are unsuitable in tunnel environment due to the products of combustion in the vehicles exhaust emissions. However the main problem even with the last detectors generation is the substantial lag time from the starting of fire to the detection (normally fire starts al floor level and it is detected at ceiling level), it would be reasonable to conclude that the fire has had time to increase at significant size (see in this article the fractal fire models). The problem of knowledge of location of a fire accident in a tunnel is critical to complete safety response. It is always possible to use loop detectors or thermal detectors or heat linear detectors to determine the fire first location zone, but video monitors and men could be the most accurate finder of the accident zone, because they are free of mistakes or false alarm. In this paper we describe the overall safety approach to catastrophic fires in tunnel used in FIT and SITI Projects, the improvement in risk analysis performed, and the main results obtained in prototypes and advanced studies, using an Intelligent Approach System based on ITS technologies and Nuclear praxis. 2 THEORETICAL BACKROUND – STATIC VERSUS DYNAMIC CONSIDERATION If we go, in a special kind of rarefied abstraction, looking at meaning of tunnel as link between two social and economic environments, in a more deep insight at light of mathematical topology, the land surface could be assimilate to an abstract mathematical set of points, some connected and some other disconnected, in such view it can be possible to see the tunnel like a relationship or a connecting function between two o more disconnected points. Continuing this mathematical similarity, it will be possible to describe a generic tunnel as the mathematical multi-parameters function, that links two disconnected points or two disconnected areas in the land-space under consideration. But as many parameters we use, each couple of points or areas will be obviously different. It means, that all tunnel could be described by the same general function, but every specific tunnel, in this space, will differ in each connection, for his numerical value. In this very abstract approach, built on disconnected points or areas into the Environment (the system), and a special connecting function describing the “Generic Tunnel”, it easy to understand that, however the general function is similar, his specific numerical value will be different for each two connecting points. The previous clear understandable rule could be applied, in the same way for the safety systems into the tunnels: a general view common to all tunnels,

but for each tunnel the specific configuration of the application of safety systems will again be different. Even more, if in the general function describing the “Generic Tunnel” are considered not only static or geometric parameters, but also the dynamical ones, then it will be possible to introduce, in very natural way, the concept of “Dynamic Tunnel” as the time dependent form of function describing the Generic Tunnel. The conclusion of this speech is that: from a more philosophical, than operative point of view, it is possible to tell that a specific tunnel is never equal to itself. If we remember the variational operator used to denote a change in a given quantity then let u=u(t) be the configuration of a given tunnel system at time t, an admissible safety configuration of the system is of the form:

u

u

(1)

v

with small. Therefore, for any fixed t: v can be viewed as a change or variation in the actual evolution u, this variation is often denoted by . u v next consider a tunnel function of t, of the dependent variable u and its derivate u’. F(t,u,u’) its variation for fixed t will be: F

F v u

F v u

F u u

F u u

(2)

From the previous reasoning, it accrues that every “Dynamic Tunnel” changes his numerical value during the time and then, for this reason, the safety system must be changeable in time for a correct application of the previous “abstract” safety concept. In general terms, the safety in this abstract approach could be see as an operator minimizing some values of the generic tunnel function, however in term of real applicable safety to a specific dynamic tunnel, the safety will minimize some variable values of a time series decomposition of the specific dynamics tunnel function. The first meaningful extension of this analysis was that it allows the complete parallel approach among highway, railroad and underground tunnels. This comparative view showed clearly that the classical safety vision utilized so far must be improved, in the light of new knowledge obtained by this comparative analysis. A more deep study enables us to understand not only the well known differences regarding highway, roadway and underground tunnel safety, but also unknown similarities and helpful methodological interconnections. In many countries, until now, safety against catastrophic accident, as fire in tunnel, is: a matter of “passive” approach to the problem; a collection of separate view of problems among road, rail and underground tunnels; a problem very far from the ap-

plication of more integrate and innovative technologies, normally applied in the advanced transportation fields like ITS or in other structures, safety advanced, like Nuclear Power Plant; as far as we know, a vision restricted to the Tunnel structure boundaries only. 3 THE SAFETY STATE OF THE ART a. HIGHWAY For long-one tube tunnel, the last word on safety is certainly the Monte Bianco equipment structured system. The technological apparatus of the tunnel increased until the actual 41 video cameras for traffic control, 116 safety boxes (with visual variable messages), 72 emergency places (every 300 m with phone and extinguisher ), 15 turning-bays (every 600 m allows rescue vehicles to operate in the tunnel), 116 new aspiration ores, and fresh air ventilation on one new underway emergency exit, two control room a computerized help for emergency, three special "Giano" fire brigade trunk, posted 24 h a day, (one in the middle of tunnel the other two at the opposite extremities) (Alamberti 2001). In Japan, some time ago, tunnels were categorized in five levels, in term of dangerous structures and decreasing severity class: AA, A, B, C, D, and “….the categorization is determined in such way, that a probability of one accident every 22 million which is ensured, regardless of tunnel length and traffic volume…". As matter of fact, it means to consider one unified safety approach providing a nearly equal safety level, for all tunnels! In Japan many further studies have been performed also about evacuation systems. From the experience of earthquake, vantages and drawback of principles, utilized in the system for people evacuation, are analyzed in order to find the best way to obtain easier exit, and to lower time-queues, very dangerous during fire accidents. b. RAILWAY The most advanced safety organization for railway tunnel is, up to now, in the opinion of the author, the systems utilized in the undersea Euro Tunnel. As it is well known, this tunnel was designed with a service tunnel between the north and the south running tunnels. After the 1996 fire accident, inside safety systems have been improved much more. Fortunately, this accident occurred with no deaths or seriously injured, but the damages were very high, nevertheless the very big emergency organization involved: about 450 firemen, from France and England, who exhausted at least 250 oxygen portable reservoirs (Brouke 1996). In spite of this very effective and rapid safety response the Channel Tunnel was closed no less than 15 days for restoration. No deaths also in the last fire

accident (lorry on freight shuttles) September 2008 the tunnel was closed for few days. Except for one fire (Hokuriku 1972) all the fires with multiple fatalities occurred when a train was forced to stop in a tunnel. c. UNDERGROUND The underground safety system could represent the most serious and complex situation, especially for the evacuation problems from the tunnels or the stations remembering the 1100 people in Philadelphia in Sept.6 1979. Among other serious accidents, it is worth to mention the fire in the Metro of Baku in 1995 (289 people killed) and the fire on the cableway to Kitzsteinhorn, Austria, in 2000 (155 people killed). Both tunnels had relatively small cross sectional area (Kitzsteinhorn 10 m2 and Baku Metro 28 m2). In the annual of statistics it is possible to find that between the years 71 -87 it was possible to count around the world about 30 significant metro fires, including also the tragedy example of the King' s Cross Underground Station fire (1987), with 31 passengers killed in the hall of the station. 4 RISK EVALUATION Quantification of risk is the main way to perform a good safety approach. Risk analysis started in the US for the nuclear power plants. This methodology was also applied to road and rail, in E.U. especially in the northern countries. In Denmark, it was applied to both road and railway about in 1970. In Germany in 1983 was edited the Safety Concept for Tunnels in New Railway, applied to the new high-speed rail. One study on the risk (H j 2002) concerning the dangerous goods through the tunnel was prepared by (OECD-PIARC-E.U.) in 1999, the total risk of open route versus tunnel route was evaluated as usual (frequency /yr - fatalities). In Italy, until now, PRA is very far from tunnels, roads or, other transport systems, but an E.U. Directive 2004/54/EC, on minimum safety requirements for tunnels in the trans-European road network, was published in 2004. The comparative analysis and the study of the state of the art of tunnel safety allowed us to introduce and define the concept of “Dynamic Tunnel”, a point of view very different from the generic unified safety vision, normally utilized by engineering groups or corporation. Starting from the obvious consideration that: each tunnel is different from the other ones’. This underlines the incorrect vision, previous introduced, of the “Tunnels Unified Safety Criteria” based on the concept that we called “Static Tunnel”. In general we may argue that for some parameters one tunnel is never equal to itself, in other words, each tunnel has its own degree of risk, not constant

in life, but variable during the daytime (Pacilio, Sacripanti. 2001). This statement shows the dynamic evolution not only of the simple “Tube” but also, that ones of the connected neighboring, the time dependence of the kinetic parameters, and also the importance to have a systemic vision of tunnel. Would this approach be useful to Safety, both from the theoretical and operative point of view? Substantially it is! At first, it underlies the importance of the dynamic evolution of the whole system, and the role of the kinetic parameters. Then the “Dynamic Tunnel” usefulness goes through two steps: a. a systemic global vision and the Tunnel System individuation; b. a parametric tunnel description and the application to the specific tunnel parameters: i of proper weighting functions; ii of proper time intervals. The birth of “Dynamic Tunnel” concept let us to introduce the time dependent risk analysis. And the use of time dependent risk analysis applied to innovative technologies breeds the concept of “Effective Safety” for each Tunnel in is System. Until now the tunnels are described by “Static” parameters as length, width, height, and historic numbers of accidents and so on. Now in the “Dynamic” view, the tunnel description will be extended by new random parameters, changeable in time, as: intensity of traffic flow, ageing of the system structures, etc. Of course, this tunnel “dynamical” description is connected to the stochastic properties linking the risk state to a straightforward time dependence, (i.e. in the “dynamic” roadway tunnels description, it is necessary to bring in new parameters such as “dynamic probabilistic occupancy” (Pacilio, Sacripanti 2001), hazard rate functions, etc. (Hensher 1994). The most important parameter in term of “probabilistic occupancy”, for road tunnels, is the passenger number connected to car number point flow, while, for railway and underground tunnel, this parameter is again the passenger number connected to van number flow. In time-space, a Poisson discrete distributions describes road traffic phenomena, for which the average probability of a passage is constant in time and independent of the number of previous events, (Pacilio, Sacripanti 2001) in this situation, the order of events cannot be interchanged Baccelli, Hasenfuss and Schmidt (Leitner 2001) give some probabilistic conditions based on coupling, for a Poisson point process; these authors connect these conditions to the queuing theory in a homogeneous Markov process. In fluid queues with several merging inputs, stochastic integrals are used for describing the total

number of cars arriving during a time interval of random length. If we pay attention to the probability of a traffic accident in a tunnel, the time dependent hazard function rate could have the special form derived by the Weibull three parameters distribution. This function could describe traffic phenomena for which the accident is a random occurrence. Thus the related hazard rate function (t) would be decreasing for safety systems or increasing with a snowballing effect for the rise of cars dynamic probabilistic occupancy in the tunnel. (t )

1

t

(3)

This last tragic occurrence will be verified if, regarding the Weibull parameters, the following conditions are satisfied: >1;

>0; 0

t

(4)

The most general models of hazard rate functions are the Multivariate Conditional Hazard Rate Functions. With these general functions it is possible to introduce a hazard rate ordering, among some “dynamic” parameters, by introducing the notion of positive dependence for dependent random lifetime parameters. This notion is useful for ordering kinetic parameters by special weights in order to obtain the most important parameters that contributed to the Dynamic Risk evaluation. 5 EFFECTIVE SAFETY IN DYNAMICAL TUNNELS As it is shown by the time dependent risk analysis it is not possible to consider the tunnel as a simple tube, for example in a mountain. In the new dynamic vision, it is vital for highway tunnel analysis to pay attention to the connected route system before and after the tunnel. The availability of access for enabling support to be available early and the capability of rapid evacuation are fundamental components in train accident emergency planning. However, it will be useful to consider these aspects also during the design phase of a tunnel project as part of the system safety planning organization. The systemic vision underlies the same problems for the underground. The system for the underground by the time dependent risk analysis is formed not only by the upper road system (for intervention), but also by the complex incoming of the tunnel, station and the out-coming of the tunnel. In addition, also in this case it is easy to understand that project engineering must take account of these problems in the early phase of the new underground project. Safety is driven by two principles: a. Minimization of technical breakdowns;

b. Minimization of human error. But it is worth to remember that it is not possible to improve a safe system like a Nuclear Power Plant over the value of less one accident per 10 million (in term of safety unity- like running time, passenger per km and so on). Achieving a total safety beyond this minimum quantity of accident is practically impossible. The solution of the overall problem depends on the global safety of the complete system. Then, when safety of one part of one old system is improved it means that you must implement this by an addition, rather than an optimization. These actions make the system increasingly complex. Really this means that the old system tends to be ageing or over regulated, in one world “rigid”, and the accident very often results from a combination of factors none of which can alone cause an accident, especially a serious one. Therefore such combination of “marginal” situations is difficult to detect or to recover using traditional analysis, (Alamberti 2001) (remember the burning role of butter in Mont Blanc Tunnel fire) (Fig. 1–2).

Figure 1. View of Mont Blanc Tunnel after the destructive fire.

Figure 2. View of Mont Blanc Tunnel after the destructive fire.

6 THE “EFFECTIVE SAFETY” CONCEPT In this new approach, there are neither tracks of the unified categorization of tunnel risk, independent from his one’s length or his one’s traffic volume, nor of classic risk analysis result like a certain number of death every year and so on. In effect, the use of time dependent risk analysis added to ITS technologies (Brouke 1996) breeds the methodological concept of

“Effective Safety” for each tunnel in his own system. For example it seems obvious that a short tunnel does not need the same safety equipments of a long one, but what about a long connected series of short ones? The concept of “Effective Safety” is a flexible methodology applicable to every tunnel in his system situation. It comes from different linked sources, like the concept of “Dynamic Tunnel”, derived by the time dependent risk analysis, the studies about the structures ageing or the response capability against fire accident, the application of the proper technologies, the optimization of emergency response, the integration of the very special system conditions to the tunnel conditions, (for example: if a serious accident happens in one short tunnel very near to one other short tunnels connected by a roadway like: “Autostrada dei trafori” in Italy, under particularly weather condition it is possible to have the maximum smoke concentration in the neighboring tunnel, than in the accident tunnel), the consideration of different time scenarios, etc. This flexible methodology is expressed by the following three time steps, considering for example the proper road tunnel system description: entrance roads system-Tunnel – exit roads systems. a. Prevention: i Interactive safety between tunnel and aware vehicle (see SITI overview); ii Active safety between intelligent tunnel or control robot (for short tunnels) and all vehicles; b. Intervention: Fire detection and suppression systems; c. Synergetic: Emergency optimization: by interactive Virtual Reality, intelligent decision support systems and emergency operators. The division in three logical and connected time blocks of the safety measures, allows us to approach the risk in more flexible way for each kind of tunnel, and with the ITS technologies utilized it is possible also to minimize the heavy time dependence of the problem, but it is important to remember that a chain is no stronger than its weakest link. The main seven technological categories for the ITS, that will grow in the next future are connected with the increasing complexity of traffic, coupled with advances in computational methods and computer architectures. They are: Decision Aids, Computer Vision, Virtual Reality, System Evaluation, Vehicle Control, Traffic Control, and Commercial Operation. The comparative safety evaluation among different tunnels, considering the Jack Linch Tunnel in Ireland; the Oslo Tunnel, in Norway opened in the January 1990 with three hierarchical orders of control, and the Ted Williams Tunnel in US monitored and controlled by the Integrated Project Control Sys-

tem (IPCS) formed by a central unit and nine workstations with operators; force us to observe that “Effective Safety” for roadway claims that road tunnel needs to become “ intelligent”, with first target car safety, and indirect primary target people safety. Instead “Effective Safety” shows us that for rail and underground tunnel, such kind of approach is unsafe, it is more effective in this case, that: train becomes “intelligent” and not tunnel. With target train safety and indirect primary target people safety. 7 THE RESEARCH ORGANIZATION We show the complete expansion of the connected organization of the SITI Cluster Project for the 35 research activities projects: Underground diagnostic system Intelligent roadway Tunnel system On site follow me system Human Factors pre and post accident. “Aware Trunk”, intelligent on board control system both for engine status and dangerous goods Training organization systems Classical risk analysis Classical Risk monitoring Source term from different materials 3D smoke propagation study Thermal qualifications of tunnel materials Thermo structural superstructure models Underground coaches project principles Emergency simulation systems Queues management simulation systems Emergency lighting systems Optimization of drop diameter Optimization of water droplet heat exchange between air-water and concrete Intelligent training systems Control and / or emergency robot Interactive Virtual Reality Tunnel system model Intelligent decision support system Instrumentation for infrastructure Time dependent risk analysis Application to individual and collective risk Cost benefit analysis Validation risk analysis Validation RAMS Reliability of electronic safety system Fractal flame interaction model In-field operation with advanced on fire suppression system. Layout study for a multipurpose laboratory tunnel In-field experimentations for prototypes

8 THE MAIN RESULTS The SITI project was organized in 15 major work packages; the most important systems are in the work package N°1. In it there were the roadway tunnel intelligent control systems applied to an Italian case study and the rail/underground control systems also applied to an underground station, both ones under patent pending (H j et al. 2002). Road tunnel intelligent control, it works for an open tunnel system like freeway or highway tunnels, and it is focused on traffic flow (Messmer et al. 1995, Smith et al. 2002) control by television outside the tunnel. This system is able to detect also driver behavior and to forecast the accident possibility in such way: the traffic flow outside the tunnel is under a control unit system which assesses an optimal car speed and separation distance, depending by the traffic conditions, by an integrated system of variable messages; in the tunnel the car flow is controlled by a connected “follow me system” with a changeable light (green, yellow and red) based on the best distance and optimal speed allowed in the tunnel in the specific time and flow situation among the cars. In the following figures (Fig. 3-4) there are shown the innovative system aiming at monitoring and steering vehicular flows nearby tunnels, based on the integration of IC, video, agent and soft computing technologies.

Figure 3. Hierarchical Fuzzy Controller, architecture.

Figure 4. Cyberspace projection, of the controlled infrastructure.

The hard real-time requirements of the applicative context impose that the goal being achieved through

a parallel Hierarchical Fuzzy Controller (HFC) implemented by means of a Fuzzy Control Agents Network (FCAN) allowing the system to gain higher performances (Galdi et al. 2008). Traffic Incident management, is the primary tool in minimizing the impact and reducing the probability of secondary incidents. It primarily includes incident detection, verification, response, clearance and recovery operations. Besides these operations, traffic management in the post-accident scenario is a crucial step to minimize the negative consequences on network efficiency and safety. Traffic management includes the dissemination of information to drivers and the activation of proper traffic control measures at the incident site and on the roadway infrastructures affected by the traffic incident. In SITI was developed a DSS (TRIMTraffic and accident management) designed to assist traffic control centre (TCC) operators to effectively and safely mitigate traffic congestion associated with incidents in tunnel and on the surrounding road network. In particular TRIM was designed to support on-line and off-line tasks of TCC operators required to define, implement and control the appropriate traffic response plans on the basis of the incident severity level, the predicted duration, the estimated delay and extension of the impact area and the current/predicted traffic demand approaching the incident site (Valenti et al. 2007), as shown in Fig. 5.

ent source term materials (like butter, plastic and so on) by heat and smoke production, these data are the input in one original software program of ‘fire flame’ which calculates the interaction among fractal front flame, concrete and the geological structure of the tunnel system. The outputs of the program are connected to a CFD smoke program that provides the concentration of smoke in the scenario; all these programs are connected to a Virtual Reality model of the specific tunnel system, for example in the underground (tunnel, station, tunnel) to produce a number of web based bi-dimensional maps (at ground and eye-level), with the time evolution of smoke concentration and heat field, that are ready to be sent to the fire brigade for the emergency strategy optimization. In off line approach the same results can be utilized for architectural optimization of smoke control strategy or evacuation of the underground station. For the roadway tunnels, smoke influence on the jet fan ventilation system at first on the basis of Memorial Tunnel data are analyzed and then applied to an Italian case study (Messmer et al. 1995, Smith et al. 2002). Figures 6-7 shows the jet fan arrangement and a time evolution of heat field and smoke.

Figure 6. Jet Fan .

Figure 5. TRIM: General Architecture.

Underground intelligent system, it is related to rail and underground operations. The system is connected with different kinds of sensory signals which come from the train; such signals discharged, on board, in a special mass memory are connected by wireless with a receiver in each station. Each receiver is connected to a central unit in which a special genetic algorithm gives forecasting information on possible system failure. Advanced in mathematical models, this group of important studies belongs to different work packages but the aim is to characterize unitary mass of differ-

Figure 7. Smoke evolution in time.

Materials special studies, other packages are connected to the experimental study of the thermo structural behavior of materials in the Tunnels (concrete, rail, coaches and trunks materials, etc.). New suppression systems, experimental studies are performed on the optimization of water drop diameter (Fig. 8) and on the optimization of heatexchange between tunnel concrete and water drops

for cooling the tunnel wall (Fig. 9) to prevent violent concrete spalling (Fig. 10). Drops generation system

cohibentation (insulating low-density briks and ceramic-fiber blanket)

Air inlet

heated low-density bricks Air outlet Nozzle

gas heater

Thermocouples Glass window

electrical insulator

Water outlet

Figure 8. Layout of drop diameter experience. PARETE RISCALDATA

ACQUA

T T ELETTRODI DI RAME

VIBRATORE

T

T

T VASCHETTA DI RACCOLTA

Figure 9. Layout of cooling wall experience.

Figure 10. Violent Concrete Spalling in tunnel.

These researches are the means to test a new kind of suppression fire system. Underground Fractal Fire and Explosion; normally CFD (Sing et al. 2002, Middelham 2001) simulations are utilized to study both the evolution of: The fire process and The movement of buoyant smoke stratified under forced ventilation action. This strategy to create a safe route upstream clear of smoke for evacuation and fire fighting, utilized in all tunnel emergency plans, is normally known as prevention of back-layering. The first assessment is very sensitive to several different parameters, for example, heat input rate, com-

bustion model, fuel type, quantity and type of combustible material, and last but not least, a good turbulence model description. The second significant parameter is the presence and the dimension of the upstream turbulent layer, and the downstream stratified layer, these parameters is sensitive for example to the slope of the tunnel to the inclination, to the roughness of the wall, to the convective and radiative heat transfer. The problem is to know the so-called “critical velocity”, that is, the minimum air velocity to prevent the smoke spreading. There are a few uncertainties in the current methods of prediction of the critical ventilation velocity. Certainly the first is the influence of the firepower; the second is the influence of the tunnel geometry. Experimental data show the influence of the cross sectional geometry of the tunnel on the “critical velocity”, it was also clearly shown that the “critical velocity” has two regime of variation, at a low rate of heat release and at a high rate. In the first, the “critical velocity” varies as onethird power of the heat release rate; in the second, it is independent of it. With reference to two critical problems, fire temperature and downstream stratified layer dimension, good model approximations are affected by an error difference with experimental measurements that can range from 20 to 30%. However, if we consider the quantitative evaluation of the physical part of the model as natural convective and radiative heat transfer, or wall roughness and turbulence models the overall uncertainties would be around the 40- 50%. That is the best situation, in my knowledge, that physics could promote, until now, in computational fluid dynamics programs. In our project were modeled both fire and explosion for an underground system like tunnel +station +tunnel, with the improvement of the fractal form of the flame interacting with the structures. This improvement changed dramatically the results of the saturation time (Fig. 11 a,b), respect the fire with buoyancy and heat radiation normally used. In Fig. 12, it is shown one of the explosion results.

800 780 760 740 720 700 680 660 640 620 600 580 560 540 520 500 480 460 440 420 400 380 360 340 320 300

6

4

2

0 0

2

4

6

8

10

12

a

14

800 780 760 740 720 700 680 660 640 620 600 580 560 540 520 500 480 460 440 420 400 380 360 340 320 300

6

4

2

0

0

2

4

6

8

10

12

14

16

16

b

Figure 11: a. Buoyancy and radiative heat; after 60 sec. b. Buoyancy, radiative heat and fractal flame; after 60 sec.

Figure 12. Explosion in underground tunnel

Short tunnel control systems, other research projects are focused on a very cheap wire-guided control robot, for short connected roadway tunnels able to control and give also indication on accident scenarios like heat field intensity, dangerous smoke component and maximum smoke concentration, that in some situation, depending on the weather condition the fire could be in one short tunnel and the maximum smoke concentration in the previous or successive tunnels, under patent pending (Italy has the 97% of the European short road tunnels). Underground people number evaluation and evacuation, other studies are been focalized on the capability to find the number of people in very complex multi connected underground station, and with the consequent evacuation problems, in Fig. 13 there is an underground station example and in Fig.14 some evacuation studies from the station.

Figure 13. Underground Type Station.

Figure 14. Evacuation study.

Risk analysis and evaluation, further studies are connected to the global risk evaluation as Classical

risk analysis, Classical Risk monitoring, Application to individual and collective risk, Validation risk analysis and the new Time dependent risk analysis of the tunnel, in connection there is also the global Cost benefit analysis to evaluate the price impact of these innovation on the Italian Tunnel system (Williams 1999, Williams et al. 1997, Webby 1996, Krzysztofowicz 1999). The mathematical theory has been applied to “Dynamic Tunnels” flows, in a general way, through the concept of tunnel Availability and Reliability (Zang 1998, Smith et al. 2002, Giuli et al. 2004, Jeffrey 2004). The Tunnel availability A(t) is defined as the probability that the tunnel is operating in his system at time t. If the cars, in Tunnel system, are independent variables, it is well known that the tunnel availability as function of time can be calculated by using the Boolean theory. Let Aj(t), (j = 1,…, n), be the availability of the cars in tunnel system and n representing the number of cars in this tunnel system. Then the tunnel availability could be evaluated by ( ( ) = Boolean system function): A j (t ) ( A1 (t ),...., An (t )) For Tunnel systems the system transport theory (Dubi et al.) could be applied in a favorable way. Now, for simplicity, we present the basics of the theory for a single parameter. For the tunnel availability calculation, for non exponential distributions, as suggested by Dubi & Gurvitz, a general state equation can be used (Giuli et al. 2003). In contrast to the Markov model, we are interested to the entering into a state i at time t as main focus instead of to being in a state i at time t, this approach based on time evolution clarifies that we speak about non Markov models. Dubi & Gurvitz define the number density of entries into a state i at time t, per time unit, denoted by i(t). Let t be a time interval and Ni(t, t+ t) denote the average number of entries into state i in the interval (t - t, t + t), the event density can be defined as: i(t

) = lim

t® 0

N i ( t ,t + t ) dN( t ) = t dt

(5)

Another important quantity is the ‘state reliability’ Ri(t - t´), as the probability that the tunnel will remain in state i without accident up to time t, given that it will be happen an accident and the tunnel entered the state at time t´. The mathematical description is based on the (car or people) probability density function f(t) and the conditional density distribution F(t), with f(t)=dF(t)/dt. In this approach the tunnel reliability will be described by R(t)= –F(t) and the tunnel failure rate or tunnel hazard function by (t)=f(t)/R(t).

The usual description of the failure behaviour of highway dynamic tunnel systems or rail dynamic tunnel system could be provided by the threeparametric Weibull distribution, with the following probability density function:

f (t )

b

t t0 (T t0 ) T t0

b 1

e

t t0 T t0

b

,t

t0

0

(6)

Where b = shape parameter; T = characteristic lifetime (scale parameter); and t0 = failure-free time. It is obvious that the failure rate of a Weibull distribution is a function of time which could depict all three sections of the well-known bath-tub curve depending only on the shape parameter. Considering the tunnel probability to have a serious accident or not, as a stochastic point process with two possible states; let one state be numbered as 1 and called ‘operational’, and the other numbered as 0 and called ‘failed’. With these definitions it is possible to introduce the state indicator function: 1 if thetunnel is operating at timet

S (t )

0 if thetunnel has a serious accident at timet Let the probability density function f(t) describe the transition from 1 to 0 (failure) and the probability density function g(t) the transition from 0 to 1 (repair). Remembering that, the event density before defined is:

Ni (t , t ) dN (t ) (7) t dt Due to this definition, it is clear that i(t´)dt´ is the probability that the system will enter state i at an infinitesimal interval dt´. Therefore i(t´)Ri(t - t´)dt´ is the probability that the tunnel is in state i at time t conditioned upon entering at time t´ into the state. Integrating over t´ yields to the probability of being in the state i at time t. i

(t )

lim t

0

t

Pi (t )

i

(t ' ) Ri (t t ' )dt '

(8)

0

Consider the two time t and t´(t´ < t) and let i and j be two different tunnel states. Using the ‘state reliability’ in combination with the transition rate ji(t - t´), the expected number density of entries into state i at time t resulting from former entry into state j at t´ is given by: dP*= i (t’)Ri (t-t’) ij (t-t’)dt’. This equation describes the following sequence of events: The tunnel enters state j at t´, remains in that state up to time t and then transfers into state i. The total event density is obtained by integration of this equation over t´ and summing over all states j i. This leads to the general state equation fulfilled by the event densities:

i (t )

Pi 0 (t )

n

t i

j i 0 j i

(t ' ) R j (t t ' ) ij (t t ' ) dt '

(9)

Where Pi0 is the probability that the tunnel starts at t=0 in state i and (t) is the Dirac delta function. The general state equation, however, is extremely complicated and provides an analytic solution to just a few simple cases. Even numerical solutions are rather difficult. But the mathematical framework is sufficient to be useful in the application of dynamic tunnel concept. 9 CONCLUSIONS The main theoretical and useful goals by the time dependent risk analysis, is to find tunnels’ potential “weak points” during their life time. The use of finite elements analysis, CFD models, Virtual Reality models and the in field experiences would increase the interconnected knowledge on serious fire accidents in tunnels. From the practical point of view, the readiness of many prototypes and intelligent systems, it introduces really practicable solutions, not very expensive, (if it is possible) to minimize the fire accident probability and to improve the emergency response. The main lesson achieved by this meta project is that the correct view is a systemic time dependent approach to the tunnel, thing that give back more flexibility to the overall safety organization. We hope with this global effort to go further in a real ambitious plan: to rationalize the matter in comparative way, and simultaneously to increase rescue people from the fire, trying to minimize the severe structural tunnel damages derived by catastrophic fire accidents. REFERENCES Alamberti, R. 2001. The Paradoxes Of Almost To-tally Safe Transportation Systems, Safety Science 37, 109-126 Baccelli, F., Hasenfuss S. & Schmidt V. 1997, Tran-sient And Stationary Waiting Times - Linear Sys-tems With Poisson Input, Queueing System, Vol.26, 301-342 Brouke, M., Varaiya P. 1996, A Theory Of Traffic Flow In Automated Highway Systems, Transp. Research C, Vol.4, 181-210 Galdi, V., Loia V., Piccolo A. & Mario V. 2008, Computational Intelligence And Agent Paradigm For Intelligent Tunnels Design, World Academic Press, Uk.Vol.2, No.2, Pp.113-122 Giuli, G., Giorgiantoni G. & Zampetti P. 2003, Analysis Of Fire Test Measurement In An Ex-perimental Tunnel, Tmi, 9 Giuli G., Giorgiantoni G., Zampetti P. – Data Analy-sis Of Natural Ventilation In A Fire In Tunnel - International Conference “Tunnel Safety And Ventilation” 19-21 April 2004, Graz, Austria Hensher, D.A., Mannering F.L. 1994. Hazard Based Duration Model And Their Application To Transport Analysis, Transport Review, Vol 14 63, 82 H j, N.P., Kröger W. 2002, Risk Analysis On Highway And Railroad In E.U, Safety Science 40, 337-357

Jeffrey, A. 2004, Bergamini; Preliminary Research Survey For A Decentralized Multi-Agent Its For Highway Congestion Avoidance; California Poly-technic State University, 3 Krzysztofowicz, R. 1999, Bayesian Forecasting Via Deterministic Models, Risk Analysis, Vol.19, N° 4 Leitner, A. 2001, The Fire Catastrophe In The Tau-ern Tunnel: Experience And Conclusion For The Austrian Guidelines, Tunnelling And Under-ground Space Technologies 16, 217-223 Messmer, A., Papageorgiu M. 1995, Route Diver-sion Control In Motorway Networks Via Nonlin-ear Optimization, Ieee, Transactions On Control Systems Techno, Vol.3, 144-153 Middelham, F. 2001, Predictability: Some Thoughts On Modeling, Fut. Generation Computer System 17, 627-636 Pacilio, N., Sacripanti A. 2001,Tunnel Intelligenti, Rome, Enea, Ed Isbn 88-8286-0004-3 Sing, A., Li S. 2002, A Predictability Analysis Of Network Traffic, Computer Network, 39, 329-345 Sacripanti, A. 2004, Siti (Safety In Tunnel Intelli-gence), An Italian Global Project, Ieee Intelli-gence Transportation Systems Conference, Wash-ington, D.C., Usa. Smith, B.L., Williams B.M. & Oswald. R.K. 2002, Comparison Of Parametric And Non Parametric Models For Traffic Flow Forecasting, Transp. Research C, Vol.10, 303-321 Valenti, G., Mitrovich S. & Mancini M. 2007, A Dss For “Dynamic Tunnel” Traffic And Incident Management, Gratz Vuilleumier, F., Weatherill A & Crausaz B. 2002, Safety Aspects Of Railways And Road Tunnels, Tunnelling And Underground Space Technolo-gies, 17, 153-158 Webby, R., O’connor M. 1996, Judgemental And Statistical Time Series Forecasting: A Review Of The Literature, Int. Journal Of Forecasting, 12, 91-118 Williams, B.M. 1999, Modeling And Forecasting Vehicular Traffic Flow As A Seasonal Stochastic Time Series Process, Doc. Dissertation Univer-sity Of Virginia Williams, M.M.R., Thorne M.C. 1997, The Estima-tion Of Failure Rates For Low Probability Events, Progr. In Nuclear Energy, Vol.31, 343-476 Woodburn, P.J., Britter R.E. 1996, Cfd Simulation Of Tunnel Fire –Part Ii, Fire Safety Journal 26, 63-90 Woodburn, P.J., Britter R.E. 1996, Cfd Simulation Of Tunnel Fire –Part I, Fire Safety Journal 26, 35-62 Zang, G., Patuwo E. & Hu M.Y. 1998, Forecasting With Artificial Neural Networks: The State Of The Art, Int. Journ. Of Forecasting, 14, 35-62