[ieee 2006 international conference on computational inteligence for modelling control and...

Simulating Radiological Accidents through Software Agents

Tadeu Augusto de Almeida Silva1 Oscar Luiz Monteiro de Farias2 Neide dos Santos3

1, 2, 3{Universidade do Estado do Rio de Janeiro Programa de Pós-Graduação em Engenharia de Computação

Rua S.Francisco Xavier, 524, sala 5017, CEP 20550-900, Maracanã, Rio de Janeiro , RJ, Brasil}

1{Instituto de Radioproteção e Dosimetria Av. Salvador Allende s/n , CEP22780-160 , Recreio dos Bandeirantes,

Rio de Janeiro, RJ, Brasil} [email protected], [email protected], [email protected]

Abstract

Radiological accidents can result in significant radiation exposures of workers and of the public. These accidents usually occur as a consequence of human activities in military, industry, medicine, agriculture and research, involving the use of radiation and radioactive substances that cause radiation exposure in addition to the natural exposure. The effects of radiological disasters must be promptly investigated and the assessment of doses estimated, in order to provide adequate medical treatment to the persons affected, study their impact in society, as well learning lessons for the future and prevent further accidents. In this paper we introduce the use of software multi-agents systems immersed in a geographical representation of the world, as a viable and economic option to simulate radiological accidents and assess doses. Build an environment, where the researcher can simulate a non-natural radioactive dispersion, in a delimited area, can furnish parameters for analyzing, studying and managing these dynamic systems. 1. Introduction

The growing use of radioactive sources with specific objectives in several sectors of society (industry, hospitals, medical and odontological clinics, energy generation, food conservation, equipment sterilization and aviation) gave rise to the need of international and national controls related to radiation exposures.

All these apparatus are necessary in function of the magnitude of some radiological accidents. For

example, in September, 1987, occurred a radiological accident of Goiania, Brazil. This accident involved a radiotherapy source of 137Cs. The activity of the source was of 50.875 GBq or 1.375 Ci and it was composed of agglomerated powder. The net effect of this accident can be described by the following figures: 249 people with significant contamination; 20 people with haematology alterations; 4 deaths; 79 people with external contamination.

Defensive measures and actions involved 575 professionals and last 6 months; the evacuation of 41 houses by 200 inhabitants; the monitoring of 3.500 tons of radioactive garbage [1].

So, planning immediate answers and actions to emergency situations is a critical factor to establish security policies. Simulation models incorporating software agents are one of the tools that can be successfully employed in these tasks. They can mimic the ‘‘real world’’, and the interaction among people, materials and radioactive sources.

2. Agent-based systems

For thousand years people have created models to help them understand the world. These models, not only help them to better understand the phenomena they study, but they also help them to transmit their ideas to other people. Through computational simulation of agent-based models, people can perceive, with details, the effect of radioactive contamination in case of accidents, what makes easy the understanding of the risks involved and the damages caused in such situations.

The word “agent” refers to all beings who possesss the ability, capacity and permission to act on behalf of

International Conference on Computational Intelligence for ModellingControl and Automation,and International Conference onIntelligent Agents,Web Technologies and Internet Commerce (CIMCA-IAWTIC'06)0-7695-2731-0/06 $20.00 © 2006

another one. Speaking in human terms, agents would be people with a more knowledge or more specialized resource in determined area, by means of which they could help other people in their tasks [2].

Autonomous agents, or agents, are the software counterpart of real world entities like human beings, animals, viruses, ships, cars, etc. For this reason several definitions for autonomous agents coexist, for authors adopt different perspectives on the subject [3],[4],[2]. Here we give the following definition, paraphrased from reference[5]: “A system situated within a given environment, that senses that environment through its perception mechanism and acts on that environment and/or on other agents, as time flows, in pursuit of its own agenda, plans or beliefs. Eventually the agent´s perception/action mechanism evolves with time”.

Agents can be classified according to the amplitude of their perception and action capacity and effectiveness of action. In the lower and higher extremes of this classification scale, we found, respectively, reactive and cognitive agents. Reactive agents merely react, in an oportune way, and according to very simple pattern behaviours, to changes they perceive on their environment. The cognive agents, are more elaborated. They not only interact with their environment, but they are also capable of remembering previous experiences, learn with them, communicate among themselves, and follow a intended goal, strategy or plan.

Multi-agent systems are composed of several agents, which, additionally to the already previously commented agent characteristics, can interact ones with others through communication mechanisms. In the most general case these agents can differ completely among them and have a competitive or colaborative behaviour, according to a specific situation [2].

Multi-agent simulation systems provide a computational platform where the dynamic of spatio-temporal systems can be studied. To do this we can use reactive (simpler) or cognive (more elaborated) agents [6],[7].

We can model the behavior of people in some determined environment, as well as consider other elements that interact with these people, observing their behavior and the consequences of these interactions, through the computational technology of agents.

In this article we’ll show how to create an environment where software agents simulate radiological accidents. Our agents will have some properties, as: mobility, reactivity and objectivity. People and radioactive sources and elements will be also represented by software agents, In this way,

through agent-based simulation, we intend to study, analyse and manage these complex systems (radiological accidents).

3. Radioactivity

“Ionizing radiation represents electromagnetic waves and particles that can ionize, that is, remove an electron from an atom or molecule of the medium through which they propagate. Ionizing radiation may be emitted in the process of natural decay of some unstable nuclei or following excitation of atoms and their nuclei in nuclear reactors, cyclotrons, x-raymachines or other instruments”[8].

“A gamma ray is a packet of electromagnetic energy-a photon. Gamma photons are the most energetic photons in the electromagnetic spectrum. They are emitted from the nucleus of some unstable (radioactive) atoms. They have no mass and no electrical charge--they are pure electromagnetic energy. Gamma radiation is very high-energy ionizing radiation.

Because of their high energy, gamma photons travel at the speed of light and can cover hundreds to thousands of meters in air before spending their energy. They can pass through many kinds of materials, including human tissue.

Because of the gamma ray's penetrating power and ability to travel great distances, it is considered the primary hazard to the general population during most radiological emergencies”[9].

“The basic quantity used to express the exposure of material such as the human body is the absorbed dose, for which the unit is the gray (Gy). However, the biological effects per unit of absorbed dose vary with the type of radiation and the part of the body exposed. To take in account those variations, a weighted quantity, called effective dose is used, for which the unit is the sievert (Sv).

Radiation effects are caused by the damage inflicted in cells by the radiation interactions. The damage may result in cell death or modifications that can affect the normal functioning of organs and tissues.

Deterministic (acute) effects occur only if the radiation dose is substantial, such as in accidents. Stochastic effects (cancer and hereditary effects) may be caused by damage in a single cell. As the dose to the tissue increases from a low level, more and more cells are damaged and the probability of stochastic effects occurring increases”[8].

4. Radiological accidents

Radiological accidents are accidents that occur in


nuclear or radioactive installations and are characterized by the existence of intense fields of not intentional radiation, not controlled release in the environment of amounts of radioactive material, and involve exposition or contamination of human beings or the environment, being able to cause causing serious damages or death.

The incidental release of radionuclides in an environment might cause the contamination of areas and people. So, it is necessary to make use of tools that allow us to foretell the effects of the exposition of the population and to evaluate the consequences and to suggest measures of protection [10].

First of all, we need to identify the amount of contaminated people, degree or dose of radiation received and a more elaborate map (georeferenced information) from the areas impacted [11].

If we know with antecedence the magnitude of a radiological accident, we can avoid alarming the population unnecessarily, what it would be an inadequate answer to the event. This would come to cause a bigger social damage than the radiological accident properly said.

A radiological accident can affect the public transportation system, generate zones of exclusion in the contaminated areas, leading to the displacement of people, damage the water supply, and overcrowd the hospital services, causing serious social problems [11].

4.1. Measures of radioactivity

For the evaluation of the dose received for

individuals it is necessary to take in consideration the release rate of the radioactive source, its exact distance from the exposed individuals, the existence of materials (shield) between the radioactive source and the exposed individuals, and the time of exposition of the individuals. i) Activity of a radioactive source

The activity of a radioactive source is characterized by the number of nuclear disintegrations or transformations that occur in a certain interval of time. The activity is proportional to the number of excited atoms existing in the radioactive element, and it can be expressed by the following formula:

A=A0e-λ.t (equation 1)

Where: A0, is the activity at time t=0, and λ is the constant of disintegration, meaning the rate in which disintegration proceeds.

The International System (SI) unit for activity is the becquerel (Bq), which is that quantity of radioactive

material in which one atom is transformed per second. The curie (Ci) is also commonly used as the unit for activity of a particular source material.

ii) Exposure

Exposure is a measure of the strength of a radiation field at some point in air. The most commonly used unit of exposure is the Roentgen (R).

The exposition rate is defined as the variation of the exposition with the time, usually measured in roentgens per hour. (R/h), The exposition rate can be associated to the gamma activity of a radioactive source by the following formula:

X = Γ.A / d2 (equation 2)

Where, X = exposition rate, in R/h ( Roentgen / hour) A = source activity, in Ci ( Curie ) d = distance between the source and the point of measure, in meters. Γ = a characteristic constant of each radioactive source, also known as factor gamma, in (R.m2) / (h.Ci) iii) Absorbed dose

The basic quantity used to express the exposure of material such as the human body is the absorbed dose, for which the unit is the gray (Gy). Whereas exposure is defined for air, the absorbed dose is the amount of energy that ionizing radiation imparts to a given mass of matter. It is the amount of energy yielded to the matter for photons or ionizing particles. The SI unit for absorbed dose is the gray (Gy), which is defined as a dose of one joule per kilogram [12].

The absorbed dose is expressed by the formula: over which

D = X.t (equation 3) Where, D = absorbed dose, in Gray (Gy) X = exposition rate, in R/h t = time, in hours (h) iv) Dose equivalent (effective dose)

When considering radiation interacting with living tissue it is also important to take in account the type of radiation. Although the biological effects of radiation are dependent upon the absorbed dose, some types of radiation produce greater effects than others for the same amount of energy imparted. For example, for equal absorbed doses, alpha particles may be 20 times as damaging as beta particles. In order to compute these variations when describing human health risks from radiation exposure, the quantity called “dose equivalent” (DE) is used. It is the absorbed dose multiplied by certain “quality” or “adjustment” factors


indicative of the relative biological-damage potential of the particular type of radiation.

The quality factor (Q) is a factor used in radiation protection to weight the absorbed dose with regard to its presumed biological effectiveness [12].

DE = D.FQ (equation 4)

Where, DE = dose equivalent , in Sievert (Sv) D = absorbed dose, in Gray (Gy) FQ = quality factor of radiation, for gamma radiation FQ equals 1

Time, distance and shield are factors that influence in the dose received for a person.

5. Architecture of agent-based systems to simulate radiological accidents

The goal of our agent-based system is to furnish

useful information related to a given radiological accident. This class of information is usually related to: i) the quantity of persons that were esposed; ii) the effective dose they received; iii) determination, localization and extension of contaminated areas.

Not always it is possible to determine who where the persons exposed, neither their exact number. This would be the case, for example, of a terrorist attack with a radioactive source in the public system of transportation, for example in a train or in a subway. Notwithstanding it is very important for the public health system, to have a estimate of the people involved and of the effective dose they received, in order to calculate the medical resources that will be need to face the problem, to warn the population against the risks of exposure, and also to manage the entire situation, all the risks involved, descontamination procedures, etc.

Our model is based on the formula of radioactive dispersion in the air (equation 2), leading in account the radionuclide and its activity, the distance of the people exposed to the radioactive element, the time of exposition, as well as possible shields. We also consider eventual movements of persons and of radioactive sources. Both of them are represented by very simple agents, People agents suffer the effects of radioactive source agents, based on equation 2, and these effects are registered on the state variables of people agents. Shield effects could, as well, be incorporated into the model.

The source agent has the following state variables: i) The source Id; ii) Its position, given by a tuple of coordinates (x, y, z); iii) Its activity A; iv) The factor Γ for the specific source; v) The quality factor (FQ), a

factor used to weight the absorbed dose with regard to its presumed biological effectiveness. A typical agent, representative of a person, has the following state variables: i) Person Id; ii) Its position, given by a tuple of coordinates (x, y, z); iii) Time of exposition; iv) Shield effect; v) The absorbed dose; vi) The effective dose.

In our model we need also a representation for the space, the environment from where our agents take sensory inputs and produce as output actions, that is, their movements. Normally Geographical Information Systems (GIS´s) use raster or vector structures to represent space in bi-dimensional models. In some cases a third dimension is represented through digital elevation models (dem) of a terrain. Given a GIS spatial representation (a shape file, for example), we’ll add to it software agents and the structure of a dynamic spatial model, in order to simulate the dynamic of radiological accidents. The GIS spatial representation is the environment or locus, where agents of our Agent-Based Models will operate.

For each specific phenomenon, we are interested, only, in particular information about the environment. So, considering the geographical space where the phenomenon develops, we need to filter only the aspects of the environment we are interested in, that is, the features that will compose the environment as seen by the agents, and where all the process will be simulated. This is not a difficult task, for data is organized on GIS´s in different layers, such that: utilities, river and lakes, roads and rails, soil maps, land parcels, etc. We need to select only the layers we are interested in. Objects that are not of interest must be discharged, for an environment full of superfluous objects would unnecessary complicate the implementation and reduce the simulation performance. So, depending on the problem, we can simplify considerably the simulation, as we´ll see below in our case study [6],[7]. 6. Case study: the radiological accident at Cochabamba, Bolívia.

We will take as our case study the radiological accident occurred in Cochabamba, Bolivia, in 2002. This accident was minutely documented and physicaly reconstructed by the IAEA [13], and the assessment of doses estimated. So, we have a valuable data to compare with the results of our agent-based simulation and to validate our model.

In April 2002 an accident involving an industrial radiography source containing 192Ir occurred in Cochabamba, Bolivia, some 400 km from the capital, La Paz. The source, in a remote exposure container,


remained exposed within a guide tube, although this was not known at the time. The container, the guide tube and other equipment were returned from Cochabamba to the headquarters of the company concerned in La Paz as cargo on a passenger bus. This bus carried a full load of passengers for the journey of about eight hours from Cochabamba to La Paz. The equipment was subsequently collected by company employees and transferred by taxi to their shielded facility. Routine radiation measurements made there established that the source was still exposed and actions were then taken to return the source to its shielded container [13].

In our case study, due to the source characteristics, the effect of the radiation was relevant only to people situated a few meters from the source, that is, for people inside the bus. If we look at the exposition rate, absorbed dose and dose equivalent (effective dose) formulas, we’ll see that all these measures - given a specific source - are dependent only of the relative position between the source and the people exposed and of the time of exposition. So, in this very special case, it doesn’t matter the path followed by the source in its journey from Cochabamba to La Paz. Assuming that all passengers, after entering the bus, didn’t change their seats, what is relevant to assess doses is: i) the distance between each passenger and the radioactive source that remained fix during the entire journey; ii) the exposure time.

Figure 1. Bus passenger

Figure 1. shows the relative distance between the

passengers and the source, as well a classification of doses received.

“The time frame of the exposures of the bus passengers is reasonably well defined. For most of the passengers on the bus that day, this is an eight-hour period from 16:00 to 24:00. There are some variations on this; for example, the duration would have been 30 minutes shorter for those passengers picked up at Quilacollo and it would have been longer for those passengers from Cochabamba who spend some time on

the bus before it departed. Also some time would have been spent off the bus during a meal stop” [13].

The Bolivian Institute of Nuclear Science and Technology (IBTEN) had estimated the doses the passengers received (see table 1) assuming the height of the passenger seats, which respect to three possible heights at which the source may have been stowed in the cargo compartment of the bus. Table 1. Ranges of estimated doses (Gy) to the bus passengers by the IBTEN [13] Seat number High Medium Low 1–4 0.02 0.02 0.02 5–6 0.03 0.03 0.03 7–10 0.04 0.04 0.04 11–14 0.07 0.07 0.06 15–18 0.14 0.12 0.09 19–22 0.38 0.24 0.16 23–26 1.64 0.52 0.24 27–30 2.77 0.52 0.24 31–34 2.29 0.52 0.23 35–38 0.52 0.29 0.17 39–42 0.17 0.14 0.11 43–46 00.9 0.08 0.07 47–50 0.05 0.05 0.04 51–55 0.03 0.03 0.03

In our Agent-Based simulation we employed essentially the same formula (equation 2) for doses estimation. However, we summed 14 cm to the seat heights, for the whole body dose may be better represented at an eight of about one third of the way up the torso (see table 2).

The received dose (d) for the passengers, after reconstruction of the dose and technical analysis conducted by an IAEA technical mission, was calculated between a minimum 10mGy (~DE = 10mSv) to a 190 maximum mGy (~DE = 190 mSv) for 8 hours of the bus trip (See table 3). The dose received annually for the worldwide population due to natural radiation is calculated in 1,12 mSv/year and in the case of a source of 192Ir, it is necessary only 4 seconds of exposure to surpass the limit of annual dose (1,12mSv/year).

As we can see from tables 2 and 3, the simulation gave us upper limits for doses absorbed, basically for not considering shield effects.

7. Conclusions

This very simple example gave us a glance of the broad possibilities of simulating radiological accidents making use of Agent-Based Systems. We could, for example, very easily simulate the consequences for the


Table 2. Ranges of estimated doses (Gy) to the bus passengers by Software Agent Simulation Seat number High Dose (Agent simulation) 1–4 0.02 5–6 0.03 7–10 0.04 11–14 0.07 15–18 0.13 19–22 0.31 23–26 0.85 27–30 1,08 31–34 1,00 35–38 0.40 39–42 0.16 43–46 00.9 47–50 0.05 51–55 0.03

Table 3. Ranges of estimated doses (Gy) to the bus passengers based on the dose reconstruction performed by the IAEA team [13] Seat number Upper Lower (by row) 1–6 0.010 0.001 7–10 0.015 0.002 11–14 0.025 0.005 15–18 0.040 0.010 19–22 0.070 0.020 23–26 0.160 0.040 27–30 0.190 0.050 31–34 0.160 0.040 35–38 0.070 0.020 39–42 0.040 0.010 43–46 0.025 0.005 47–50 0.015 0.002 51–55 0.010 0.001 population, of a radioactive source left in a subway, estimate absorbed doses, number of people involved, physical consequences, spatial distribution of people contaminated, the possibilities of contamination, etc. With this we could have a first idea of the magnitude of the radiological accident, its impact on the healthy system, statistics about the number of people contaminated, the kind of injuries they suffered, etc.

Through Agent Based Systems we can build scenarios of radiological accidents that enable us to evaluate the consequences of accidental contaminations.

8. References [1] International Atomic Energy Agency, IAEA, The Radiological Accident in Goiânia, Vienna, Austria, 1988. [2] Wooldridge, M. J. An Introduction to MultiAgent Systems. John Wiley & Sons, Ltd, April, 2002. [3] M. J. Wooldridge, and N. R. Jennings, “Agents theories, architectures, and languages: a survey.” In Proceedings of ECAI’94 Workshop on Agent Theories, Architectures and Languages, 1995, pp.1-22. [4] J. Ingham, “What is an agent?” Technical Report #6/99, Centre for Software Maintenance, University of Durhan, Durhan, London. [5] S. Franklin, and A. Graesser, “Is it an agent, or just a program?: a taxonomy for autonomous agents.” In Inteligents Agents III: Agent Theories, Architectures, and Languages, edited by J. P. Muller, M. J. Wooldridge, and N. R. Jennings, Berlin: Springer, 1997, pp.21-35. [6] O. L. M. Farias and N. Santos. Agent-Based Geographical Information System. In: International Conference on Intelligent Agents, Web Technologies and Internet Commerce - IAWTIC'2005, 2005, Vienna. Proceedings of the International Conference on Intelligent Agents, Web Technologies and Internet Commerce - IAWTIC'2005. Los Alamitos, California : Edited by M. Mohammadian, 2005. v. 1. p. 998-1005. [7] O. L. M. Farias, and L. T. Leite. Simulation and Control of Maritime Area through Multi-Agent Systems and Geographical Information Systems . In: International Conference on Intelligent Agents, Web Technologies & Internet Commerce - IAWTIC 2005, 2005, Vienna. Proceedings of the International Conference on Intelligent Agents, Web Technologies and Internet Commerce - IAWTIC'2005. Los Alamitos, California : Edited by M. Mohammadian, 2005. v. 1. p. 734-738. [8] United Nations, “Report of the United Nations Scientific Committee on the Effects of Atomic Radiation to the General Assembly”, 2000. [9] http://www.epa.gov/radiation/understand/gamma.htm, accessed in July, 3, 2006 [10] International Atomic Energy Agency, IAEA, Radiation Safety, Vienna, Austria, 1996. [11] A.M. Agape, R.V. Arutyunyan, A.Yu.Kudrin, A.M. Eliseev “Experience of past radiation accidents and problems of response to possible dispersion of radioactive material in urban conditions”, , International Conference on the safety and security of radioactive sources towards a global system for the continuous control of sources throughout their life cycle, 27 June-July 2005. [12] http://www.ndt-ed.org/EducationResources/CommunityCollege/RadiationSafety/quan_units/units.htm, accessed in July, 3, 2006. [13] International Atomic Energy Agency, IAEA, The Radiological Accident in Cochabamba, Vienna, Austria, 2004.


[ieee 2006 international conference on computational inteligence for modelling control and...

Documents