supporting multi-group emergency management with multimedia

12
Supporting multi-group emergency management with multimedia Y. Ikeda a , G.E.G. Beroggi b, *, W.A. Wallace c a Engineering Science, Rensselaer Polytechnic Institute, Troy, New York, USA b Systems Engineering, Policy Analysis and Management, Delft University of Technology, Netherlands c Decision Sciences and Engineering Systems, Rensselaer Polytechnic Institute, Troy, New York, USA Abstract Many dierent groups of experts are involved in the management of complex large-scale operations during normal operation and also in cases of emergencies. Advanced communica- tions and information processing technologies (e.g. multimedia and the Internet) provide the basis for real-time decision support for emergency response. However, to assure eective information processing and decision-making, the technology must be complemented by sophisticated analysis and decision-making models. In this paper we discuss the concept of emergency management for multiple groups of experts, and propose a decision support approach which capitalizes upon advanced multimedia technology. The concept is based on an expanded paradigm of operational risk management which is based on the concept of assessing and revising courses of action in a real-time environment. The approach is discussed for the case of an emergency response organization in a nuclear power generation plant. # 1998 Elsevier Science Ltd. All rights reserved. Keywords: Decision support systems; Operational risk management; Multimedia; Emergency response; Nuclear power generation 1. Introduction An important consideration in industrial operations is the appropriate handling of unexpected events which occur very infrequently but which could cause detrimental damage. The crucial aspect in emergency management refers to the timely collection, processing, and dissemination of data, as a basis for decision making. To support response operations, a set of alternative response plans are devised in a strategic manner (Beroggi and Wallace, 1994a). However, unexpected events occur (called real-time events, RTEs) which have not been considered in the strategically devised SAFETY SCIENCE Safety Science 30 (1998) 223–234 0925-7535/98/$—see front matter # 1998 Elsevier Science Ltd. All rights reserved. PII: S0925-7535(98)00038-1 * Corresponding author. E-mail: [email protected].

Upload: y-ikeda

Post on 04-Jul-2016

220 views

Category:

Documents


7 download

TRANSCRIPT

Page 1: Supporting multi-group emergency management with multimedia

Supporting multi-group emergency managementwith multimedia

Y. Ikedaa, G.E.G. Beroggi b,*, W.A. Wallace caEngineering Science, Rensselaer Polytechnic Institute, Troy, New York, USA

bSystems Engineering, Policy Analysis and Management, Delft University of Technology, NetherlandscDecision Sciences and Engineering Systems, Rensselaer Polytechnic Institute, Troy, New York, USA

Abstract

Many di�erent groups of experts are involved in the management of complex large-scaleoperations during normal operation and also in cases of emergencies. Advanced communica-tions and information processing technologies (e.g. multimedia and the Internet) provide the

basis for real-time decision support for emergency response. However, to assure e�ectiveinformation processing and decision-making, the technology must be complemented bysophisticated analysis and decision-making models. In this paper we discuss the concept ofemergency management for multiple groups of experts, and propose a decision support

approach which capitalizes upon advanced multimedia technology. The concept is based onan expanded paradigm of operational risk management which is based on the concept ofassessing and revising courses of action in a real-time environment. The approach is discussed

for the case of an emergency response organization in a nuclear power generation plant.# 1998 Elsevier Science Ltd. All rights reserved.

Keywords: Decision support systems; Operational risk management; Multimedia; Emergency response;

Nuclear power generation

1. Introduction

An important consideration in industrial operations is the appropriate handling ofunexpected events which occur very infrequently but which could cause detrimentaldamage. The crucial aspect in emergency management refers to the timely collection,processing, and dissemination of data, as a basis for decision making. To supportresponse operations, a set of alternative response plans are devised in a strategicmanner (Beroggi and Wallace, 1994a). However, unexpected events occur (calledreal-time events, RTEs) which have not been considered in the strategically devised

SAFETY SCIENCE

Safety Science 30 (1998) 223±234

0925-7535/98/$Ðsee front matter # 1998 Elsevier Science Ltd. All rights reserved.

PII: S0925-7535(98)00038-1

*Corresponding author. E-mail: [email protected].

Page 2: Supporting multi-group emergency management with multimedia

plans. When RTEs occur, the operators must analyze, and possibly change, theplans to assure the best possible response.Large industrial processes require that the assessments and decision-making pro-

cesses be completed rapidly under stressful conditions, involving di�erent groups ofexperts. To cope with the combinatorial complexity caused by the large numberof possible solutions, a graph±theoretic approach to revising plans in real time hasbeen proposed (Beroggi and Wallace, 1994a). This new risk assessment and decision-making paradigm was originally devised for an individual decision maker (opera-tor), and then extended to include multiple experts, called multi-expert operationalrisk management (Beroggi and Wallace, 1998).This paper proposes an approach to e�ective decision support for groups

of experts engaging in emergency response. Special attention is given to the dis-semination of information among the groups. For example, emergency response toan incident at a nuclear power generating facility involves at present three groupsstationed at di�erent sites. Communication between these groups was traditionallyverbal with limited use of video (Housel et al., 1986).

2. Multiple groups in decision making for emergency management

2.1. The groups in an emergency response organization

The di�erent groups in an emergency response organization are responsible fortheir unique tasks (Fig. 1). For most cases, one group cannot overrule anothergroup's recommendations or perform its tasks. There is usually one single ultimatedecision maker, who has the ®nal responsibility for the actions taken in response toan emergency. In the process of communications and information exchange, thegroups generate multiple courses of action. It is up to the groups together to decideon the ranking of the alternative courses of action to present to the decision makeras aggregated recommendations.

Fig. 1. Multi-group emergency response organization.

224 Y. Ikeda et al./Safety Science 30 (1998) 223±234

Page 3: Supporting multi-group emergency management with multimedia

Emergency response organizations dealing with large-scale operations consist offour groups, each group containing at least one expert. The groups are characterizedbased on the members' responsibilities during an emergency response:

1. Operators, who execute procedures of the operation under the decisionmaker's supervision in order to accomplish the objectives of the emergencyresponse.

2. A recommendation-generating group, which generates recommendations forthe decision maker to consider. The members of this group have specialknowledge of the technologies of the operations and generate recommenda-tions based on their expertise.

3. A management group, whose responsibility is to consider the impact on theorganization responsible for the operations.

4. A decision maker, who has the authority and responsibility to make decisionsabout the operations in an emergency situation. The decision maker will changewith the increasing severity of the event. For the purpose of this discussion, thedecision makers will be either the operators or the management group.

The goal is to resolve the emergency situation and to return the operation tonormal operating conditions. Any proposed decision support technology must aideach of the groups in: (1) monitoring, (2) real-time information exchange, (3) inter-connectedness, (4) real-time recommendation generation with preference values,(5) user-friendly interfaces, and (6) targeted information display.If the system displays are not designed to provide only the information appro-

priate for particular groups based on their responsibilities, information overloadcould occur. It is, therefore, important that the technology includes a modelingcapability which provides the groups with the appropriate cognitive support tofacilitate decision making in stressful situations of emergency response.

2.2. Preference structure

Real-time events can make an emergency more severe and the ongoing and plan-ned response activities harder to control (Beroggi and Wallace, 1994a). In orderto respond best to any RTE, the groups and their members must be aware of allreal-time events that occur or that might occur.E�ective decision support can be provided through a reasoning model which

generates a set of recommendations in real-time and orders them according to thegroups' preferences. The principle of such a reasoning logic is described by Beroggiand Wallace (1994a). An emergency plan is seen as a course of action, consisting of aconcatenated set of decisions and subsequent actions. A visually intuitive way torepresent all possible courses of action and their relations is to use a graph structure.The actions are the edges of the graph and the decisions to take the actions thevertices. A response plan is, therefore, a course of action or a `path' on the decision±action graph. It starts with the identi®cation of the emergency situation and endswith the decision to resume normal operations.

Y. Ikeda et al./Safety Science 30 (1998) 223±234 225

Page 4: Supporting multi-group emergency management with multimedia

We assume that the preference of taking an action under normal conditions (i.e. asplanned strategically) is assessed in terms of di�erent attributes, such as risks tohumans, operational costs, and risks to the environment. The attributes should notbe seen as evaluation criteria. They are preference classes which are related to eachother by a complete strong preference structure. This preference structure couldstate, for example, that high risks are of greater concern than operational costs.Whenever a RTE occurs, the preferences of the actions which are a�ected by theevent must be reassessed in terms of the chosen attributes. Using these reassessedpreferences, new optimal courses of action can be computed.The reasoning logic employed to compute new courses of actions takes into

account four components: (1) the large-scale operational system (LSOP), which issupposed to behave as considered in the strategically devised courses of action(CAs); (2) the unexpected RTEs which have an impact on the ongoing responseoperations, and which cannot be considered strategically; (3) the human±machinesystem which (i) monitors the LSOP as part of passive ORM, (ii) assesses impacts ofRTEs on the planned CAs as part of active ORM, (iii) if necessary revises the CAsas part of active ORM, and (4) has recommendation generating capability thatranks recommendations according to the emergency situation, including the poten-tial for catastrophic impacts, and the responsibilities of the users.After the preferences of those activities which were a�ected by the RTE are

assessed, the new overall preference of a course of action can be computed. Thepreference of two activities is the `sum' of their preferences. Thereby, the sum isbased on a lexicographic principle, such that the preferences within a preferenceclass are added up, but the preference classes remain incommensurable. For exam-ple, if high risks are of higher concern than operational costs, then there is noamount of operational costs to override concerns for high risks. That is, high risksmust be avoided regardless of the ®nancial implications.The preference of a CA is, therefore, the sum of the preferences of its activities.

The most preferred CA, among the feasible CAs, is the one with highest overallpreference value. The most preferred CA can be found by an appropriate graph±theoretic search algorithm (Beroggi and Wallace, 1994a).In addition to high risks, costs, and low risks, two boundary preference classes are

introduced. Activities which do not have any signi®cant risks or costs are assigned tothe ®rst class (!); i.e. their risk and cost concerns are negligible. Activities whichunder no circumstances are to be engaged in are assigned to the second class (�);i.e. they must be avoided. The resulting ®ve preference classes have the followingordinal relation:

�< High Risks < Costs < Low Risks < !;

where `<' means `less preferred'.An activity or a course of action is feasible only if its overall preference is higher

than �. After the preferences of all activities on the graph are assessed, the graph iscalled a preference graph. The principle of the algorithm to compute optimal coursesof action is based on a hierarchical procedure (Beroggi and Wallace, 1994a):

226 Y. Ikeda et al./Safety Science 30 (1998) 223±234

Page 5: Supporting multi-group emergency management with multimedia

1. never recommend an activity with Risks or Costs of value �,2. avoid as much as possible links with High-Risk values;3. minimize Costs; and4. avoid as much as possible links with Low Risk values.

2.3. Preference ranking with multiple groups of experts

In case multiple groups are involved in emergency response in a distributedenvironment, each group might use a di�erent preference structure to assess pre-ferences for actions. A reason for this can be due to the di�erent responsibilities ofthe groups. Let us assume we have the following meta-preference structure consist-ing of the seven preference classes �, High Risk (HR), Low Feasibility (LF), Cost,Low Risk (LR), High Feasibility (HF), and !:

�< HR < LF < Cost < LR < HF < !:

The di�erent preference spectra for the di�erent groups are given in Table 1.Group 1's preference spectrum is limited to the preference classes �, HR, LR, and !.Therefore, Group 1's preference order is:

�< HR < ÿ < ÿ < LR < ÿ < !;

which can be written as:

�< HR < LR < !:

For the personnel in Group 2, the preference structure is:

�< ÿ < LF < ÿ < ÿ < HF < !;

which can be written as:

�< LF < HF < !:

For Group 3 we have:

� < Cost < !:

Whenever a RTE occurs, the group members must assess the impact of the RTEon all a�ected activities using their designated preference spectrum. It is assumedthat the a�ected activities are presented to the experts one by one.

Table 1

Responsibility for assessing preference classes

� HR LF Cost LR HF !

Group 1 X X X X

Group 2 X X X X

Group 3 X X X

�, not acceptable; HR, High Risk; LF, Low Feasibility; LR, Low Risk; HF, High Feasibility; !, negligible.

Y. Ikeda et al./Safety Science 30 (1998) 223±234 227

Page 6: Supporting multi-group emergency management with multimedia

An example is given to illustrate the aggregation of the preferences across thegroups. Let us assume that Group 1 recommends two alternatives with the followingpreferences: Alt. 1=[0, 2, ÿ, ÿ, 0, ÿ, 0] and Alt. 2=[0, 2, ÿ, ÿ, 0, ÿ, 0]. Group 2assesses the preferences of the same two alternatives as follows: Alt. 1=[0,ÿ, 1,ÿ,ÿ,3, 0] and Alt. 2=[0, ÿ, 1, ÿ, ÿ , 4, 0]. Group 3 assesses the alternatives as: Alt. 1=[0,ÿ, ÿ, 5, ÿ, ÿ, 0] and Alt. 2=[0, ÿ, ÿ, 4, ÿ, ÿ, 0], where the preference functions aremonotonically decreasing, such that a $50 cost alternative is less preferred than a $40cost alternative.The two alternatives have the same preference order Group 1, while Group 2

prefers Alt. 1 and Group 3 prefers Alt. 2. The aggregation of preferences across thegroups is done with the meta-preference structure:

� < HR < LF < Cost < LR < HF < !:

The resulting preferences for the two alternatives are shown in Table 2. Thus withall the group's assessments the order of the alternatives has been determined asAlt. 2>Alt. 1. The decision maker now knows both the assessments by the groups,as well as the aggregated preferences. To operationalize this proposed decision logic,the groups involved in the decision-making process must be supported in theirreal-time communication process. We propose that this be accomplished by the useof multimedia technology.

3. Multimedia

Most of the past research has focused on individual modes of media presentationand on decision support systems (DSSs), such as voice (DeHaemer and Wallace,1992) video conferencing (Housel et al., 1986; Chidambaram and Jones, 1993), gra-phics (Jessen et al., 1992; Chau, 1993). Little attention was given to integrated multi-media and DSSs (Beroggi and Wallace, 1994b; Beroggi et al., 1995; Burger, 1995;Hatcher, 1995a; Ikeda et al., 1995; Ramesh and Sengupta, 1995). The advantage ofintegrated multimedia presentations over single types of media is that they can betailored much better to aspects of the information the presenters want to convey.Distributed multimedia systems for dispersed groups of experts provide, therefore,important information about a situation to appropriate individuals and groups. Thisprovides for allmembers of the emergency response organization the basis for the sameunderstanding of the situation. Their recommendations will, therefore, be based onthis understanding, while avoiding unnecessary information processing.

Table 2

Assessments of the alternatives by each group

� HR LF Cost LR HF !

Alt. 1 0 2 1 5 0 3 0

Alt. 2 0 2 1 4 0 4 0

�, not acceptable; HR, High Risk; LF, Low Feasibility; LR, Low Risk; HF, High Feasibility; !, negligible.

228 Y. Ikeda et al./Safety Science 30 (1998) 223±234

Page 7: Supporting multi-group emergency management with multimedia

3.1. Multimedia on network

It must be recognized that multimedia requires a large amount of space for data,making compression and decompression (coded) technology extremely importantfor sending and receiving multimedia data over a computer network. For example, a640�480 picture with 24 bits of color (millions of colors) needs 7.4 Mbits (920 kB);1 min of the same size and color digital movie needs 13.32 Gbits (1.66 GB); and 1min of compact disk (CD) quality sound needs 10 MB. An Ethernet local area net-work (LAN), which is the most popular LAN, can transfer data at a speed of 10Mbits/s (bps). A 1-min digital movie needs 22 min to transfer on an Ethernet LAN.Coded technology must be employed in order to construct a real-time system on a

computer network. Unlike still media, time-dependent dynamic presentation has themost impact on multimedia networking. Time-dependent media include digitalvideo, voice, live music, and live video.

3.2. QuickTime

QuickTime (a trademark of Apple Computer, Inc.) uses the concept of standard-izing a digital movie on a personal computer (Macintosh, Windows). This includescommon user-interfaces, video/audio coding, ®le format, and application program-ming interfaces (APIs). Besides video (or time-based data), the standard also appliesto still media, such as pictures, and it integrates time-based data into mainstreamapplications (Apple Computer, 1993), including multimedia authoring tools. Videoand audio in QuickTime are handled in separate tracks, where each track has atimetable which makes the synchronization of video and audio possible.The exchange of time-based data among computers connected to a computer

network is simpli®ed through the use of QuickTime Conference. This is done withQuickTime's coded method and its standardization, in addition to Apple's Apple-Talk (a trademark of Apple Computer, Inc.) protocol and TCP/IP computer net-working. The user-interfaces, conference events, transportation of the conferencedata, initialization of connections, etc., are managed by the programming environ-ment of QuickTime Conference. Moreover, programmers can exchange their owntypes of data in addition to conference data, such as video and sound. These func-tions allow one to send and receive vital data for multimedia communicationsinvolving multiple groups of experts. The use of QuickTime Conference on eitherLAN or TCP/IP provides the possibility to connect a system with either LAN orTCP/IP; the change from LAN to TCP/IP is easy to realize (Apple Computer, 1996).Consequently, QuickTime Conference technology is what we use to develop the DSSand to integrate the decision logic for the multi-group emergency management layout.

4. Prototype system

We have developed a prototype DSS that incorporates the proposed decision logicand multimedia technology for an emergency response organization in a nuclear

Y. Ikeda et al./Safety Science 30 (1998) 223±234 229

Page 8: Supporting multi-group emergency management with multimedia

power generation plant. The prototype system supports three sites: the control room(CR), the technical support center (TSC), and the emergency operation facility(EOF). The members of the emergency response organization (ERO) can share vitalinformation due to the connections among the three sites. The operators and theoperational decision maker on duty are located in the CR. The recommendation-generating group is sta�ed in the TSC, and the management group stays in the EOF.The connections among these three sites have been established with QuickTime

Conference, including video-conferencing systems, the capability to exchange com-mand messages, and several other information exchange capabilities. The multi-media data are stored on one workstation and controlled with the commandmessages via QuickTime Conference linkages. This setup reduces the amount ofcomputer tra�c on the computer network.

4.1. Functions of the prototype system

Fig. 2 shows a screen view of the prototype DSS supporting the CR duringnormal or routine operations, i.e. no emergency has been declared. Under normaloperations, no communication links to any other sites are active.The system was designed for each site, so each system has its own function based

on the responsibility of the groups in the sites.The system functions are as follows:

1. `video conferencing' (all the sites);2. `talk' that enables a `chat' function between the sites (all the sites);3. `event board' that displays the status of the emergency including the declared

emergency classes (all the sites);4. `plant status board' that indicates the power plant's parameters input by an

operator in the CR (all the sites);5. `connection window' that displays the conditions of system connections

between the sites (all the sites);6. `geographic information system (GIS) map' that indicates wind direction and

speed of radioactive plumes (TSC, EOF);7. `plant overview' that displays the plant condition using graphical view (CR,

TSC);8. `station blackout procedures' that indicate emergency procedures (CR, TSC);

and9. `RPV control graphs' that track plant conditions (CR, TSC).

4.2. Emergency response

As soon as an emergency occurs, the system shift supervisor (SSS), who is thedecision maker in the CR, reports the emergency to the plant authorities. Then,the TSC and EOF start to be sta�ed. The system connection is immediately estab-lished as soon as personnel of the TSC and EOF report in, and information starts tobe exchanged (Fig. 3).

230 Y. Ikeda et al./Safety Science 30 (1998) 223±234

Page 9: Supporting multi-group emergency management with multimedia

If another RTE occurs which causes station blackout, the operators must conductthe appropriate emergency procedures in a correct and timely manner. Theprocedures are stored in the system for activation by the operator.

Fig. 2. Prototype of multi-group operational risk management decision aid. CR, control room.

Y. Ikeda et al./Safety Science 30 (1998) 223±234 231

Page 10: Supporting multi-group emergency management with multimedia

The wind direction and speed that are obtained in CR and transferred to EOF inorder to make recommendations for evacuation of the plant or surrounding com-munity are displayed on the GIS. The spreading radioactive plumes are calculated atTSC and the information is transferred on the GIS display (Fig. 4).Control graphs for the reactor pressure valves (RPVs) are displayed during the

emergency operations (Fig. 5). Eight graphs are used to monitor the plant condition,where ERO tries to track the plant parameter, keeping safety range with con-sequences of some procedures. Data readings for the graphs are displayed in the CR;data are transferred to the TSC by clicking with the computer mouse on the graph.When the graph indicates that a plant parameter is about to enter the unsafe range,the personnel in the TSC must recommend appropriate procedures for the operators.

Fig. 3. Connection window (site area emergency). TSC, technical support center; EOF emergency

operation facility.

Fig. 4. Geographical information system (GIS) around the power plant.

232 Y. Ikeda et al./Safety Science 30 (1998) 223±234

Page 11: Supporting multi-group emergency management with multimedia

Fig. 6 indicates the plant overview. The status of pumps and valves is displayedgraphically. If the ERO ®nds a malfunctioning pump, the information is transferredon this view by clicking on the equipment. A cross shows up when the equipment isdown.

5. Future research

Our current work includes the development of an experimental assessment of theproposed multi-group decision logic during an exercise at a nuclear power genera-tion plant. The experiment will be conducted with nuclear plant operators/engineersfollowing a drill scenario. Our objective is to evaluate both the multimedia technol-ogy and the decision logic. Because models are a way of providing the cognitivesupport needed to process data into information, they must be assessed togetherwith the technology within which they are embedded. Past research (Beroggi andWallace, 1995) has shown the concept of the proposed logic to be useful for theindividual decision maker in an emergency response setting. Therefore, the multi-group formulation has yet to be tested.

Fig. 6. Plant overview.

Fig. 5. Reactor pressure valve (RPV) control graphs.

Y. Ikeda et al./Safety Science 30 (1998) 223±234 233

Page 12: Supporting multi-group emergency management with multimedia

References

Apple Computer, 1993. Inside Macintosh: QuickTime. Addison-Wesley, Cupertino, CA.

Apple Computer, 1996. Inside Macintosh: QuickTime Conference. Addison-Wesley, Cupertino, CA.

Beroggi, G.E.G., Wallace, W.A., 1994a. Operational risk managementÐa new paradigm for decision

Making. IEEE Transactions On Systems Man and Cybernetics 24, 1450±1457

Beroggi, G.E.G., Wallace, W.A., 1994b. A prototype decision support system in hypermedia for opera-

tional control of hazardous material shipments. Decision Support Systems 12, 1±12

Beroggi, G.E.G., Wallace, W.A., 1995. Operational control of the transportation of hazardous materials:

an assessment of alternative decision models. Management Science 41, 1962±1977.

Beroggi, G.E.G., Wallace, W.A., 1998. Multi-Expert Operational Risk Management. School of Systems

Engineering, Policy Analysis, and Management, Delft University of Technology, Delft, Netherlands.

Beroggi, G.E.G., Waisel, L., Wallace, W.A., 1995. Employing virtual reality technology to support deci-

sion making in emergency management. Safety Science 20, 79±88.

Burger, 1995. Multimedia for Decision MakersÐa Business Primer. Addison Wesley.

Chau, P.Y.K., 1993. Decision support using traditional simulation and visual interactive simulation.

Information and Decision Technologies 19, 63±76.

Chidambaram, L., Jones, B., 1993. Impact of communication medium and computer support on group

perceptions and performance: a comparison of face-to-face and dispersed meetings. MIS Quarterly 17,

465±488.

DeHaemer, M., Wallace, W.A., 1992. The e�ects on decision task performance of computer synthetic

voice output. International Journal of Man±Machine Studies 36, 65±80.

Hatcher, M., 1995a. Introduction to multimedia supported group/organizational decision system. Deci-

sion Support Systems 15, 179±180

Hatcher, M., 1995b. A tool kit for multimedia supported group/organizational decision systems

(MSGDS). Decision Support Systems 15, 211±217

Housel, T.J., Sawy, O.A.E., Donovan, P.F., 1986. Information systems for crisis management: lessons

from southern California. MIS Quarterly 10, 389±400.

Ikeda, Y., Beroggi, G.E.G., Wallace, W.A., 1995. Real-time air-raid command (RARC): and application

of operational risk management. In: Military, Government and Aerospace Simulation, Michael J.

Chinni (Ed.), Simulation Series, 27, 4, The Society for computer simulation.

Jessen, D.M., Grabowski, M.R., Triscari, T., Wallace, W.A.A., 1992. A test of stereoscopic imaging in a

tactical command and control decision situation. Symposium of Command and Control Research,

Monterey, CA.

Ramesh, B., Sengupta, K., 1995. Multimedia in a design rationale decision support system. Decision

Support Systems 15, 181±196.

234 Y. Ikeda et al./Safety Science 30 (1998) 223±234