Regulatory Guide 5.44

(Task SG 479-4), Perimeter Intrusion Alarm Systems

ML003740097
Person / Time
Issue date:	05/31/1980
From:	Office of Nuclear Regulatory Research
To:
References
RG-5.44, Rev 2
Download: ML003740097 (17)

Revision 2 May 1980

U.S. NUCLEAR REGULATORY COMMISSION

REGULATORY GUIDE

OFFICE OF STANDARDS DEVELOPMENT

REGULATORY GUIDE 5.44 (Task SG 479-4)

PERIMETER INTRUSION ALARM SYSTEMS

B. DISCUSSION

.4' Part 73, "Physical Protfection of Plants and Materials," of Perimeter intrusion alarm systems can be used to detect Title 10, Code of Federal Regulations, specifies performance intrusion into or through the isolation zone at the perimeter requirements for the physical protection of special nuclear of the protected area. A system generally consists of one or materials and associated facilities. Section 73.20 describes more sensors, electronic processing equipment, a power the general performance objective and requirements that supply, signal lines, and an alarm monitor. Detection of an must be met through the establishment of a physical intruder is accomplished by the alarm system responding to protection system. Performance capabilities necessary to some change in its operating condition caused by the meet the requirements of §73.20 are described in §73.45. intruder, e.g., interruption of a transmitted infrared or Paragraph 73.45(c) requires that only authorized activities microwave beam or stress exerted on a piezoelectric crystal.

and conditions be permitted within protected areas, material The choice of a perimeter alarm system is influenced by access areas, and vital areas through the use of detection considerations of terrain and climate. At present, no single and surveillance subsystems and procedures to detect, perimeter intrusion alarm system is capable of operating assess, and communicate any unauthorized access or effectively in all varieties of environment.

penetrations or such attempts by persons, vehicles, or materials. Furthermore, §73.46 outlines typical specific The mode of installation of the perimeter alarm system safeguards measures that will often be included in an overall influences its effectiveness. In general, dividing the site system that meets the requirements of § § 73.20 and 73.45. perimeter into segments that are independently alarmed The use of an intrusion alarm subsystem with the capability and, uniquely monitored assists the security organization to detect penetration through the isolation zone is specifically responding to an alarm by localizing thearea in which the called out in paragraph 73.46(e)(1). For power reactors, alarm initiated. Segmenting of the perimeter'alarm system paragraph 73.55(c)(4) requires that detection of penetra also allows testing and maintenance of a portion of the tion or attempted penetration of the protected area or the system while maintaining the remainder of the perimeter isolation zone adjacent to the protected area barrier ensure under monitoring. It is generally desirable that the individual that adequate response by the security organization can be segments be limited to a length that allows observation of initiated. the entire segment by an individual standing at one end of the segment.

This guide describes six types of perimeter intrusion alarm systems and sets forth criteria for their perform Effective use of a perimeter intrusion alarm system is ance and use as a means acceptable to the NRC staff for facilitated by a regular program of system testing. Operability meeting specified portions of the Commission's regula testing can be performed by a guard or watchman penetrating tions. It also references a document (SAND 76-0554) that the segment protected by the alarm system during routine provides additional information in this area, especially on patrols. Performance testing, i.e., manufacturer's specifica the subject of combining sensors to yield a better overall tion testing and detection probability testing, however, is performance. usually more elaborate. In any case, testing can be conducted without compromising security only if performed under controlled circumstances such as direct visual observation or by closed-circuit television of the area being tested while Lines indicate substantive changes from Revision 1. a specified test is conducted.

USNRC REGULATORY GUIDES Comments should be sent to the Secretary of the Commission, U.S. Nuclear Regulatory Commission, Washington, D.C. 20555, Regulatory Guides are Issued to describe and make available to the Attention: Docketing and Service Branch.

public methods acceptable to the NRC staff of implementing specific parts of the Commission's regulations, to delineate tech The guides are issued in the following ten broad divisions:

niques used by the staff in evaluating specific problems or postu lated accidents or to provide guidance to applicants. Regulatory 1. Power Reactors 6. Products Guides are noi substitutes for regulations, and compliance with 2. Research and Test Reactors 7. Transportation them Is not required. Methods and solutions different from those set 3. Fuels and Materials Facilities 8. Occupational Health out in the guides will be acceptable If they provide a basis for the 4. Environmental and Siting 9. Antitrust and Financial Review findings requisite to the issuance or continuance of a permit or 5. Materials and Plant Protection 10. General license by the Commission.

Copies of issued guides may be purchased at the current Government Comments and suggestions for improvements in these guides are Printing Office price. A subscription service for future guides in spe encouraged at all times, and guides will be revised, as appropriate, cific divisions is available through the Government Printing Office.

to accommodate comments and to reflect new information or Information on the subscription service and current GPO prices may experience. This guide was revised as a result of substantive com be obtained by writing the U.S. Nuclear Regulatory Commission, ments received from the public and additional staff review. Washington, D.C. 20555, Attention: Publications Sales Manager.

To ensure secure operation, the system may periodically beam can cause nuisance alarms. Since the beam is wider monitor the sensor transducer and signal processing circuits. than other systems, care must be taken to ensure that,.-'

This self-checking feature can vary depending on the type authorized activities do not create nuisance alarms. Systems and design of the alarm system. Many systems require self using the Doppler shift for motion detection are especially excitation of the sensor transducer (e.g., vibration, strain, sensitive to the motion of trees and grass and to falling rain pressure) while others monitor the signal level at the'receiv and snow.

ing transducer (e.g., microwave, infrared). However, several worthwhile commercially available perimeter alarm systems The maximum and minimum separation of the transmitter provide little or no self-checking circuitry. To ensure normal and receiver is usually specified by the manufacturer.

operation for those alarm systems that do not incorporate Typically, a microwave perimeter alarm system will operate self-checking circuitry, the licensee should institute a test effectively in the range between 70 and 150 meters.

program that will periodically test each segment of a perim eter alarm system to verify that it maintains the proper sensitivity to detection. 2. E-Field Perimeter Alarm System In order to increase the probability of detection and lower An E-field perimeter alarm system consists basically of a the false alarm rates, a combination of sensors may be desir field generator that excites a field wire, one or more sensing able in certain environments. Additional factors to be con wires, and a sensing filter; an amplifier; and a discriminatory sidered in the selection and application of single sensors or and annunciator unit. The field wire transmits essentially an a combination of sensors are presented in a Sandia Labora omnidirectional E-field to ground. A large body approaching tories report prepared for the Department of Energy entitled the system changes the pattern of the E-fiel

d. When sensing

"Intrusion Detection Systems Handbook" (IDSH), SAND wires are placed at different locations within the transmitted

76-0554, and in particular Sections 8.3 and 3.2. Additional E-field pattern, they pick up any changes occurring in that information in this area, i.e., integrated perimeter systems, pattern. If the changes are within the frequency bandpass is scheduled for development by the NRC. An important of human movement, an alarm signal is generated. The field element of an intrusion detection system is the assessment wire and one or more parallel sensing wires can be either capability associated with the perimeter intrusion alarm connected to a chain link fence or mounted as an above system. Alternative assessment capabilities such as video ground, freestanding system of an isolation zone.

assessment, hardened observation posts, and armored response vehicles are discussed in Regulatory Guide 5.61, The E-field system can offer about 300 meters of perim- .

"Intent and Scope of the Physical Protection Upgrade Rule eter protection per segment, but shorter lengths of 100

Requirements for Fixed Sites," in the discussion of para meters are recommended in order to have effective alarm graph 73.46(h)(6). System design considerations for video assessment and response capabilities. The system can be assessment systems are discussed in Section 6.3 of the IDSH. mounted on metal, plastic, or wooden posts using specially designed electrical isolators that allow for small movements The following discussion describes the operations, limita of the posts without disturbing the field and sensing wires.

tions, and environmental considerations of six basic types Both the field and sensing wires need to be under a high of commercially available perimeter alarm systems: micro degree of spring tension to produce high-frequency vibra wave, E-field, ferrous metal detector, pressure-sensitive, tions when they are struck by small foreign objects or infrared, and vibration- or stress-fence protection systems. blown by the wind, both of which are out of the passband of the receiving circuitry. In addition, in order to keep the sensitivity of the system from varying, the E-field

1. Microwave Perimeter Alarm System detector needs to be well grounded.

Each link of a microwave perimeter alarm system is com The E-field detector is not a line-of-sight system and posed of a transmitter, receiver, power supply, signal pro therefore can be installed on uneven terrain and in an cessing unit, signal transmission system, and annunciator. irregular line. The surrounding terrain should be kept clear The microwave transmitter produces a beam-like pattern of of shrubs, tree limbs, and undergrowth since they act as microwave energy directed to the receiver, which senses the moving ground objects. The basic system is a two-wire microwave beam. A partial or total interruption of the beam system with the sensing wire located between 200 and 450

will cause an alarm. The microwave beam can be modulated to millimeters above the ground and the field wire located reduce interference from spurious sources of radiofrequency approximately 1 meter above and parallel to the sensing energy, to increase sensitivity, and to decrease the vulner wire. The width of the detection zone is variable and ability to defeat from "capture" of the receiver by a false depends to a large degree on the size of the target. Generally, microwave source. it is approximately 0.6 meter wide on either side of the field wire. To prevent an intruder from jumping over the Successive microwave links can be overlapped to form a top of the E-field detector, a second sensing wire can be protective perimeter around a facility. Since the transmitter/ installed approximately 1 meter above the field wire.

receiver link is a line-of-sight system, hills or other obstruc When installed on a chain link fence, standoffs approximately tions will interrupt the beam, and ditches or valleys may 0.5 meter long are used for mounting the wires. The E-field provide crawl space for an intruder. Moreover, objects such generated in this configuration does not penetrate the fence as tumbleweed, paper, and bushes moving in the path of the but parallels it.

5.44-2

3. Ferrous Metal Detector Perimeter Alarm System alarms. Features to compensate for wind-generated noise can be designed into the equipment but in turn may cause a A ferrous metal detector system consists of buried decrease in system sensitivity. Pressure systems will lose electrical cables, amplifiers, inhibitors, power supply, signal sensitivity when the buried sensors are covered by snow, by processing unit, signal transmission lines, and annunciator. snow with a frozen crust that will support the weight of a The system is passive and is susceptible to changes in the man, or by frozen ground. Other natural phenomena earth's ambient magnetic field. Such changes are caused such as hail and rain can cause nuisance alarms.

either by electromagnetic disturbances such as lightning or by ferrous metal being carried over the buried cables. The The sensitive area consists of a narrow corridor, usually change in the local ambient magnetic field induces a current about 1 meter in width. A greater degree of security can be in the buried cable which is filtered and sensed by the achieved by employing two such corridors to prevent an electronics. If the change exceeds a predetermined threshold, intruder from jumping over the buried transducers. A

an alarm is generated. To reduce nuisance .alarms from typical length monitored by a transducer (i.e., set of external electromagnetic sources (e.g., electrical power piezoelectric crystals, a liquid-filled tube, or an electrical transmission lines), the electrical cable is laid in loops that cable) is about 100 meters.

are transposed at regular intervals. Also, an inhibitor loop can be used to reduce nuisance alarms from electromagnetic 5. Infrared Perimeter Alarm System interference. The inhibitor, which operates on the same principle as the sensor cable loops and is buried near the Like the microwave system, each link of an infrared sensor cable, senses strong temporary electromagnetic system is composed of a transmitter, receiver, power interferences (e.g., lightning) and disables the alarm system supply, signal processor, signal lines, and alarm annunciator.

for approximately one second, thus reducing nuisance The transmitter directs a narrow infrared beam to a receiver.

alarms. If the infrared beam between the transmitter and receiver is interrupted, an alarm signal is generated. As with the The ferrous metal detector system is not a line-of-sight microwave system, the infrared system is a line-of-sight system and therefore can be installed on uneven ground in system. In addition, the infrared beam is usually modulated.

an irregular line. The sensor subloops formed by the cables Since the infrared beam does not diverge significantly as must be fairly regular, however. Since the system will detect does the microwave beam, multiple infrared beams between only ferTous metal, animals, birds, or flying leaves will not transmitter and receiver can be used to define a "wall." If initiate alarms. However, electromagnetic interferences this "wall" is then penetrated by an individual, an alarm can cause nuisance alarms or disable the alarm system when will result.

the interference is severe.

Fog both attenuates and disperses the infrared beam and Each sensing cable (and amplifier) can monitor a security can cause nuisance alarms. However, the system can be segment up to 500 meters in length. Increasing the length designed to operate properly with severe atmospheric of the security segment beyond 500 meters usually results attenuation. Dust on the faceplates will also attenuate the in a high nuisance alarm rate. Multiple cables and amplifiers infrared beam as will an accumulation of condensation, can be used to extend the monitoring length. frost, or ice on the faceplates.

4. Pressure/Strain-Sensitive Perimeter Alarm System Such condensation, frost, or ice, however, may be eliminated through the use of heated faceplates. Sunshine Buried pressure/strain transducers detect small variations on the receiver may cause an alarm signal. Misalignment of in the mechanical stress exerted on the surrounding soil by transmitter and receiver caused by frost heaves may also the presence of an individual passing above the sensor. The cause an alarm signal. Like the microwave system, vegetation signals produced by the transducers are amplified and such as bushes, trees, or grass and accumulated snow will compared with a preestablished threshold. If the signal interfere wlith the infrared beam, and ditches, gullies, or exceeds the threshold, an alarm occurs. The transducer may hills will allow areas where the passage of an intruder be a set of piezoelectric crystals, a fluid-filled flexible tube, may go undetected.

a specially fabricated stress/strain electrical cable, or an insulated wire in a metallic tube. The typical distance between transmitter and receiver is about 100 meters; some systems are capable of monitoring Like the ferrous metal detector system, the pressure a distance up to 300 meters under ideal conditions.

sensitive system does not require line-of-sight installation and can be sited on uneven terrain. However, soil condition 6. Vibration- or Strain-Detector Perimeter Alarm System and composition have a significant effect on sensor sensitivity.

Installation in rocky soil may result in damage to the A variety of devices that detect strain or vibration are pressure transducers either during installation or as a result available for use as fence protection systems. Although the of soil settlement after installation. Wind-generated move devices vary greatly in design, each basically detects strain ment in trees and poles can create nuisance alarms. High or vibration of the fence such as that produced by an winds can produce pressure waves on the ground surface intruder climbing or cutting the fence. In the simplest which, if sensed by the transducer, could necessitate devices, the vibration or strain makes or breaks electrical operation at reduced sensitivity in order to avoid nuisance continuity and thereby generates an alarm. Vibration- or

5.44-3

strain-detection devices for fence protection generally are All controls that affect the sensitivity of the alarm susceptible to nuisance alarms caused by wind vibrating the system should be located within a tamper-resistant enclosure fence or by hail stones or large pieces of trash blowing All signal lines- connecting alarm relays with alarm monito.

against the fence. The frequency of nuisance alarms due to should be supervised; if the processing electronics is separate-'r the wind can be reduced by rigidly mounting the fence and from the sensor elements and not located within the detection thereby lessening the propensity of the fence to vibrate in area of the sensor elements, the signal lines linking the sensors2 the wind. This situation is especially common with post to' the Processing electronics should also be supervised.

mounted switch-contact-type alarm systems. The use of electronic signal processing equipment in conjunction with All key locks or key-operated switches used to signal-generating strain transducers can effectively reduce protect equipment and controls should have UL-listed nuisance alarm rates without sacrificing sensitivity to locking cylinders (see Regulatory Guide 5.12, "General Use climbing or cutting the fence. However, most fence alarm of Locks in 'the Protection and Control of Facilities and systems can be easily bypassed by a variety of methods. Special Nuclear Materials").

Depending on the variety of sensor, each sensor can (3) Environment. Perimeter intrusion. alarm systems monitor a length of fence ranging from about I meter to should be capable of operating throughout the climatic several hundred meters. extreme of the environs in which they are used; as a mini mum, the outdoor systems should be capable of effective I

C. REGULATORY POSITION

operation between -35 0 C and +50 0 C. Components that necessarily must be located out of doors should be protected

1. Minimum Qualification for Perimeter Intrusion Alarm from moisture damage by such methods as hermetic sealing, Systems potting in an epoxy compound, conformal coating, or watertight enclosures.

a. General

(4) Alarm Conditions. Perimeter intrusion alarm sys

(1) Electrical. All components-sensors, electronic tems, whether using single or complementary sensors, should processing equipment, power supplies, alarm monitors generate an alarm or indication under any of the following should be capable of meeting the typical design require conditions:

ments for fire safety of nationally recognized testing laboratories such as Underwriters Laboratory (UL) or (a) Detection of stimulus or a condition for which Factory Mutual (FM). The system should contain provisions the system 'was designed to react, for automatic switchover to emergency battery and generator (b) Indication of a switchover to the emergency Ior emergency battery power without causing an intrusion system alarm in the event primary power is interrupted.

Emergency power should be capable of sustaining operation or secondary source(s) of power and also upon loss of emergency power, for a minimum of 24 hours2.777778e-4 days <br />0.00667 hours <br />3.968254e-5 weeks <br />9.132e-6 months <br /> without replacing or recharging batteries or refueling generators. If sufficient battery or fuel (c) Indication of tampering (e.g., opening, short capacity is not attainable for 24-hour operation as stated ing, or grounding of the sensor circuitry) which renders the above, additional batteries or fuel should be stored on site device "incapable of normal operation, expressly for augmenting the emergency power supply. If emergency power is furnished by battery, all batteries (d) Indication of tampering by activation of a (including stored batteries) should be maintained at full tamper switch or other triggering mechanism, charge by automatic battery-charging circuitry. Batteries should be checked in accordance with IEEE Standard (e) Failure of any component(s) to the extent

450-1975 as endorsed by Regulatory Guide 1.129, "Main that the device is rendered incapable of normal operation.

tenance Testing and Replacement- of Large Lead Storage Self-checking circuitry is normally used for detecting Batteries for Nuclear Power Plants," and IEEE Standard components'that have failed in a device.

308-1974 as endorsed by Regulatory Guide 1.32, "Criteria for Safety-Related Electric Power Systems for 'Nuclear Under normal environmental conditions, includ Power Plants." ing *seasonal extremes, the total perimeter alarm system should not average more than one false alarm per week per

(2) Tamper Indication. All enclosures for equipment segment and should not average more than one nuisance should be equipped with tamper switches or triggering alarm per week per segment while maintaining proper de mechanisms compatible with the alarm systems. The tection sensitivity. Where the segment can be fully observed electronics should be designed so that tamper-indicating at all times, either visually or by closed circuit television, devices' remain in operation even though the system itself the false alarm rate and nuisance alarm rate may be increased may be placed in the access mode. 1 to one alarm per day per segment. False alarms are defined as those alarms that have been generated without any appar ent cause. Nuisance alarms are alarms generated by an iden

1Access mode means the condition that maintains security tified input to a sensor or monitoring device that does not over the signal lines between the detector and annunciator and over 2Signal line supervision is discussed in NUREG-0320. "Interior the tamper switch in the detector but allows access into the protected area without generating an alarm. Intrusion Alarm Systems," issued in February 1978.

5.44-4

to obtain proper ground coverage. The distance between a represent a safeguards threat. Properdetection probability is transmitter and its receiver should be in accordance with defined as the ability to detect an intruder with at least 90%

the manufacturer's specifications and site-specific require probability for each segment of the isolation zone under ments. Neither the transmitter nor the receiver should be the conditions stated in the Performance Criteria of each mounted on a fence. To prevent passage under the microwave type of alarm system.

beam in the shadow of an obstruction, hills should be An automatic and distinctly recognizable indica leveled, ditches filled, and obstructions removed so that the area between transmitter and receiver is clear of obstructions tion should be generated by the alarm monitor upon and free of rises or depressions of a height or depth greater switchover to emergency power. Loss or reduction of than 15 cm. The clear area should be sufficiently wide to power (either primary or emergency) to the degree that the preclude generation of alarms by objects moving near the system is no longer operating properly, should also be microwave link (e.g., personnel walking orvehicular traffic).

indicated in the central alarm station.

Approximate dimensions of the microwave pattern should be provided by the manufacturer.

Placement of any portion of a perimeter intrusion alarm system into the access mode should be indicated auto If the microwave link is installed inside and matically and distinctly by the alarm monitor. Moreover, roughly parallel to a perimeter fence or wall, the transmitter the segment(s) of the system placed in the access mode and receiver should be positioned so as to prevent someone should be indicated clearly.

from avoiding detection by jumping over the microwave beam into the protected area from atop the fence or wall.

(5) Installation. It is recommended that perimeter Typically, a chain link security fence with an overall height intrusion alarm systems be located inside the perimeter of 2.4 meters will necessitate a minimum of 2 meters physical barrier at a distance that prohibits use of the barrier to illicitly traverse the alarm zone. If, however, installation between the fence and the center of the microwave beam.

is outside the perimeter barrier, a second barrier or a fence Successive microwave links and corners should (e.g., a cattle or snow fence) should be erected so that the overlap at least 3 meters to eliminate the dead spot (areas alarm system is located between the barriers. The second where movement is not detected) below and immediately in barrier or fence will serve to reduce the incidence of nuisance front of transmitters and receivers. The overlap of successive alarms from animals and passersby. The separation between links should be arranged so that receiver units are within the second barrier and the perimeter barrier should be the area protected by the microwave beam.

sufficient to preclude bridging of the perimeter alarm

':- system; in all cases, it should not be less than 6 meters.

Fence protection systems should be located on an inner c. E-Field Perimeter Alarm System fence.

Where possible, the perimeter should be segmented (1) PerformanceCriteria. An E-field perimeter alarm system should be able to detect an individual weighing a so that an individual standing at one end of a segment will minimum of 35 kilograms at least 0.5 meter from the have a clear view of the entire segment. In no case should sensing wire whether crawling and rolling under the lower any segment exceed 200 meters in length. Each segment sensing wire, stepping and jumping between the field and should independently and uniquely indicate intrusion and sensing wires, or jumping over the top sensing wire of the should be capable of placement into the access mode system. The field and sensing wires should be supervised to independently of the other segments.

prevent the undetected cutting or bypassing, of the system through electronic or clandestine means. The system design b. Microwave Perimeter Alarm System should employ techniques to minimize alarms caused by

(1) Performance Criteria. A microwave perimeter high winds, thunderstorm-related electrical phenomena, and small animals.

alarm system should be capable of detecting an intruder weighing a minimum of 35 kilograms passing between the transmitter and receiver at a rate between 0.15 and 5 meters per second, whether walking, running, jumping, crawling. ," (2) Installation Criteria. The E-field sensor should consist of a minimum of one field wire and two. sensing rolling. The beam should be modulated, and the reccr:

should be frequency selective to decrease susceptibility to wires. One sensing wire should be located no more than

0.45 meter above ground level with the second located receiver "capture." Generally, because of susceptibility to approximately 2.6 meters above ground level. The field wire motion beyond the area to be protected, monostatic should be located between the sensing wires approximately Doppler microwave systems should not be used as perim eter intrusion alarms. I meter above ground level. The surrounding terrain within

3 meters of E-field wires should be free of all shrubs, trees, and undergrowth. The control unit should be well grounded

(2) InstallationCriteria.The transmitters and receivers using a I-meter or longer grounding rod or equivalent elec should be installed on even terrain clear of trees, tall grass, trical ground. When mounted to a chain link fence, the fence

~ and bushes. Each unit :should be mounted rigidly at a distance of about 1 meter above the ground. Because of should also be weli grounded approximately every 23 meters variances in the antenna pattern of different microwave using a I-meter or longer grounding rod or equivalent elec trical ground.

systems, this height may have to be varied slightly in order

5.44-5

d. Ferrous Metal Detector Perimeter Alarm System (2) Installation Criteria. The sensors should be installed at the depth below the ground surface stated by

(1) Performance Criteria. A ferrous metal detector the manufacturer. To obtain a high probability of detection, perimeter alarm system should be able to detect a 400-pole the sensors should be in two separate parallel lines at a distance of 1.5 to 2 meters apart. The sensors and electronic centimeter (CGS units) magnet moving at a rate of 0.15 meter per second within a radius of 0.3 meter of a sensor circuitry buried in the ground should be of a durable, cable. The detection system should be equipped with moistureproof, rodent-resistant material. When a pressure inhibitor coils to minimize nuisance alarms due to electro sensitive perimeter alarm system is being installed in rocky soil, all rocks should be removed during backfilling to magnetic interference. No more than six sensing loops per prevent damage to sensors. If the frost line exceeds 10 cm, a inhibitor coil should be used in order to prevent simulta buried pressure-sensitive system should not be used unless neous desensitizing of the entire system.

the soil is specifically prepared to eliminate freezing above the sensor.

(2) Installation Criteria. To determine if the ferrous metal detection system will operate in the proposed environ f. Infrared Perimeter Alarm Systems ment, a preengineering site survey should be made using an electromagnetic detection survey meter. This survey meter (I) Performance Criteria. An infrared perimeter can be furnished by the manufacturer. If the electromagnetic alarm system should be a multibeam modulated type disturbances are within the limits prescribed .by the manu consisting of a minimum of three transmitters and three facturer, this type of system can be used effectively. Special receivers per unit. An infrared perimeter alarm system looping configurations can be made in areas of high electro should be capable of detecting an individual weighing a magnetic interference to reduce the incidence of nui sance alarms. minimum of 35 kilograms passing between the transmitters I

and receivers at a rate between 0.15 and 5 meters per The sensing loops of electrical cable should be second, whether walking, running, jumping, crawling, or rolling. Furthermore, the systems should be able to operate buried in the ground according to the manufacturer's stated depth. Multiple units (cable and amplifier) should be used as above with a factor of 20 (13db) insertion loss due to atmospheric attenuation (e.g., fog) at maximum range to protect a perimeter. All associated buried circuitry

(100 meters).

should be buried within the protected zone and packaged in hermetically sealed containers. The cable should be laid in

(2) Installation Criteria. An infrared perimeter alarm accordance with the manufacturer's recommended geometri cal configurations to reduce nuisance alarms from external system should be installed so that, at any point, the lowest beam is no higher than 21 cm above grade and the highest sources. When cable is being installed in rocky soil, care

.beam at least 2.6 meters above ground. Sufficient overlap should be taken to remove sharp rocks during backfilling over the cable. of beams should exist such that an individual could not intrude between the beams and remain undetected. The ground areas between the infrared beam posts should be Inhibitors should be buried in the ground at least prepared to prevent tunneling under the lower beam with

6 meters from the cable inside the protected perimeter.

in at least 15 cm of the surface. This may be accomplished by using concrete, asphalt, or a similar material in a path at Continuous electromagnetic interference obstructs least 1 meter wide and 15 cm deep or alternatively 15 cm the detection of an intruder carrying metal over the buried wide and 1 meter deep between tht posts.

cable by keeping the inhibitor activated, thereby preventing the alarm unit from responding to a change in flux caused by the intruder. The device should therefore be used only The transmitters and receivers should be mounted rigidly (e.g., installed on a rigid post or concrete pad) to where the environment is relatively free of severe man-made prevent nuisance alarms from vibrations. Each transmitter electromagnetic interference (e.g., overhead power cables, and receiver post should be provided with a pressure-sensitive pole-mounted transformers, generators). The cable should cap to detect attempts at scaling of or vaulting over the never be installed close to overhead power transmission infrared beam post. The maximum distance between lines. Moreover, the cable should be placed at least 3 meters transmitter and receiver should be selected to permit proper from parallel-running metal fences and at least 20 meters operation during conditions of severe atmospheric attenua from public roads to minimize nuisance alarms.

tion that are typical for the site, generally a maximum of

100 meters.

e. Pressure-Sensitive Perimeter Alarm System It is recommended that the infrared perimeter

(1) Performance Criteria. A .pressure-sensitive perim alarm system be installed inside the physical perimeter eter alarm system should be capable of detecting an indi

• vidual weighing more than 35 kilograms crossing the barrier with the transmitter and receiver units positioned a minimum of 3 meters from the barrier. Installation of the sensitive area of the system at a minimum speed of 0.15 meter per second, whether walking, crawling, or rolling. infrared alarm system inside and directly adjacent to the perimeter barrier should be avoided since the barrier may The system design should employ techniques (e.g., electronic provide a solid base from which an intruder can jump over I

signal processing) to eliminate nuisance alarms from wind the beams into the protected area.

I and other adverse environmental phenomena.

5.44-6

g. Vibration or Strain Detection b. Performance Testing This vibration- or strain-detection system should be At least quarterly, i.e., once each 93 calendar days, used only as a secondary or backup perimeter alarm system after each inoperative state, and after any repairs, the except when one of the other five types of perimeter alarm perimeter intrusion alarm system should be tested against systems will not work (e.g., because of the environment) its manufacturer's design specifications and for proper I and after the NRC's approval has been received. If there is a detection probability. An inoperative state for an alarm need to use this system, the following criteria should apply: system or component exists when (1) the power is discon nected to perform maintenance or for any other reason, (2)

(1).Performance Criteria. Vibration- or strain-detec both primary and backup power sources fail to provide tion systems used for fence protection should detect an power, and (3) when power is applied and one or more intrude1r weighing more .than 35 kilpgrams attempting to components fail to perform their intended function. Placing climb the fence. The system sl~ould also detect any attempt a properly operating alarm system in the access mode to cut the fence or lift the fence more than 15 cm above would not constitute an inoperative state unless accompany grade. The system should not generate alarms due to wind ing or followed by any of the above three conditions.

vibration of the fence from a wind force of up to 48 kilometers/hour. (1) Specification Testing. The test procedure tecom mended by the manufacturer should be followe

d. While the

(2) Installation Criteria.The vibration or strain sensors test is being conducted, the area under test should be should be attached firmly to the fence (post or fabric, as maintained under visual observation by a member of appropriate) so that the vibration/stress caused by an intruder the security organization. For all perimeter systems, tests climbing, cutting, or lifting the fence will generate an alarm. should be conducted to verify that no obvious dead spots exist in the segment of protection. As a minimum, the tests

2. Testing of Perimeter Intrusion Alarm Systems should include line supervision and tamper proofing when testing in both the access and secure modes. If the perimeter All tests and test results should be documented. The docu alarm system does not meet the manufacturer's specifica mented test results will establish the performance history of tions, corrective actions should be taken and documented.

each perimeter alarm system and each segment of the isola tion zone. The test results should be available for inspection (2) Detection Probability Testing. Proper detection and analysis. probability is defined as the ability to detect an intruder with at least 90% probability in each segment of the isolation a. Operability Testing zone, .with 95% confidence, under the conditions stated in the Performance Criteria of each type of alarm system.

Perimeter intrusion alarm systems should be tested on While the detection probability testing is being conducted, all segments of the isolation zone at least once each 7 days. the area under test should be maintained under visual Testing may be conducted during routine patrols by the observation by a member of the security organization. One members of the licensee security force. The testing should sample testing method for demonstrating compliance with be conducted by crossing the segment of the isolation zone detection probability and confidence levels is given in the where the alarm system is located or by climbing the fence detection probability testing section of Appendix A to this to which the system is attached to provide the required alarm guide. Other testing methods may be used if the methods stimulus. Where appropriate, a specific test procedure should are fully documented and approved by the NRC.

be followed. Prior to making the test, the individual making the test should notify the central alarm station that a test is

D. IMPLEMENTATION

about to be conducted. The area under test should be main tained under visual observation by a member of the security The purpose of this section is to provide information to organization. applicants and licensees regarding the NRC staff's plans for using this regulatory guide.

All segments of the isolation zone should be tested in a different, preferably random, order every 7 days and the Except in those cases in which the applicant or licensee testing should be conducted throughout the week, not all proposes an acceptable alternative method, the staff will tests on I day. The operability testing should result in use the methods described herein in evaluating an applicant's

100% detections on all segments each 7 days. If the perimeter or licensee's capability for and performance in complying alarm system fails to detect an intrusion on one or more with specified portions of the Commission's regulations segments, corrective actions should be taken and documented. after April 1, 1980.

See the operability testing section of Appendix A to this guide for a sample method for determining the testing order If an applicant or licensee wishes to use the method for the segments and a suggested method for determining if described in this regulatory guide on or before April 1, the detection rate of the perimeter alarm system has decreased 1980, the pertinent portions of the application or the to below 90%. Other testing methods may be used if the licensee's performance will be evaluated on the basis of methods are fully documented and approved by the NRC. this guide.

5.44-7

VALUE/IMPACT STATEMENT

ysis prepared for the proposed amendments was made Room at A separate value/impact analysis has not been prepared available in the Commission's Public Document The This for the proposed revision to this regulatory. guide. the time the proposed amendments were published.

the as well as changes were made to make the guide consistent with analysis is appropriate for the final amendments to the regula guide revisions appropriate to those upgraded physical protection amendments for the regulatory Register of tions published in final form in the Federal amendments.

28, 1979 (44 FR 68184). A value/impact anal- November

5.44-8

APPENDIX A*

EXAMPLES OF TESTING METHODS FOR

PERIMETER INTRUSION ALARM SYSTEMS

6. Jumping - Leaping from a height above the zone of BACKGROUND detection to a point at ground level across the zone of detection, e.g., standing on the fence and attempting to The purpose of this appendix is to provide an example of a leap across the zone of detection.

testing method to determine detection capability of perim eter intrusion alarm systems. This example should not be 7. Rolling - Entering and leaving the zone of detection interpreted as a regulatory requirement. Other testing meth prone to the.ground with a low profile, parallel to the ods for determining compliance with detection probability zone of detection, and rolling slowly at an approximate and confidence levels may be used if fully documented and velocity of 0.15 meter per second.

approved by the NRC. The purpose of testing a perimeter in trusion alarm system is to ensure that the installed system is operating according to the three testing criteria stated belo

w. TESTING

1. Operability Testing - Paragraph C.2.a of this guide states: Operability Testing

"Perimeter intrusion alarm systems should be tested on all segments of the isolation zone at least once each Operability testing is a check to ensure that the alarm

7 days.... The operability testing should result in 100%

system is operating and that the detection sensitivity of the detections on all segments each 7 days."

alarm system has not decreased from the 90% detection rate. The perimeter alarm systems should be tested on each

2. Specification Testing - Paragraph C.2.b of this guide segment of the isolation zone at least once during a 7-day states: "At least quarterly, ... the perimeter intrusion period. For example, the guard may violate the detection alarm system should be tested against its manufacturer's field by walking through the sensitive zone. The ordering of design specifications ..."

the tests on the segments should be in a different, prefer ably random, order each week, and the testing should be

3. Detection ProbabilityTesting, Paragraph C.2.b(2) states:

conducted throughout the wee

k. For an example of

"Proper detection probability is defined as the ability to randomizing the segments, assume that there are 10 seg detect an intruder with at least 90% probability in each ments and 21 shifts per week (3 shifts per day and 7 days segment of the isolation zone, with 95% confidence ... "

per week). Select at random (using a random number table or a random number generator) 10 of the shifts out of the DEFINITIONS

21 possible shifts, retaining the order in which the shifts were drawn. Then pair these 10 shifts with the segments I

In order to ensure uniform testing, the following terms through 10. In this example, let the 10 shifts selected be 6, are defined:

14, 9, 6, 20, 16, 19, 18, 10, 7.

I. Zone (Isolation Zone) - The entire perimeter adjacent to Table 1 the protected erea.

Shift No. Segment No.

2. Segment - A portion of the isolation zone that is inde pendently alarmed and monitored. 6 1

14 2

3. Running - Entering and leaving the zone of detection at 9 3 an approximate velocity of 5 meters per second. 6 4

20 5

4. Walking - Entering and leaving the zone of detection 16 6 with a normal stride. 19 7

18 8

5. Crawling - Entering and leaving the zone of detection by 10 9 lying prone to the ground, perpendicular to the zone of 7 10

detection, with a low profile at an approximate velocity of 0.15 meter per second.

The segment to be tested on each day of the week and the specific shift (1, 2, or 3) can be seen more clearly by Although this appendix is a substantive addition to Revision 2, reorganizing this information (see Table 2).

no lines are added in the margin.

5.44-9

Table 2 manufacturer's specifications, the recommended ýi, ions include retesting and calling the manufacturer's represen ,tive for repairs or upgrading of the system.

Shift No. Day - Shift Segment No.

Mon.- I None Detection Probability Testing

1 None

2 Mon. - 2 Mon. - 3 None., The following is one example of a method for detc, tion

3 Tues. - I None probability testing:

4 Tues. - 2 None

5 Tues. - 3 1,4 i. Determine the most vulnerable area of each segment,

6 Wed. - 1 10 -and determine the method of approach most likely to

7 Wed. - 2 None penetrate that segment, i.e., walking, running, jumping,

8 Wed. - 3 3 crawling, rolling, or climbing. This determination will, in

9 Thurs. - 1 9 most cases, be terrain dependent.

10

Thurs. - .2 None

11 Thurs. - 3 None 2. Test all segments using all the applicable penetration

12 Fri.- I None approaches at the most vulnerable area 30 times initially,

13 Fri.- 2 2 after installing a new system, after repairing or upgrading

14

15 Fri..- 3 None the system, or after the system failed to meet the mini Sat.- I 6 mum number of the successful detection criterion

16 Sat. - 2 None given below. All 30 tests must have resulted in successful

17 Sat. - 3 8 detections of the intrusion in order to have at least a

18 Sun.- I 7 90% probability of detection, with 95% confidence.

19 Sun. - 2 5

20

Sun. - 3 None If the minimum number of successful detections is not'

21 achieved, -the system should be checked. If no problems tests with the system are discovered, 10 more tests should be The testing could be conducted such that no shift made and if the minimum number of successful detections less more than one segment if ,the number of segments issible is achieved for the new number of tests (given in Table 4),

than the number of shifts. There are many other pos the in this case 39 out of 40, the testing can be ended for methods for ordering the segments, depending, on this segment for this quarter. If no problems with the nple, number of segments and the number of shifts. For exar system can be discovered and the minimum number of thod if there are more segments than shifts,.the ordering me nent. successful detections is not achieved after one more tef could require that each shift test. at least one segn of 10 intrusions, the system would need to be upgrade,,.

ilure to increase the detection probability to the required The test results should be documented on a success/fa level. If problems with .he systems are discovered, the ctive basis. If the test on a segment results in a failure, corre le, if system should be repaired and 30 new tests performed. If actions should be taken and documented. For examp 'stem there are 30 successful detections, testing can be ended.

the test of a segment results in no alarm, the alarm syas an should be checked for an obvious problem .such times For the subsequent tests at 90-day intervals, each incorrect setting and should be retested four more t tests segment should be tested 10 times. Each segment should during the same shift if possible. If all four of these Id be show at least 9 successful detections out of 10 approaches result in alarms, the alarm system on the segment shou five and.the cumulative results (combining the present results tested five more times on. the next day. If all- these this with the results from previous quarters) should have at r

tests result in alarms, the weekly testing schedule fo can least the minimum number of successful detections given e

segment can be resumed since the 90% detection rat nine in Table 4.

be confirmed. If any failures* occurred during the need additional tests, the alarm system for the segment will A.,-, Table 4 to be thoroughly checked, repaired, and retested accoJLstrate5 to the detection probability testing method to demon ecting Total No. Minimum No. of Maximum No. of that the alarm system for the segment is now dete 95% Successful Detections Failures to Detect of Tests intrusions with at least a 90% detection rate, with-4-11)

confidence. A table similar to Table 3 (see page 5.4 30 0

30

may be used for recording the test results. 40 39 1

48 2

50

Specification Testing 57 3

60

67 3 design 70

The licensee should conduct a manufacturer's re the 76 4

80

specification test of the system under test beforon all 85 5

90

detection probability tests have been conducted hould 95 5

100

segments and the results documented. The licensee snufac 104 6

110

6 follow the test procedures recommended by the maet the 120 114 turer of that system. If the system does not me'

5.44-10

Table 3 OPERABILITY TESTING RESULTS

(Success = 1, Failure = 0)

Week x, Quarter y, 19zz Environmental Conditions Result 4 Retests 5 Retests Date Time Segment I

Segment22- , , '

Segment 3 , ' .

• Attempt all applicable penetration approaches for a man One of the problems in testing intrusion-detection on-the-ground target. The penetration approach most systems is the need for a large number of tests to be performed likely not to be detected should be attempted more on each segment to estimate well the probability of detec frequently if an equal number of tests per approach tion in each segment. One example of a method to be used is not possible. For example, if the applicable penetra to avoid performing a large number of tests on each segment tion approaches for a given segment in the system are each quarter is to use an empirical Bayesian approach to running, walking, and crawling, the 10 quarterly tests estimate the probability of detection. The empirical Bayesian would be divided among the 3 approaches. If crawling method' combines the present quarter's data with those has the worst detection record, running would be of previous quarters. Using the empirical Bayesian method, attempted three times, walking three times, and crawling the performance criterion can be tested without a large four times. number of tests being performed each quarter.

4. Randomize the order in which the segments are tested. For the total number of tests less than 100 on each Randomization is a means of ensuring that environmental segment, the performance criteria are relaxed to be "at least effects and other unknown factors that may affect the 88% probability of detection in a segment with 95% con test results (detection or nondetection) do not always fidence." When the number of tests is 100 or more, the favor or handicap the same segment or method of performance criterion of "at least 90% probability of approach. For example, if Segment 1 is always tested in detection in a segment with 95% confidence" is used.

the morning and Segment 2 is always tested in the afternoon and if the detection equipment is slightly Table 6 gives the probability statements for the number more sensitive to intrusions in the morning, the conclu of tests between 30 and 120 with a given minimum number sion might be drawn, based on the test results, that of successful detections.

Segment 2 is less protected than Segment 1. However, the difference noted between the two segments might Table 6 be due only to the morning vs. afternoon difference.

Similarly, by randomizing the methods of approach, no Statement:

The probability of approach will be continually favored if the time sequence Table No. Minimum No. of detection is at least _%

(ordering) affects the test results. Randomization is of Tests Successful Detections with 95% confidence protection against disturbances that may or may not occur and that may or may not be serious if they do

30 30 90.5 occur. Randomization can be accomplished by using a

40 39 88.7 random numbers table to assign the order in which the

50 48 87.9 segments will be tested.

60 57 87.6

70 67 89.3

5. Maintain records of the results of all tests performed.

80 76 88.9 Included in these records should be the segment number,

90 85 88.7 date, time, and relevant environmental conditions when

100 95 89.8 tests were performed. Table 5 (see page 5.44-13) provides

110 104 89.6 a suggested format for recording the test result

s. The test

120 114 90.4 results in the "Overall" (totals) row in the columns headed (b), (c), (bW), and (c') are the important summary values.

For the initial testing or when retesting the perimeter For example, one is 95% sure that the probability of alarm system after it has failed to meet the minimum detection is at least 89.8% for the test results of 95 successful number of successful detections given in Table 4, the (b) detections out of 100 tests, i.e., the lower 95% confidence and (c) values should be 30 and 30, or 39 and 40, or 48 limit for the probability of detection is 89.8%.

and 50. For the subsequent quarterly testing, (b) must be 9 or 10 and (c) is 10 and (b') must be at least the Appendix B to this guide gives the details for deriving number under "Minimum No. of Successful Detections" these statements. Table 1 in Appendix B gives the probabil for the (cW)value ("Total No. of Tests") in Table 4. ity statements associated with all the numbers of successful detections out of the total number of tests performed that result in at least a 90% probability of detection with a 95%

Detection Probability Statements confidence level. The total number of tests covered in this table range from 30 to 120 in increments of 10 tests.

One method for assessing the probability of detection of the entire detection system is to use the "chain model," Using Table 1 in Appendix B, stronger statements can be i.e., the weakest "link" in the system determines the made about the probability of detection for the number of probability of detection for the system. In this case, the approach to a particular segment that has the lowest probability of detection would equal the probability of 1For a discussion of Bayesian methods, see H. F. Martz, Jr., and detection for the system. This is a "worst case" approach; R. A. Waller, "The Basics of Bayesian Reliability Estimation from however, it is the vulnerable areas of the system that need Attribute Test Data," Los Alamos Scientific Laboratory Report to be discovered and eliminated. LA-6126, February 1976.

5.44-12

( (

Table 5 DETECTION PROBABILITY TESTING RESULTS

Date:

Time:

Environmental Conditions:

Data Combined from Quarter s, 19tt to Segment x Quarter y,.19zz Data Only Quarter y, 19zz (a) (b) (a') (b')

No. of No. of (a)+(b)=(c) (b)/(c) Combined No. Combined No. (a')+(b')=(c') (b')/(c')

Method of Failures Successful Total No. Prob. of of Failures of Successful Combined Total Combined Prob.

Approach to Detect Detections of Tests Detection to Detect Detections No. of Tests of Detection

~ Running -

Walking -

Crawling -

Jumping Rolling -

Climbing -

Overall --

successful detections greater than the minimum number. For Table 7 example, if there were 98 detections out of 100 tests, one should state: 'The probability of detection is at least 93.8% with 95% Overall Probability Probability of confidence." Quarter of Detection Detecting Crawling In addition to the overall lower confidence limit on the 1st (initial) 30/30 = 1 6/6 = 1 probability of detection for a segmentconsidered previously, 2nd 39/40 7/8 = .875 a point estimate can be computed for the probabilities of 3rd 48/50 8/10= .8 detection for each method of approach for each segment, as 4th 57/60 9/12 .75 well as a point estimate for the overall probability of detection for each segment. The point estimate of a probabil ity of detection is the number of successful detections high likelihood of not being detected. Additional testing divided by the total number of tests of the type being should be performed to verify that the particular approach considered. Note that these point estimates are different is a system weakness, not random failures that coincidentally from the lower 95% confidence limits discussed previously. occurred in the same method of approach. If the weakness is The benefit of computing point estimates for each method verified, it should be eliminated, perhaps by increasing the of approach in each segment is to recognize a segment that sensitivity of the detector or by installing an additional may be particularly vulnerable to a specific method of device to detect this type of approach with a higher probabil approach. The concept is to look for trends occurring in ity. If, on the other hand, the failures of detection come from the data. For example, if all or most of the failures to detect varying approaches and if the overall probability of detection in a segment are in one method of approach, this segment in the segment is sufficiently high, i.e., the maximum number should be suspected as being vulnerable to this method of of failures to detect for the total number of tests is not approach. As a specific example, let the initial 30 tests be 6 exceeded, no specific weakness is indicated for this segment.

tests each of running, walking, crawling, jumping, and rolling.

Assume that no failures to detect intrusion occurred. The Caution: When the data indicate a problem with the de point estimate for the overall probability of detection is tection system and the problem is corrected, do

30/30 = 100%; the point estimate for the probability of de not combine (sum) the next quarter's data with tection for a crawling approach is 6/6 = 100%. Let the sub the data from previous quarters for the problem sequent quarterly tests be two tests each of the five methods segment. Begin accumulating the data again for of approach. In the next three quarters, assume that one this segment, starting, with the 30 tests from the failure to detect occurred in a crawling approach. Table 7 current quarter's testing that were conducted below gives the point estimates for the overall probability after correcting the problem.

of detection and for the crawling approach.

A table similar to Table 5 can be used for recording and Note that the minimum number of successful detections reporting the test resultsfor each method of approach, each are achieved for the total number of tests and 9 successful segment, and each quarter. The date and time of day and detections are achieved for the 10 quarterly tests. However, relevant environmental conditions such as weather, micro by xomputing the point estimates for each method of approach wave field intensity, E-field intensity, and changes in light the trend can be seen that a crawling approach has a fairly level should be recorded.

5.44-14

APPENDIX B*

CALCULATING THE CONFIDENCE LIMIT ON THE DETECTION PROBABILITY

Assume a binomial model tor the number of successful using F.05(6,96) 2.19.

detections, i.e., the probability of a successful detection is a fixed value, designated "p", and the tests for detection are 2. For x = 95 successes and n = 100 tests, independent. Let the number of tests performed be "n" 95' _ 95 = 89.79%,

and the number of successful detections "x".

95+6(1.80) 105.8 The point estimate of p, 0, is x/n.

using F. 0 5 (12,190) 1.80.

However, the problem is to obtain a confidence interval for p, which in this case is a lower one-sided 95% confidence 3. For x = 98 successes and n = 100 tests, limit.

98 = 98 = 93.85%,

The normal approximation to the binomial distribution 98 + 3(2.14) 104.42 is a valid approximation only when nt and n(l - P) are both equal to or greater than 5. For example, for the perform using F. 0 5 (6,196) 2 2.14.

ance criterion of 48 successes out of 50 tests, n(l -)

equals 2. Also, when there are no failures in detection, it is pot possible to use the normal approximation since var(ft) Table 1 gives the lower 95% confidence limits for the nft(l - P) = 0. probability of detection for n = 30, 40, 50, 60, 70, 80, and

90 beginning with x values such that the lower confidence The exact lower 95% confidence limit on p is given by limit is approximately equal to 88%; and for n = 100, 110,

and 120 beginning with x values such that the lower con x fidence limit is approximately equal to 90%. The lower x+ (n - x + 1) F.F0 5 (2n - 2x + 2,2x)] confidence limits for n = 30, 40, and 50 were abstracted from "Percentage Points of the Incomplete Beta Function,"

where F 0 5 (a,b) is the value of the F distribution with "a" Robert E. Clark, Journal of the American Statistical Asso and "b"' degrees of freedom which leaves 5% in the upper ciation 48: 831-843 (1953). The lower confidence limits tail of the distribution. for n = 60, 70, 80, 90, and 100 were abstracted from

"Tables of Confidence Limits for the Binomial Distribu Three examples given in. Appendix A to this guide can be tion," James Pachares, Journal of the American Statistical derived as follows: Association 55: 521-533 (1960). The lower confidence limits for n = 110 and 120 were computed using Formula (1).

I, For x = 48 successes and n = 50 tests, Clark's article gives confidence limits for all values of n

48 = 48. = 87.96%, from 10 to 50' for all values of x from I to n. Pachares'

48 + 3(2.19). 54.57 article gives confidence limits for values of n from 55 to

100 in increments of 5 for all values of x from I to n.

The confidence limits for any values of n and x can be Although this appendix is a substantive addition to Revision 2, no lines are added in the margin. computed using Formula (1).

5.44-15

Table 1 LOWER 95%CONFIDENCE LIMITý FOR p No. of Statement:

No. Successful The probabilit of detection of Tests Detections is at least %with 95% confidence.

n = 30 x = 30 90.5 n = 40 x = 39 88;7

40 92.8 n = 50 x = 48 87.9

49 90.9

50 94.2 n=60 x=57 87.6

58 89.9

59 92.3

60 95.1 n= 70 x= 67 89.3

68 91.3

69 93.4

70 95.8 n 80 x=76 88.9

77 90.6

78 92.3

79 94.2

80 96.1 n= 90 x 85 88.7

86 90.1

87 - 91.6

88 93.2

89 94.8

90 96.7 n= 100 x=95 89.8

96 91.1

97 92.4

98 93.8

99 95.3

100 97.01 n=110 x= 104 89.6

105 90.7

106 91.9

107 93.1

108 94.4

109 95.7

110 97.3 n 120 x 114 90.4

115 91.4

116 92.5

117 93.7

118 94.8

119 96.1

120 97.5

5.44-16

UNITED STATES

FIRST CLASS MAIL

NUCLEAR REGULATORY COMMISSION POSTAGE b FEES PAID

WASHINGTON, D.C. 20555 USNRC

PERMIT No. G-67 OFFICIAL BUSINESS

PENALTY FOR PRIVATE USE, $300