【文章內(nèi)容簡介】 were applicable to sprinkler and smoke detection systems reliability were analyzed，with descriptive estimates of the mean values and 95 percent confidence intervals for the operational reliability of these in situ systems reported. ELEMENTS OF RELIABILITY ANALYSIS There is considerable variation in reliability data and associated analyses reported in the literature. Basically，reliability is an estimate of the probability that a system or ponent will operate as designed over some time period. During the useful or expected life of a ponent，this time period is “reset” each time a ponent is tested and found to be in working order. Therefore，the more often systems and ponents are tested and maintained，the more reliable they are. This form of reliability is referred to as unconditional. Unconditional reliability is an estimate of the probability that a system will operate “on demand.” A conditional reliability is an estimate that two events of concern，.，a fire and successful operation of a fire safety system occur at the same time. Reliability estimates that do not consider a fire event probability are unconditional estimates. Two other important concepts applied to operational reliability are fuiledsafe and failed dangerous. when a fire safety system fails safe，it operates when no fire event has occurred. A mon example is the false alarming of a smoke detector. A fire safety system fails dangerous when it does not function during a fire event. In this study，the faileddangerous event defines the Operational probability of failure (1reliability estimate). A sprinkler system not operating during a fire event or an operating system that does not control or extinguish a fire are examples of this type of failure. The overall reliability of a system depends on the reliability of individual ponents and their corresponding failure rates，the interdependencies of the individual ponents that pose the system，and the maintenance and testing of ponents and systems once installed to veri@ operability. All of these factors are of concern in estimating operationaz reliability. Fire safety system performance is also of concern when dealing with the overall concept of reliability. System performance is defined as the ability of a particular system to acplish the task for which it was designed and installed. For example，the performance of a fire rated separation is based on the construction ponent’s ability to remain intact and provide fire separation during a fire. The degree to which these ponents prevent fire spread across their intended boundaries defines system performance. Performance reliability estimates require data on how well systems acplish their design task under actual fire events or full scale tests. Information on performance reliability could not be discerned directly from many of the data sources reviewed as part of this effort due to the form of the presented data，and therefore，it is not addressed as a separate effect. The cause of failure for any type of system is typically classified into several general categories： installation errors，design mistakes，manufacturing/equipment defects，lack of maintenance，exceeding design limits，and environmental factors. There are several approaches that can be utilized to minimize the probability of failure. Such methods include (1) design redundancy，(2) active monitoring for faults，(3) providing the simplest system (.，the least number of ponents) to address the hazard，and (4) a well designed inspection，testing，and maintenance program. These reliability engineering concepts are important when evaluating reliability estimates reported in the literature. Depending on the data used in a given analysis，the reliability estimate may relate to one or more of the concepts presented above. The literature review conducted under the scope of this effort addresses these concepts where appropriate. Most of the information that was obtained from the literature in support of this paper were reported in terms of unconditional operationaZ reliability，.，in terms of the probability that a fire protection strategy will not faiZ dangerous. LITERATURE REVIEW A literature search was conducted to gather reliability data of all types for fire safety systems relevant to the protection strategies considered： automatic suppression，automatic detection，and partmentation. The objective of the literature search was to obtain systemspecific reliability estimates for the performance of each type of fire safety system as a function of generic occupancy type (.，residential，mercial，and institutional). Sources of information included national fire incident database reports，US Department of Defense safety records，industry and occupancy specific studies，insurance industry historical records and inspection reports documented in the open literature，and experimental data Reports on experimental work and fire testing results were utilized only when fire detection，automatic suppression，or partmentation strategies were explicitly evaluated. Tests of systems used for qualification，approval，or listing were also reviewed for information on failure modes. Published data from the United Kingdom，Japan，Australia，and New Zealand were included. General Studies Several broad based studies were identified that reported reliability estimates for fire detection and fire suppression systems as well as construction partmentation. These included (1) the Warrington Fire Research study [1996] in the United Kingdom，(2) the Australian Fire Engineering Guidelines [Fire Code Reform Center，19961，(3) a pilation of fire statistics for Tokyo，Japan [Tokyo Fire Department，19971，and(4)results from a study of in situ performance of fire protection systems in Japan [Watanabe， Warrington Fire Research study addressed the reliability of fire safety systems and the interaction of their ponents. A Delphi methodology was used to develop dis