Advanced fire detection using multi-signature alarm algorithms

The objective of this work was to assess the feasibility of reducing false alarms while increasing sensitivity through the use of combined conventional smoke detectors with carbon monoxide (CO) sensors. This was accomplished through an experimental program using both real (fire) and nuisance alarm sources. A broad selection of sources was used ranging from smoldering wood and flaming fabric to cooking fumes. Individual sensor outputs and various signal-conditioning schemes involving multiple sensors were explored.The results show that improved fire-detection capabilities can be achieved over standard smoke detectors by combining smoke measurements with CO measurements in specific algorithms. False alarms can be reduced while increasing sensitivity (i.e., decreasing the detection time for real fires). Patented alarm criteria were established using algorithms consisting of the product of smoke obscuration and the change in CO concentration. Alarm algorithms utilizing ionization detector smoke measurements proved to be more effective than measurements from photoelectric detectors.     

Investigation of Multi-Sensor Algorithms for Fire Detection

There is widespread interest in the development of advanced fire detectors. A primary objective of fire detection is to provide prompt indication of the presence of a fire, without responding to deceptive signatures from nuisance sources. The principal purpose of this project is to identify the characteristics of a discriminating fire detector for Naval shipboard applications incorporating ionization, photoelectric, carbon dioxide and carbon monoxide sensors. Test data from previously conducted full-scale tests involving fire and nuisance sources are being analyzed to develop an algorithm involving combinations of the magnitude or slope of the response signal from each sensor. Acceptability of a particular algorithm is judged based on the number of correct classifications (fire vs. nuisance) and response time to fire sources.     

A Relative Time Analysis of the Performance of Residential Smoke Detection Technologies

A statistical study was conducted to compare the performance of different residential smoke detector technologies when exposed to different fire types. In order to facilitate comparisons between different fire and smoke growth rates, a non-dimensional smoke detector activation relative time was employed. Data from four major experimental studies was analyzed utilizing the relative time approach. The Common Language Effect Size, a measure of the probability that a particular detector technology will be the first to detect a fire of a particular type, was used to assess detector performance. The analysis confirmed previous results that ionization detectors, on average, respond faster to flaming fires, and that photoelectric detectors, on average, respond faster to smoldering fires. More importantly, this study also determined that the responses of ionization, photoelectric, and combination technologies are statistically equivalent for any given future residential fire That is, it cannot be determined with confidence which detector technology will alarm first to the next fire. Additionally, the analysis found that this statistical equivalence between detector technologies has not changed in the last 35 years despite increased fire growth rates associated with changes in furniture materials over that same time span.     

Video image fire detection for shipboard use

The objective of this work was to evaluate the effectiveness of commercial video image fire detection systems for small, cluttered spaces as would be found on Navy ships. The primary goal was to establish an understanding of the performance sensitivity and limitations of the video image detection (VID) systems to various setup and environmental conditions that may occur onboard ship while exposed to a range of flaming and smoldering fire sources and potential nuisance alarm sources. The response of the VID systems was benchmarked against standard fire alarm systems using addressable ionization and photoelectric smoke detectors.     

Laboratory evaluation of smoke detectors for use in underground mines

Laboratory experiments were conducted to determine the responses of a prototype smoke detector and a commercially available photoelectric smoke detector to smoke particles generated from various combustion sources. The prototype smoke detector combines optical scattering measurements with ionization chamber measurements in order to reduce/eliminate nuisance alarms due to the presence of airborne dusts or diesel exhaust particles. The commercially available smoke detector is designed for use in harsh environments where airborne dust represents a major problem due to both nuisance alarms and detector contamination. In the experiments, the responses of the two detectors were measured when exposed to smoke particles from the exhaust of a diesel engine and from a variety of fire sources, including wood, coal, styrene butadiene rubber, and No. 2 diesel fuel. For the solid fuels, data were obtained for both smoldering and flaming combustions. This report describes the experiments, their results, and the use of these results as they apply to early-warning fire sensors capable of the rapid and reliable detection of fires in atmospheres that may or may not be contaminated by either airborne dust or the products produced from diesel engines.     

Performance of Dual Photoelectric/Ionization Smoke Alarms in Full-Scale Fire Tests

Data from two full-scale residential smoke alarm fire test series were analyzed to estimate the performance of dual sensor photoelectric/ionization alarms as compared to co-located individual photoelectric and ionization alarms. Dual alarms and aggregated photoelectric and ionization alarm responses were used to estimate dual alarm performance. It was observed that dual alarms with equivalent or higher sensitivity settings performed better than individual photoelectric or ionization alarms over a range of flaming and smoldering fire scenarios. In one test series, dual alarms activated 539 s faster than ionization alarms and 79 s faster than photoelectric alarms on average. In another test series, individual alarm sensor outputs were calibrated against a reference smoke source in terms of light obscuration over a path length (percent smoke obscuration per unit length) so that alarm thresholds could be defined by the sensor outputs. In that test series, dual alarms, with individual sensor sensitivities equal to their counterpart alarm sensitivities, activated 261 s faster on average than ionization alarms (with sensitivity settings of 4.3%/m smoke obscuration for the ionization sensors) and 35 s faster on average than the photoelectric alarms (with sensitivity settings of 6.6%/m, for the photoelectric sensors.) In cases where an ionization sensor was the first to reach the alarm threshold, the dual alarm activated 67 s faster on average than the photoelectric alarm. While in cases were a photoelectric sensor was the first to reach the alarm threshold, the dual alarm activated 523 s faster on average than the ionization alarm. Over a range of ionization sensor settings examined, dual alarm response was insensitive to the ionization sensor setting for initially smoldering fires and fires with the bedroom door closed, while dual alarm response to the kitchen fires was very sensitive to the ionization sensor setting. Tests conducted in the National Institute of Standards and Technology (NIST) fire emulator/detector evaluator showed that the ionization sensors in off-the-shelf ionization alarms and dual alarms span a range of sensitivity settings. While there appears to be no consensus on sensitivity setting for ionization sensors, it may be desirable to tailor sensor sensitivities in dual alarms for specific applications, such as near kitchens where reducing nuisance alarms may be a goal, or in bedrooms where higher smoke sensitivity may be a goal.     

Rapid Detection and Suppression of Mining Equipment Cab Fires

The National Institute for Occupational Safety and Health (NIOSH/PRL) conducted a series of large-scale experiments to evaluate the effectiveness of optical flame detectors, photoelectric smoke detectors, and combined ionization and photoelectric smoke detectors for rapidly detecting mining equipment cab fires. The detector alarm times were then used to trigger the discharge of a fire inerting system inside the cab to suppress cab material fires. This paper discusses the types of fire detectors tested, the experiments that were conducted, and the results that were obtained. Conclusions are that rapid detection of equipment cab fires can be achieved to trigger the discharge of a fire inerting system inside the cab to protect the operator in the cab.     

Fire Test Comparisons of Smoke Detector Response Times

As part of an effort to develop objective smoke detector selection criteria, smoldering smoke and flammable liquid fire tests were performed on photoelectric, multicriteria (ion/heat/photo), and laser-photoelectric smoke detectors. Results showed that, when evaluated at similar sensitivities, the laser-photoelectric and multicriteria detectors had response times close to those of the photoelectric detector.     

Monitoring Multiple Aspects of Fire Signatures for Discriminating Fire Detection

Numerous researchers are exploring multisensor detection as the principal means of discriminating between fire and nuisance sources. Multisensor detectors can monitor multiple aspects of a wide variety of signatures produced by flaming fires, non-flaming fires, and nuisance sources. This paper describes one program of small- and large-scale experiments that has been conducted using a prototype advanced fire detector with multiple gas sensors. An elementary analysis is applied to demonstrate that spacing guidance can be rationally developed for multiple gas sensors to detect fires of a particular threshold fire size, i.e., heat release rate. Discriminating between flaming fires, non-flaming fires, and nuisance sources could be achieved using either a threshold concentration or CO₂ rate-of-rise to identify flaming fires and a CO or CO₂ rate-of-rise for non-flaming fires. Time to detection was also compared to commercial smoke detectors, and the reductions in time were noted.     

Multisensor/Multicriteria Fire Detection: A New Trend Rapidly Becomes State of the Art

The detection performance of fire alarm systems has recently improved substantially with the development of multisensor/multicriteria detection technology, a new generation of products that derive various alarm and diagnostic criteria from a combination of input signals from sensors responding to different fire phenomena. In an actual case, the signals from a photoelectric smoke sensor and a temperature sensor were combined using modern techniques of signal analysis, such as neural networks and fuzzy logic, which by far exceed commonly used simple logic. The underlying algorithms are parametrized to allow application-specific adaptation of the fire alarm system response behavior.Results from laboratory experiments and computer simulations, field tests, and a rapidly increasing number of real installations clearly demonstrate that systems using multisensor/multicriteria detection technology outperform systems that depend on single sensor inputs, such as ionization or photoelectric smoke detection or temperature sensing. Indeed, the results show that such systems can be adapted to respond to a substantially wider spectrum of fire phenomena, such as visible/invisible and black/white smoke, aerosols, and temperature, while remaining much less sensitive to deceptive phenomena that result from cigarette smoke, welding, spray aerosols, dust, humidity, and so on. This new technology not only contributes to improved life safety, it also reduces the probability of nuisance alarms.In this paper, the principle building blocks that lead to the improved detection behavior will be outlined and results from actual installations will be presented. In particular, the effect of application-specific parametrization will be demonstrated. The performance of multisensor/multicriteria fire detection systems will also be compared to that of systems that depend on single sensor inputs only.     

变电站电缆沟火灾探测试验与数值仿真研究

为探讨火灾探测器灵敏度对电缆沟火灾响应的影响程度,验证火灾监控系统的准确性,应用点式温度探测器、点式烟雾探测器、电缆式感温探测器和吸入式烟雾探测器4种火灾探测器,在标准变电站电缆沟防火分区内进行了全尺寸电缆火灾试验,测试不同灵敏度火灾探测器的火灾报警序列。根据电缆沟的实际尺寸,利用FDS搭建模拟明火和阴燃两种不同火灾场景,对不同灵敏度火灾探测器的温度场、有毒气体质量浓度和火灾报警顺序进行了分析和讨论。试验结果表明,在变电站电缆沟火探测中,电缆式感温探测器、吸入式烟雾探测器、点式感烟探测器火灾平均响应时间分别为41.2,111.4,331.8 s;吸入式感烟探测器的灵敏度对探测结果影响较大,其火灾报警响应时间比电缆型线性温度探测器约延迟50%。     

Analysis of the Response of Smoke Detectors to Smoldering Fires and Nuisance Sources

Since the initial smoke detector was developed, research has been ongoing to reduce the response times to fire sources and improve the ability of detectors to ignore nuisance sources. Research has been conducted to analyze the signatures from cooking activities, ranging from normal cooking to the flaming ignition of food products with the intent of identifying the precursors to flaming ignition. In particular, the goal of the research is to provide an alarm sufficiently prior to flaming ignition to allow homeowners to take corrective actions to prevent a fire. For the 11 experiments analyzed, the optical density measured in the range hood was the most accurate precursor signal, having the greatest ability to predict the imminent transition to a flaming fire with a minimum of false positives. Though plagued by false positives, the temperature of the heating element and response of an ionization detector provided a faster response than the optical density measure. Given that this research only included 11 experiments, further research should be conducted on a broader range of cooking styles and items being cooked.     

Early Warning Fire Detection System Using a Probabilistic Neural Network

The Navy program, Damage Control-Automation for Reduced Manning is focused on enhancing automation of ship functions and damage control systems. A key element to this objective is the improvement of current fire detection systems. As in many applications, it is desired to increase detection sensitivity and,more importantly increase the reliability of the detection system through improved nuisance alarm immunity. Improved reliability is needed, such that fire detection systems can automatically control fire suppression systems. The use of multi-criteria based detection technology continues to offer the most promising means to achieve both improved sensitivity to real fires,and reduced susceptibility to nuisance alarm sources. A multi-criteria early warning fire detection system, has been developed to provide reliable warning of actual fire conditions, in less time, with fewer nuisance alarms,than can be achieved with commercially available smoke detection systems. In this study a four-sensor array and a Probabilistic Neural Network have been used to produce an early warning fire detection system. A prototype early warning fire detector was built and tested in a shipboard environment. The current alarm algorithm resulted in better overall performance than the commercial smoke detectors, by providing both improved nuisance source immunity with generally equivalent or faster response times.     

Results from a Full-Scale Smoke Alarm Sensitivity Study

A series of 24 full-scale experiments was conducted to examine the effects of alarm type (photoelectric, ionization, and dual sensor), alarm location, fabric type (100% cotton and 100% polyester), polyurethane foam density, ignition scenario, and room configuration, on smoke alarm performance. A two-level, fractional factorial design of eight experimental configurations was developed around the five factors: fabric type, foam density, fire location, ventilation, and ignition scenario. A structure, designed to represent a single-story home or apartment, was constructed inside the Large Fire Laboratory at the National Institute for Standards and Technology for the experiments. The fire source was a chair mockup consisting of a seat and back cushion of a specific cover fabric and foam density, weighing between 5.5 kg and 8.3 kg. It rested on a metal frame and was subjected to a small propane gas flame, or an electric cartridge heater to initiate smoldering. Each experimental configuration was replicated three times. Smoldering fires were allowed to progress until they naturally transitioned to flaming fires except for one test that was terminated early due to time constraints. The smoldering to flaming transition times ranged from (81 to 182) min. Each fire progressed for a time sufficient to produce multiple hazards (smoke, heat, and toxic gases). All alarms tested were purchased from retail outlets and activated at their preset levels. Photoelectric, ionization, and dual photoelectric/ionization alarms were co-located at multiple locations to facilitate comparisons of each alarm type, and different designs of the same type of alarm. For smoke alarms in the room of fire origin, it was observed that each of the five factors had an effect on the measured alarm times that was primarily a result of fire growth rate (fabric type, foam density, and ignition scenario), or smoke dilution and transport (fire location and ventilation). The photoelectric alarm responded quicker on average than ionization alarm in two of four smoldering fire configurations, responding before the ionization alarm in all 6 trials, while the ionization alarm responded before the photoelectric alarm in two of three trials for the other two configurations. The ionization alarm responded quicker on average than photoelectric alarm in all four flaming fire configurations, and responded before the photoelectric alarm in all 12 flaming fire trials. One dual alarm had the fastest average alarm time for all four smoldering fire configurations, and responded first in 11 of the 12 trials. It also yielded faster average alarm times than the other dual alarm in seven of eight configurations, and was the first dual alarm to respond in 22 out of 23 trials where dual alarms were present.     

Predicting response times of fixed-temperature, rate-of-rise, and rate-compensated heat detectors by utilizing thermal response time index

A project was initiated to find a way of assigning thermal response time index (RTI) of heat detectors and to see if a detector response time could be estimated by utilizing the assigned RTI value. In phase 1 of the project, the concept was applied to fixed-temperature rating detectors and detector response times were successfully estimated. In the second part of the project, rate-of-rise detectors and rate-compensated detectors were investigated. Test data employing a plunge tunnel of varying air temperatures were used to extract RTI values of a rate-of-rise detector and rate-compensated detectors. In the final stage of the project, adequacy of estimating detector response times of rate-of-rise and rate-compensated detectors based on RTI values in conjunction with other threshold values for activation was assessed. Twenty-seven real-scale fire tests were conducted to see if the estimated response times match well with the measured detector response times. Test results showed that the detector response time estimation based on RTI values matched well with the measured ones. Having the ability of predicting detector response times in association of fire scenarios provides a great flexibility in deploying heat detectors in field operations. Detectors can be installed at performance-based spacing with respect to detector types and anticipated fire growth scenarios. Bench-scale tests to assign RTI values of detectors can replace the real-scale fire tests for the maximum spacing that are costly while less than reliable. Detector response sensitivity can be classified by the value of RTI and the maximum spacing of a detector can be assigned based on its RTI value.     

New methods for reducing the number of false alarms in fire detection systems

In this paper, we present new methods of reducing the number of false alarms in smoke detectors and apply these methods to an ionization smoke detector. The detector is able to diagnose its working condition and its environment very precisely. When the detector's environment changes, the detector can automatically determine the cause, whether the change is fire-related or not. This is done by measuring the ionization current in two sensitivity ranges of the measurement chamber and analyzing the results with new algorithms. With the help of algorithms that use fuzzy logic, we can identify basically every potential problem an ionization detector can produce.     

厨房油烟引起感烟火灾探测器误报实验研究

利用火灾探测综合模拟实验平台研究了厨房油烟引起的感烟火灾探测器误报，对厨房油烟与真实火灾烟气对感烟火灾探测器激励作用进行了比较，测定并分析了厨房油烟与真实火灾烟气中CO与CO2浓度的差异，为研究能抵抗厨房油烟干扰的复合火灾探测器提供了实验依据。     

Video Image Detection and Optical Flame Detection for Industrial Applications

In order to provide a technical basis for the selection and use of flame detectors in challenging industrial applications, an experimental study was conducted. Multiple flame detectors were tested in numerous scenarios representative of those expected in industrial applications. The study systematically evaluated the fire detection performance of devices at different distances from fires and with different detector viewing angles relative to fuel type, a range of obstructions and exposure to nuisance/interference sources. Limited comparisons between the “state of the art” Video Image Detectors and Optical Flame Detector candidate systems to older flame detector technologies were included.     

Performance improvement of fire detectors by means of gas sensors and neural networks

Nowadays, fire detection systems are used world wide in order to protect life and goods. However, at the present time, detectors show poor features with regard to detection speed and reliability. They respond to only a few fire parameters like smoke particles and they do not take into account other important fire parameters such as gaseous products. In this paper, we present a new multi-sensor detector consisting of a commercial optical fire detector, a temperature sensor and selected semiconductor metal oxide gas sensors. The use of a multi-sensor detector requires a more sophisticated algorithm than the simple threshold rule. The new algorithms are typically based on pattern recognition systems, consisting of a pre-processing unit, a feature extraction unit and a classification unit. The choice of suitable methods for the feature extraction and the classification is difficult. Most often, the classifier depends on the type of the extracted features. In this paper two methods for the feature extraction with their suitable classifiers are presented and compared. However the classification is based on neural networks.The first algorithm consists of (i) a pre-processing unit; and (ii) a FFT-based feature extraction unit to resolve characteristic fire signatures. For that purpose a moving window has been introduced and a composed signal has been generated from the different pre-processed sensor responses. The algorithm also consists of (iii) a classification unit with a Learning Vector Quantization (LVQ) neural network to classify the extracted features to fire, not fire, or disturbing event.The second algorithm consists of a pre-processing unit and feature extraction method based on the scaling of the quadratic mean value. For this kind of feature extraction a back-propagation neural network as a classifier has been chosen.An important improvement towards the use of commercial detectors has been achieved using both of the above-described algorithms. The neural networks with suitable feature extraction methods were able to detect test fires more quickly than the commercial optical fire detector without generating false alarms by disturbing events.     

火灾探测器智能探测算法实现与性能评估技术

围绕点型感烟火灾探测算法的设计与性能评估过程，论述了为了获取数据样本而建立的火灾探测应用环境与火灾信息数据库的基本情况，并在数据分析的基础上概述了火灾探测算法的实现方法。为了对算法的性能进行评估和验证，进一步研究了算法的评估模式与技术方法，以及为性能评测试验而建立的火灾探测性能评估模拟试验平台的具体内容，以便提高点型感烟火灾探测器的综合探测能力、减少漏报和误报，进而提高火灾探测器的整体技术水平。