NEARBY AND DISTANT FIRE CONDITION DISCRIMINATION METHOD
FIELD AND BACKGROUND OF THE INVENTION
The present invention relates to a method of detecting a fire condition in a monitored region and, more particularly, to such a method that also discriminates between nearby and distant fire conditions. Many methods are known in the art for detecting a fire condition in a monitored region. One example of such a method is that taught in US Patent No. 5,373,159, which is incorporated herein by reference for all purposes as if fully set forth herein. A limitation of all presently known methods is that they fail to discriminate between small, nearby fires and large, distant fires. This is problematic in facilities, such as gas and oil production facilities and oil refineries, in which it is normal for large exposed controlled flames, such as flares for burning excess natural gas, to be present. Presently known methods of fire detection fail to distinguish between a small, nearby accidental fire, which must be extinguished before it grows, and a distant regulated fire, such as a flare, that can be ignored safely. Presently known methods of fire detection also fail to distinguish between real flames and flames reflected from water or reflective metal structures.
There is thus a widely recognized need for, and it would be highly advantageous to have, a method for detecting a fire condition in a nearby monitored region that distinguishes between flames in the monitored region and distant flames.
SUMMARY OF THE INVENTION
According to the present invention there is provided a method of detecting a fire condition in a monitored region, including the steps of: (a) concurrently monitoring the region by a first sensor sensitive to radiation in a first bandwidth whose overlap with the CO2 emission band includes only a long wavelength portion of the CO2 emission band, by a second sensor sensitive to radiation within a second bandwidth which includes wavelengths mainly lower than the CO2 emission band, and by a third sensor sensitive to radiation within a third bandwidth which includes wavelengths mainly higher than the CO2 emission band, thereby producing first, second and third measurements of radiation variations emitted from the monitored region; and (b) utilizing the measurements in determining the presence or absence of the fire condition in the monitored region.
According to the present invention there is provided a method for discriminating between a nearby fire condition and a distant fire condition in a monitored region, including the steps of: (a) monitoring the region by a first sensor selected from the group consisting of sensors sensitive to radiation in a first bandwidth whose overlap with the CO2 emission band includes only a short wavelength portion of the CO2 emission band and sensors sensitive to radiation in a second bandwidth whose overlap with the CO2 emission band includes only a long wavelength portion of the CO2 emission band that is disjoint from said short wavelength portion of said CO2 emission band, thereby producing a first measurement of radiation variation emitted from the monitored region; and (b) utilizing the first measurement in determining the presence of the fire condition and the distance to the fire condition in the monitored region.
The physical principle whereon the present method is based is illustrated in
Figure 1 , which shows the infrared emission spectrum of a flame as observed from various distances through the atmosphere. The burning of organic materials produces copious amounts of hot CO2. This hot CO2 emits infrared radiation at wavelengths between about 4.1 microns and about 4.7 microns, with a peak at about 4.3 microns. The shape of the corresponding emission spectrum is close to the curve labeled "15 m" in Figure 1. Cold CO2 in the atmosphere absorbs this radiation, but over a narrower spectral range, predominantly at wavelengths shorter than about 4.4 microns. Thus, the observed emission spectrum of a flame depends on the distance between the flame and the observer, as illustrated in Figure 1: the spectrum shifts to longer wavelength with increasing distance. In order to detect a distant flame, without attenuation by cold atmospheric CO2, a sensor responsive to the long wavelengths of the CO2 emission spectrum, i.e., wavelengths longer than about 4.4 microns, is used by the present invention. The preferred spectral band is from about 4.4 microns to about 4.6 microns.
The spectral shift illustrated in Figure 1 is used by the present invention to discriminate between nearby fires and distant fires. In all prior art systems that rely on the detection of CO2 emission to indicate the presence of a fire, no effort is made to distinguish between short wavelength (below about 4.4 microns) CO2 emissions, which are characteristic of nearby flames, and long wavelength (above about 4.4 microns) CO2 emissions, which are characteristic of both nearby and distant flames. Fore example, Bright, in US Patent No. 4,220,857, teaches the detection of CO2 emissions in a 0.3 micron band centered at 4.4 microns. Kern et al., in US Patent Nos. 4,639,598 and 4,785,292, teach the separate detection of wavelengths less than 2
microns and greater than 4 microns. Kern et al. cite Muggli, US Patent No.
4,249,168, who teaches the detection of CO2 emissions in a band between 4.1 microns and 4.8 microns. Goto et al., in US Patent No. 5,153,563, teach the detection of CO2 emissions in a band between 4 microns and 5.5 microns. Barrett, in GB Patent No. 1,550,334, teaches the detection of CO2 emissions in a band between 4.19 microns and 4.45 microns. The above-referenced US 5,373,159 teaches the detection of CO2 emissions in a band between 4.3 microns and 4.6 microns, or alternatively between 4.4 microns and 4.7 microns.
According to a first aspect of the present invention, the detectors described in Figures 7 and 8 of US Patent No. 5,373,159 are improved by using a first sensor IRl that is sensitive only to wavelengths in the long wavelength portion of the CO2 emission band. Preferably, sensor IRl is sensitive only to wavelengths between about 4.4 microns and 4.6 microns. The sensitivity bands of the other two sensors remain as described in US Patent No. 5,373,159: between about 3.8 microns and about 4.2 microns for the second sensor IR2, and between about 4.8 microns and about 5.1 microns for the third sensor IR3. In this way, the detectors are rendered immune to false alarms cause by local infrared sources such as welding operations and hot objects.
According to a second aspect of the present invention, two sensors are used for detecting radiation from the CO2 emission band, in a fire detector otherwise similar to the prior art systems, one sensor that is sensitive predominantly to the short wavelength side of the CO2 emission band, and another sensor that is sensitive predominantly to the long wavelength side of the CO2 emission band. The sensitivity bands of the two sensors are disjoint, i.e., nonoverlapping. Preferably, the cutoff
between short wavelength and long wavelength is at about 4.4 microns. The sensitivity band of the short wavelength sensor may extend to wavelengths shorter than those of the CO2 emission band, and the sensitivity band of the long wavelength sensor may extend to wavelengths longer than those of the CO2 emission band, although one or the other of these sensitivity bands preferably is restricted to the CO2 emission band. The ratio between the signal from the first sensor and the signal from the second sensor is diagnostic of the distance from the fire detector to the fire. Preferably, the cross-correlation of the two signals is used to discriminate against noise. Most preferably, the detector includes a third sensor, responsive to infrared radiation at wavelengths outside the CO2 emission band. The ratio of the first signal to the third signal , and the cross-correlation of the third signal with either or both of the first two signals, are used for further anti-noise discrimination.
BRIEF DESCRIPTION OF THE DRAWINGS The invention is herein described, by way of example only, with reference to the accompanying drawings, wherein:
FIG. 1 is a collection of plots of the infrared emission spectrum of a flame measured at various distances through the atmosphere;
FIG. 2 is a block diagram of an apparatus for implementing the present invention;
FIG. 3 is an illustrative plot of the ratio of measured short wavelength CO2 emission to measured long wavelength CO2 emission as a function of distance from a flame.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
The present invention is of a method of detecting a fire condition which can also determine the distance to the fire condition. Specifically, the present invention can be used to discriminate between nearby fires and distant fires.
The principles and operation of fire detection according to the present invention may be better understood with reference to the drawings and the accompanying description.
Referring now to the drawings, Figure 2 illustrates an apparatus for detecting fires and discriminating between nearby fires and distant fires according to the principles of the present invention. The apparatus includes three sensors IRl, IR2 and IR3, for concurrently monitoring the radiation emitted from the monitored region. The outputs of sensors IRl, IR2 and IR3 are fed to bandpass filters 2, 4, and 6 respectively and thence to amplifiers 12, 14 and 16 respectively, to produce three measurements of radiation, emitted from the monitored region, whose variation in time falls within the bands of the three filters 2, 4 and 6. These measurements, as amplified and outputted from their respective amplifiers 12, 14 and 16, are indicated by the three time- varying signals Vι(t), V2(t) and V3(t).
The three amplifiers 12, 14 and 16 are tuned to amplify the signal from their respective bandpass filters 2, 4 and 6 within a frequency range of 2 - 10 Hz, the flame flicker frequency range, so that their respective output signals represent the measurements of the three sensors within their respective bandwidths at the flame flicker frequency.
The apparatus further includes five correlation circuits 18, 20, 22, 24 and 26, for producing correlation values among the various measurements. Specifically, correlation circuit 18 produces an unnormalized autocorrelation Cn of signal V,(t), correlation circuit 20 produces an unnormalized cross correlation C12 of signal Vι(t) with signal V2(t), correlation circuit 22 produces an unnormalized autocorrelation C22 of signal V2(t), correlation circuit 24 produces an unnormalized cross correlation C23 of signal V2(t) with signal V3(t), and correlation circuit 26 produces an unnormalized autocorrelation C33 of signal V3(t). The inputs to the correlators are the analog outputs of the corresponding sensor or sensors. These inputs are digitized, shifted with respect to themselves (in autocorrelators 18, 22 and 26) or with respect to each other (in cross correlators 20 and 24), and summed within a predetermined time gate, as described in US Patent No. 5,373,159. Note that all five correlations vary in time.
The ratio of autocorrelation Cu to autocorrelation C22 is determined in a ratio circuit 28, which outputs a ratio signal. This ratio signal is inputted to a comparitor 36 where it is compared with a predetermined threshold Tl. Comparitor 36 produces a signal of binary value "1" if the input ratio signal exceeds threshold Tb and otherwise produces a signal of binary value "0". Similarly, the ratio of autocorrelation Cu to autocorrelation C33 is determined in a ratio circuit 34, the ratio signal output of which is inputted to a comparitor 42 where it is compared with a predetermined threshold T4. Comparitor 42 produces a signal of binary value "1" if the input ratio signal exceeds threshold T4, and otherwise produces a signal of binary value "0".
In a normalization circuit 30, cross correlation C12 is normalized by squaring it and dividing the results by the product of autocorrelations Cπ and C22. The output of
circuit 30, a normalized cross correlation of signals V^t) and V2(t), is compared with a predetermined threshold T2 in a comparitor 38: if the normalized cross correlation output by circuit 30 exceeds T2, comparitor 38 produces a binary "1"; otherwise, comparitor 38 produces a binary "0". Similarly, in a normalization circuit 32, cross correlation C23 is normalized by squaring it and dividing the results by the product of autocorrelations C22 and C33. The output of circuit 32, a normalized cross correlation of signals V2(t) and V3(t), is compared with a predetermined threshold T3 in a comparitor 40: if the normalized cross correlation output by circuit 32 exceeds T3, comparitor 40 produces a binary "1"; otherwise, comparitor 40 produces a binary "0". The outputs of comparitors 36, 38, 40 and 42 are combined in an and-circuit
44, which produces a binary "1" if all four inputs are binary "l"s, and otherwise produces a binary "0". The outputs of ratio circuit 28 and of and-circuit 44 is monitored by a CPU 50.
It will be appreciated that the apparatus of Figure 2 is similar to the apparati illustrated in US Patent No. 5,373,159, the principal difference being the wavelength ranges to which the various sensors are sensitive. Specifically, sensor IRl is sensitive to radiation, in the CO2 emission band, of wavelengths shorter than about 4.4 microns; sensor IR2 is sensitive to radiation, in the CO2 emission band, of wavelengths longer than about 4.4 microns, and to radiation at longer wavelengths; and sensor IR3 is sensitive to radiation of the wavelengths longer than the CO2 emission band to which sensor IR2 is sensitive. Thus, and-circuit 44 produces a binary "1" only of all four of the following conditions obtain:
1. The ratio of autocorrelation Cπ to autocorrelation C22 exceeds threshold T,, indicating that signal V,(t), indicative of a nearby fire, is sufficiently
strong, compared to signal V2(t), indicative of a fire at any distance, to indicate the presence of a nearby fire.
2. The normalized cross correlation of signals V,(t) and V2(t) exceeds threshold T2, indicating that the noise levels in signals V,(t) and V2(t) is sufficiently low for the ratio of autocorrelations Cπ and C22 to be trusted as an indicator of a nearby fire.
3. The normalized cross correlation of signals V2(t) and V3(t) exceeds threshold T3, indicating that the noise level in signal V3(t) is sufficiently low for signal V3(t) to be trusted as a valid background signal. 4. The ratio of autocorrelation Cu to autocorrelation C33 exceeds threshold T4, indicating that signal V,(t), indicative of a nearby fire, is sufficiently strong compared to signal V3(t), indicative of background, to indicate the existence of a nearby fire.
The normalized cross correlations are particularly important for noise suppression in the case that sensors IRl, IR2 and IR3 are pyroelectric sensors, which are subject to sporadic noise spikes.
The ratio of autocorrelations Cu and C22 is used by CPU 50 to infer the distance to the fire whose existence is indicated by the binary "1" outputted by and- circuit 44. It will be appreciated from Figure 1 that this ratio decreases monotonically with distance to the fire. It is a straightforward matter to calibrate this ratio experimentally with fires at various distances from the apparatus of Figure 1, or alternatively to predict this ratio as a function of distance by integrating the curves of Figure 1 in the respective bandwidths of sensors IRl and IR2 to give the values of the ratio at the indicated distances and then interpolating the values of the ratio at
distances in between the indicated distances and extrapolating the values of the ratio at distances outside the range of indicated distances. Figure 3 is an illustrative example of a plot of this ratio vs. distance, obtained by integrating under curves similar to those shown in Figure 1 to obtain the points marked by "X"s, and interpolating and extrapolating to other distances.
It will be appreciated that other combinations and cross correlations fall within the scope of the present invention. For example, if the sensitivity band of sensor IRl is chosen to include wavelengths in the CO2 band below 4.4 microns and also wavelengths shorter than the CO2 band, and the sensitivity band of sensor IR3 is chosen to include the same wavelengths shorter than the CO2 band, then the normalized cross correlation between signals V,(t) and V3(t) may be used to indicate that signal V3(t) is a valid background signal.
Conventionally, the correlations between pairs of signals are obtained as described in the above-referenced US Patent No. 5,373,159: digitizing the analog signals and summing the products of the signal values in a time window. There are other ways in which the correlations may be performed. For example, the autocorrelation of a signal may be obtained by selecting the product of the signal values that has the largest absolute value, or alternatively by summing the absolute values of the products of the signal values. The cross correlation of two signals may be obtained by replacing positive products with +1 and negative products with -1 before summing.
While the invention has been described with respect to a limited number of embodiments, it will be appreciated that many variations, modifications and other applications of the invention may be made.