EP2151807A1 - Method of determining the operating period of a danger warning system and danger warning system - Google Patents

Abstract

The method involves determining a filtered signal (2) of measurement- or operation value (1) of a danger warning system i.e. fire alarm, by exponentially filtering first and second orders with filter parameters, and weighted summing of results from filtering the first and second orders. A momentary increase of the filtered signal is determined from the filter parameters and another filter parameter. Information about time period until exceeding or falling below a predetermined reference value through the filtered signal is determined by increasing the filtered signal. An independent claim is also included for a danger warning system including a signal processing device.

Description

The invention relates to a method for operating a hazard detector, in particular a scattered light fire detector with at least one sensor, which detector is provided with at least one signal processing unit, by means of which information about its remaining operating time is obtained from the development over time of measured values of the sensor or operating data of the detector. Moreover, the invention relates to a hazard detector, in particular for carrying out the aforementioned method.

Hazard detectors, including stray light fire detectors, such as those from the DE 103 00 868 A1 are now widely used, for example in the building monitoring, used and are exposed in many applications an adverse environment in the form of dust, which settles, for example, inside a scattered light fire alarm on the walls of the measuring chamber.

This leads there to an increased reflection of the light emitted to the smoke light on just these walls and thus to an increase of the basic signal. Basic signal is the signal of a smoke detector, which is measured in smoke-free air. Over the operating time of the fire detector, the basic signal continuously rises due to the increasing pollution. The so-called alarm threshold, in which the alarm is triggered by the fire detector, must therefore be increased in parallel in the signal processing unit by a value which is generally the same in order to avoid false alarms. For this tracking of the alarm threshold, however, there is a physical upper limit, resulting for example from the linearity of the measurement signal or the resolution of the analog-to-digital converter of the signal processing unit. As soon as the distance between the basic signal and the alarm threshold can no longer be kept constant by tracking, the detector indicates a fault and should be replaced.

Ideally, an annunciator should constantly calculate, or at least at the time of its maintenance, the duration until the disturbance is reached. Based on this information, the operator or service technician can decide whether the detector will survive the time until the next maintenance or test or should be replaced prematurely. Thus, the occurrence of pollution-related disturbances can be reduced or unnecessary replacement can be avoided, resulting in cost savings.

If the actually unlikely case of a constantly increasing pollution were present, then one could use the quotient of the difference between an older value z. B. calculate the starting value of the detector and the current value and the intervening period of operation, the slope of the basic signal. This could in principle still be available standing time until the fault occurs. In practice, however, the environmental conditions will change over time, for example, by moving the detector or changing the use of the room in question (carpet or parquet, warehouse or office, air conditioning, crafts, etc.). However, a prediction of pollution as described above would not take account of a variable increase or decrease in the degree of dusting and would result in values that are too high or too low. This great uncertainty makes such a prognosis pointless.

Regardless of the previous history, knowing the current slope of the fundamental signal, it would be possible to determine that, even with an already high fundamental signal but a very low slope, a long operating time is still to be expected. On the other hand, even a low fundamental signal on a fast rise could lead to a fault before the next maintenance. In order to calculate the current slope according to the methodology described above, at least one data value that is sufficiently far back must be used to compensate for short-term fluctuations. In addition, a large number of the measured values must be stored between the sufficiently long data value and the current measured value for future calculations of the gradient. However, this places great demands on the size of the internal buffer of the Signalverarbeicungseinrichtung and their internal control software, so the firmware, and therefore does not apply in practice.

It is therefore an object of the present invention to provide a method for operating a hazard detector available, which at low equipment cost, a reliable statement about the remaining service life of the Fire detector allows for changing operating conditions and thus allows a cost-reduced operation.

1. a) Ermittlung eines geglätteten Signals G_t der Mess-oder Betriebswerte g_t mittels exponentieller Glättung erster und zweiter Ordnung mit Glättungsparametern α und β und Gewichtung der Ergebnisse g, aus der Glättung erster Ordnung und g_t ^** aus der Glättung zweiter Ordnung;
2. b) Bestimmung der momentanen Steigung Ġ_t des geglätteten Signals G_t aus g_t ^*, g_t ^** und einem Glättungsparameter γ;
3. c) Ermittlung der Information über die Zeitdauer bis zum vermutlichen Über- oder Unterschreiten einer vorbestimmbaren Schwelle durch das geglättete Signal G_t mittels der Steigung Ġ_t und gegebenenfalls deren Ausgabe als restliche Betriebsdauer oder Störung;

This initially contradictory object is achieved by a method which has at least partially the following method steps:

1. a) determination of a smoothed signal G _{t of} the measured or operating values g _t by means of exponential smoothing of first and second order with smoothing parameters α and β and weighting of the results g, of the first-order smoothing and g _t ^** of the second-order smoothing;
2. b) determining the instantaneous slope Ġ _{t of} the smoothed signal G _t from g _t ^* , g _t ^** and a smoothing parameter γ;
3. c) determination of the information about the time until the presumable exceeding or falling below a predeterminable threshold by the smoothed signal G _t by means of the slope Ġ _t and, if appropriate, their output as the remaining operating time or fault;

1. a) Vorgabe wenigstens eines Glättungsparameters α, β und γ ;
2. b) Ermittlung einiger Messwerte g des Sensors, Bestimmung eines Startwertes g ₀ für einen Zeitpunkt t=0 aus der Mittelung dieser Messwerte g zu einem bestimmten Zeitpunkt, Verwendung des Mittelwerts als Funktionswert g ₀ ^* zu diesem Zeitpunkt;
3. c) Ermittlung des Mess- oder Betriebswertes g_t zu einem nach t=0 liegenden Zeitpunkt t;
4. d) Exponentielle Glättung des Mess- oder Betriebswerts g_t zum geglätteten Wert g_t ^* durch Bildung der gewichteten Summe von g_t und dem Ergebnis einer vorangegangenen Glättung g _t-1 ^* sowie dem Glättungsparameter α;
5. e) Exponentielle Glättung von g_t ^* zum geglätteten Wert zweiter Ordnung g_t ^** durch Bildung der gewichteten Summe von g_t ^* und dem Ergebnis einer vorangegangenen Glättung zweiter Ordnung g _t-1 sowie dem Glättungsparameter β;
6. f) Ermittlung eines geglätteten (Grund-)Signals G_t aus einer gewichteten Summe von g_t ^* und g_t ^** ;
7. g) Bestimmung der momentanen Steigung Ġ_t des geglätteten (Grund-) Signals G_t ;
8. h) Ermittlung der Information über die Zeitdauer bis zum vermutlichen Über- oder Unterschreiten einer vorbestimmbaren Schwelle durch das geglättete Signal G_t mittels der Steigung Ġ_t und gegebenenfalls deren Ausgabe als restliche Betriebszeit oder Störung;
9. i) Wiederholung der Verfahrensschritte c) bis h) zu Zeitpunkten t+1, t+2, ..., t+n .

A preferred variant of the method uses, in addition to the abovementioned method steps, additional method steps. This preferred variant has at least partially the following method steps:

1. a) specification of at least one smoothing parameter α, β and γ;
2. b) Determination of some measured values g of the sensor, determination a start value g ₀ for a time t = 0 from the averaging of these measured values g at a specific time, use of the mean value as a function value g ₀ ^* at this time;
3. c) determination of the measured or operating value g _t at a time t after t = 0;
4. d) exponential smoothing of the measured or operating value g _t to the smoothed value g _t ^* by forming the weighted sum of g _t and the result of a previous smoothing g _{t -1} ^* and the smoothing parameter α;
5. e) Exponential smoothing of g _t ^* to the smoothed second order value g _t ^** by forming the weighted sum of g _t ^* and the result of a previous second order smoothing g _{t -1} and the smoothing parameter β;
6. f) determining a smoothed (fundamental) signal G _t from a weighted sum of g _t ^* and g _t ^** ;
7. g) determining the instantaneous slope Ġ _{t of} the smoothed (fundamental) signal G _t ;
8. h) determination of the information about the time duration until the presumable exceeding or falling below a predeterminable threshold by the smoothed signal G _t by means of the slope Ġ _t and, if appropriate, their output as the remaining operating time or fault;
9. i) repetition of the method steps c) to h) at times t + 1, t + 2, ..., t + n.

Among other things, the method is based on the fact that the influence of older data values on the current value is determined by means of at least one smoothing factor. The smaller the smoothing value (s), the stronger the influence of the past data and the stronger the smoothing. The smoothed value follows the original value with a time interval when the trend changes. Nevertheless, in order to be able to recognize trends promptly, a second exponential smoothing (exponential smoothing of the second order) of the simply smoothed values as well as a weighted calculation of both values of the first and second order take place together. The new data value contains significantly fewer fluctuations than the raw data, but follows trend changes faster than the values that are only smoothed once or twice. The smoothing value also determines the speed of the adjustment here. Furthermore, the slope of the smoothed curve can be calculated at the same time. Again, the smoothing factor or factors influence how fast the slope calculation follows the short-term curve.

The method of exponential smoothing of first and second order can thus be used both to smooth the values of a measured signal or other operating data such as the energy consumption of a hazard alarm, ie to remove short-term fluctuations, as well as simultaneously a forecast over the time to reach a certain, the operating period of the detector limiting threshold to create such. B of the degree of pollution or the remaining an energy storage removable energy.

The basic signal to be measured and / or the corresponding operating data g _t are recorded at times t and respectively after the lapse of a corresponding time interval Δ t . To smooth out or dampen rapid fluctuations or signal increases, the smoothing parameters α, β and γ must be selected to be small. In this case, 0 <α <1, 0 <β <1 and 0 <γ <1, where as a rule, though not exclusively, α = β = γ holds.

$${g_t}^{**}=\beta \cdot {g_t}^{*}+\left(1-\beta \right)\cdot {g_{t-1}}^{**}\left(\mathrm{exponentielle\; Gl}\ddot{\mathrm{a}}\mathrm{ttung\; zweiter\; Ordnung}\right)$$

sowie $$G_t=2\cdot {g_t}^{*}-{g_t}^{**}\left(\mathrm{gewichtete\; Summe}\right),$$

wobei t-1 gleichbedeutend mit t-Δt ist.The smoothed basic signal or the smoothed operating data G _{t are} calculated according to $${G_t}^{*}=\alpha \cdot G_t+\left(1-\alpha \right)\cdot {G_{t-1}}^{*}\left(\mathrm{exponential\; Eq}\ddot{\mathrm{a}}\mathrm{first\; order}\right).$$

$${G_t}^{**}=\beta \cdot {G_t}^{*}+\left(1-\beta \right)\cdot {G_{t-1}}^{**}\left(\mathrm{exponential\; Eq}\ddot{\mathrm{a}}\mathrm{second\; order}\right)$$

such as $$G_t=2\cdot {G_t}^{*}-{G_t}^{**}\left(\mathrm{weighted\; sum}\right).$$

where t-1 is t-Δt.

The slope Ġ _{t of} the smoothed curve G _t at time t is calculated according to $${\stackrel{•}{G}}_t=m_t=\frac{\mathit{\gamma }}{1-\mathit{\gamma }}\cdot \left({G_t}^{*}-{G_t}^{**}\right)$$

die folgenden Vereinfachungen bzw. Abschätzungen machen: $$\frac{\mathit{\gamma }}{1-\mathit{\gamma }}=\frac{1/A}{1-1/A}=\frac{\frac{1}{A}}{\frac{A-1}{A}}=\frac{1}{A-1}\approx \frac{1}{A}=\mathit{\gamma }f\ddot{\mathrm{u}}\mathrm{r}\mathit{\gamma }\ll 1\mathrm{bzw}\mathrm{.}\mathit{A}\gg 1,$$

woraus sich die Steigung m_t zu $$m_t=\mathit{\gamma }\cdot \left({g_t}^{*}-{g_t}^{**}\right)$$

ergibt.It can be with you $\mathrm{\gamma }=1/A$

the following simplifications or estimates make: $$\frac{\mathit{\gamma }}{1-\mathit{\gamma }}=\frac{1/A}{1-1/A}=\frac{\frac{1}{A}}{\frac{A-1}{A}}=\frac{1}{A-1}\approx \frac{1}{A}=\mathit{\gamma }f\ddot{\mathrm{u}}\mathrm{r}\mathit{\gamma }"1\mathrm{respectively}\mathrm{,}\mathit{A}»1.$$

hence the slope m to _t $$m_t=\mathit{\gamma }\cdot \left({G_t}^{*}-{G_t}^{**}\right)$$

results.

If a linear curve is assumed in the future, the time t _v at which a defined signal height or a defined operating data value g _{v will be} achieved can be calculated. It must be assumed that the smoothed value G _t to compensate for short-term signal fluctuations, for example by dust particles or consumption peaks or the like, even if G _t may differ due to the strong attenuation of the actual measurement or operating data g _t .

lässt sich der Zeitpunkt t_V zu $$t_V=\frac{g_V-G_t}{m_t}+t$$

bestimmen.With $$m_t=\frac{G_V-G_t}{t_V-t}$$

the time t _V can be $$t_V=\frac{G_V-G_t}{m_t}+t$$

determine.

aus welchem sich die tatsächlich verbleibende Zeit durch Multiplikation mit dem Messintervall Δt(bspw. 60 Sekunden)zu $$T_C=t_c\cdot \mathrm{\Delta }\mathit{t}$$

ergibt.In practice, the determination of a so-called countdown value t _c until the occurrence of the disturbance by z. As pollution or lack of energy $$t_c=t_V-t=\frac{G_V-G_t}{m_t}$$

from which the actual remaining time by multiplying by the sampling interval Δ t (eg. 60 seconds) to $$T_C=t_c\cdot \mathrm{\Delta }\mathit{t}$$

results.

With a suitable choice of the smoothing value, therefore, temporary fluctuations of the fundamental signal or the operating data, e.g. be caused by dust particles during construction or consumption peaks due to frequent testing, are filtered out. The mode of action then corresponds approximately to a moving averaging over a multiplicity of measured values covering a range of the order of magnitude of several days.

The described method thus solves the above-described problem: the basic signal or an operating data sequence is smoothed. At the same time, the current slope of the basic signal or of the operating data sequence can be determined at any time and thus a disturbance prognosis can be created, for which, in a variant of the method, a prognostic value g _{t + r} ^P from the weighted averages of g _t ^* and g _t ^** in the future t + r is determined.

In a likewise expedient variant of the method, after a measurement, only the mathematically determined, simply smoothed value g _t ^* and the twice smoothed value g _t ^{** are} stored in a storage means for further use and are available after the time Δt as values g _{t -1} ^* and g _{t -1} ^** , because for the smoothing and prognosis in contrast to other methods no data except those of the previous calculation must be stored. This saves internal storage space. Since, in terms of the method, only two old data values have to be stored in memory per signal or data sequence, the exponential smoothing is particularly suitable for use in signal processing devices provided with microcontrollers.

In order to avoid the problem of a division by too large integer numbers in the signal processing device, a variant of the method may consist in that an offset g ^{0 is} subtracted from the respective measured value g _t , so that in the case of an analog-digital Conversion no information is lost.

In a further variant of the method, after each measurement, the slope value Ġ _t determined in method step b) or g) is compared with a predefinable value stored in a memory element. By calculating the current slope, it is then possible to check whether the maximum permitted slope of the signal tracking is exceeded by the standards. In this case, measures such as increasing the damping or stopping the tracking can be taken.

The object is also achieved by a fire detector in which the signal processing device is associated with a microcontroller, which subjects successive measured values g _{t of} at least one sensor and / or operating data of the fire detector to multiple exponential smoothing and information about the remaining service life of the fire detector generates and optionally outputs.

Further advantageous embodiments of the fire detector according to the invention will become apparent from the dependent claims.

Fig.1

ein Diagramm mit einer Kurve von Messpunkten sowie einer geglätteten Kurve;

Fig.2

ein Diagramm mit der geglätteten Kurve aus der Fig.1 mit Steigungen zu verschiedenen Zeit- punkten;

Fig.3-5

verschiedene Beispiele von digitalen Filtern zur Glättung der jeweiligen Messwerte zum ge- glätteten Grundsignal und zur Bestimmung eines Prognosewerts.

The invention is explained below with reference to figures in the drawing. It shows the

Fig.1

a diagram with a curve of measuring points and a smoothed curve;

Fig.2

a diagram with the smoothed curve from the Fig.1 with gradients at different times;

Fig.3-5

various examples of digital filters for smoothing the respective measured values for the smoothed basic signal and for determining a prognosis value.

In the Fig.1 a diagram is shown, which shows a possible development of a smoke detector measuring signal, wherein first the measured signal g _t (1) of a fire detector is plotted point by point in arbitrary units over the time t . Due to the large number of measured points after predetermined time intervals, the individual points are only to be resolved here if they stand out markedly from the other curve as outliers. The smoothed basic signal G _t (2) determined from the measuring points is laid as a white curve over the multiplicity of measuring points, whereby it can be seen that by the two-fold exponential smoothing, the smoothed values G, (2) are temporal with respect to the actual values (1) run slightly, otherwise the trend but very well mapped becomes. Next is in the diagram of Fig.1 to recognize a horizontal, dashed line (3) representing the maximum allowable fundamental signal, to the value of which a fixed amount can be added to the continuous pollution increasing, smoothed fundamental signal G _t (2). The time (5) indicated by the vertical dashed line (4), at which the measured and smoothed signals (1, 2) reach this threshold (3) substantially simultaneously, forms the instant (5) at which the detector replaces must be determined by calculating the slope of the smoothed fundamental signal (2).

In the Fig. 2 the measured signal g _t (1) is omitted in the diagram and only the smoothed fundamental signal G, (2) is plotted over time. At the same time, two dash-dotted lines (6, 7) show slopes of this smoothed fundamental signal (2) at different times. It can be seen here that in a time range of approximately 5,000 (arbitrary) units, the gradient (6) of the smoothed signal (2) initially indicates a shorter expected life of the fire detector until its replacement. Due to the later signal development, however, the slope (6, 7) of the smoothed fundamental signal G _t (2) changes and a correct prognosis about the time of the probable replacement of the detector is possible.

In the FIGS. 1 and 2 By way of example, the effect of the above-described smoothing on the measuring signal (1) of a scattered light smoke detector and the application of the inventive method to this signal has been shown. However, it is also conceivable that the method is based on an operating data sequence of parameters such. As the energy consumption or the content of an energy storage, to power the detector is used, which, like the contamination of the measuring chamber of a scattered light fire detector, has an effect on the remaining service life of the detector.

In the Fig. 3 to 5 various digital filter structures assigned to the microcontroller of the fire detector are shown, which can be implemented in software or hardware and which in each case use different measured filter operations on the basis of the measured signal (1) or the operating data g _t (1) Structure of the Fig. 3 the simply smoothed value g _t ^* (8), by the structure of the Fig. 4 the doubly smoothed value g _t ^** (9) and finally the structure of Fig. 5 the smoothed basic signal or the smoothed operating data sequence G _t (2) can be generated at a time t .

In the FIGS. 3 to 5 represent the triangles multiplier (10), the input signal z. G _t (1), g _t ^* (8) or g _{t -1} ^* (8) multiply by the adjacent factor (11). The elements marked "+" are adders (12) which sum the input signals coming from the left, top and bottom and output the sum of these signals as the output signal to the right. The elements marked with "T" represent timers (13) which as a storage means latch the values g _t ^* and g _t ^{** present} at their input and at their output the respective values of the preceding instant t - 1 as g _{t -1} ^* (8) and provide g _{t -1} ^** (9).

Claims (14)

Method for operating a hazard detector with at least one sensor, which detector is provided with at least one signal processing unit, by means of which information about its remaining operating time is obtained from the development over time of measured values g _{t of} the sensor or operating values of the detector, characterized at least by the following method steps : a) Determination of a smoothed signal G _t (2) of the measured or operating values g _t (1) by exponential smoothing of first and second order with smoothing parameters α (11) and β (11) and weighted summation of the results g _t ^* (8) from the first order smoothing and g _t ^** (9) from the second order smoothing; b) determining the instantaneous slope Ġ _{t of} the smoothed signal G _t (2) from g _t ^* (8), g _t ^** (9) and a smoothing parameter γ; c) determination of the information about the time until the presumable exceeding or falling below a predeterminable threshold by the smoothed signal G _t (2) by means of the slope Ġ _t and, if appropriate, their output as the remaining operating time or fault; Method for operating a hazard detector with at least one sensor, which detector is provided with at least one signal processing unit, by means of which the time development of measured values g _t (1) of the Sensor or operating values (1) of the detector information about the remaining operating time is obtained, characterized at least by the following method steps: a) specification of at least one smoothing parameter (11) α, β and / or γ; b) determining a starting value g ₀ for a time t = 0; c) determination of the measurement or operating value g _t (1) at a time t after t = 0; d) exponential smoothing of the measured or operating value g _t (1) to the smoothed value g _t ^* (8) by forming the weighted sum of g _t (1) and the result of a previous smoothing g _{t -1} ^* (8) and the Smoothing parameter (11) α; e) Exponential smoothing of g _t ^* (8) to the smoothed value of second order g _t ^** (9) by forming the weighted sum of g _t ^* (8) and the result of a previous second-order smoothing g _{t -1} ^** ( 9) and the smoothing parameter (11) β; f) determining a smoothed signal G, (2) from a weighted sum of g _t ^* (8) and g _t ^** (9); g) determining the instantaneous slope Ġ _{t of} the smoothed signal G _t (2) from g _t ^* (8), g _t ^** (9) and the smoothing parameter (11) γ; h) determination of the information about the time until the presumable exceeding or falling below a predeterminable threshold by the smoothed signal G _t (2) by means of the slope Ġ _t and, if appropriate their output as remaining operating time or fault; i) repetition of the method steps c) to h) at times t + 1 , t + 2, ......, t + n. A method according to claim 1 or 2, characterized in that the hazard detector is a smoke detector with a scattered light smoke sensor, and that are used as measured values g _t (1) the measured values of the smoke sensor. Method according to Claim 1 or 2, characterized in that voltage values of an energy store and / or energy consumption values of the detector and / or the remaining energy content of a store are used as operating values g _t (1). Method according to one or more of claims 1 to 4, characterized in that from the weighted averages g _t ^* (8) and g _t ^** (9) a forecast value g _{t + r} ^P for a future time t + r is determined. Method according to one of claims 1 to 5, characterized in that the information about the remaining service life of the fire alarm is expressed by a time or a period of time. Method according to one of Claims 3 to 6, characterized in that the smoothed signal G _t (2) determined at the respective instant t is interpreted as a basic signal and an alarm threshold of the fire detector is determined by adding a fixed value. Method according to one of the preceding claims, characterized in that, after each measurement, the slope value Ġ _t determined in method step b) or g) is compared with a predefinable value stored in a memory element. Method according to one of the preceding claims, characterized in that after a measurement only the mathematically determined, singly and doubly smoothed values g _t ^* (8) and g ^** (9) are stored in a storage means for further use. Hazard detector, in particular fire detector for carrying out the method according to one of the preceding claims, which is in turn assigned to a microcontroller, characterized in that the microcontroller a present measurement (1) at least one sensor or operating data (1) of the fire detector a multiple exponential smoothing and information about the duration of the fire detector from the smoothed signal G _t (2) generates and optionally outputs, wherein the smoothed signal G _t (2) from a weighted sum of a single smoothed signal g, (8) and a twice smoothed signal g _t ^** (9) is obtained. Hazard detector, in particular fire detector according to claim 10, characterized in that the microcontroller determines the instantaneous slope of the smoothed signal (2). Hazard detector, in particular fire detector according to claim 10 or 11, characterized in that the microcontroller generates an alarm threshold of the fire detector from the smoothed signal (2) and a value that can be influenced or possibly influenced by measured values of further sensors. Hazard detector, in particular fire detector according to one of claims 10 to 12, characterized in that the microcontroller causes an adaptation of calculation parameters or a fault message when exceeding allowable slope values. Hazard detector, in particular fire detector according to one of the preceding claims, characterized in that the microcontroller is provided with at least one digital filter (14) which performs the exponential smoothing of the measured value (1).

Images