JP2755973B2 - Fire alarm - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention detects a plurality of physical quantities of heat, smoke, gas, or the like based on a fire phenomenon in a time series, and based on these plural physical quantities in a time series. The present invention relates to a fire alarm device for making a fire judgment by using a fire alarm.

[Prior art] Detection information based on the physical quantity of a fire phenomenon, that is, a sensor
When detecting a plurality of levels in chronological order and making a fire judgment based on a plurality of sensor levels that change over time, create a table of patterns based on the plurality of sensor levels and fire information for each pattern. In other words, a so-called pattern discrimination has been considered in which a fire detection is performed by storing the detected sensor level in a ROM or the like and comparing the actually detected sensor level with the pattern information in the table.

It has also been considered to define a function using a plurality of sensor level values as variables, and to make a fire judgment based on the relationship between the input and output of the function (for example, the applicant of the present invention has proposed a method in 1988). Japanese Patent Application No. 63-7628 filed on March 31
No. 2).

[Problems to be Solved by the Invention] In any of the above cases, only the determination as to whether or not there is a fire is made based on the detected sensor level. In other words, fire accuracy and danger, as well as fires from smoldering fires to flaming fires can be monitored in detail as a whole, and more accurate fire monitoring that takes into account the possibility of false alarms caused by noise etc. It is very preferable to perform

Therefore, a first object of the present invention is to provide a fire alarm device that makes a fire judgment from a plurality of sensor levels detected in time series, not only to judge whether or not a fire has occurred, but also to determine whether a fire has occurred. In addition to the above, it is possible to monitor the accuracy and danger of fire, as well as the entire process from smoldering fires to flaming fires,
It is an object of the present invention to provide a fire alarm device that also eliminates the possibility of false alarm or the like due to the influence of noise or the like.

Such a purpose is achieved by using a plurality of time-series sensors as described above.
When defining the level vs. fire information in a table stored in ROM or the like, increasing the number of input points explosively increases the number of input combinations. A large ROM table is required. In addition, when describing the relationship between input and output using the above functions, there is a limit in expressing a complex relationship. Furthermore, it is practically impossible to eliminate the possibility of false alarms or the like due to the influence of noise using a method using a table or a function.

Accordingly, a second object of the present invention is to provide a fire alarm apparatus having a signal processing structure suitable for achieving the first object.

[Means for Solving the Problems] To achieve these objects, according to the present invention, fire information is obtained by performing signal processing on detection information output from fire phenomenon detection means, and various types of fire information are obtained based on the fire information. In a fire alarm device that performs a fire judgment, a set of a plurality of specific patterns of the plurality of pieces of detection information to be obtained in chronological order and fire information to be obtained when the specific pattern is given is stored. Table, and when a plurality of pieces of detection information actually obtained in time series are input, according to the degree of contribution to the fire information, the plurality of pieces of detection information actually obtained in time series are A signal processing network configured to calculate the fire information based on the weighted information, and a specific pattern of a plurality of pieces of detection information in the table to the signal. Given to the processing network Adjusting means for adjusting the weight so that the fire information calculated at the time is approximated to the fire information in the table.

In a specific embodiment, it is preferable that a storage area for storing each of the adjusted weights is provided. In this case, the signal processing network performs a process for each detection information obtained in time series. , Weighting is performed using the value read from the storage area, and the calculation is performed.

Further, the signal processing network does not directly calculate fire information from a plurality of pieces of input detection information, but once calculates intermediate information from the input information, and calculates fire information from the intermediate information. It is preferable that the calculation is performed hierarchically. The hierarchy can have a plurality of stages, and the number of intermediate information to be calculated in each intermediate hierarchy is arbitrarily set. For example, if the hierarchy has two stages, input-intermediate and intermediate-output, first, each of the input information is individually weighted first, and each intermediate information is calculated. , And output information, that is, fire information is calculated. The value of each intermediate information is not important, and the signal processing network firstly adjusts the first and second weighting values by the adjusting means so that the relationship between the input information and the output information approximates the contents of the definition table. Adjusted for.

According to another aspect of the present invention, the weight value is not adjusted after installation at the site, but a storage area in which the adjusted weight value is stored is used. The weight value in such a storage area is adjusted in advance at the manufacturing stage or the like using, for example, the table or the adjusting unit.

Specifically, as a configuration of such an aspect, a signal processing network configured to calculate fire information by assigning corresponding weights to each of a plurality of pieces of detection information obtained in time series by the fire phenomenon detection means. And the fire information calculated when a specific pattern of a plurality of pieces of detection information to be obtained in time series is given to the signal processing network, to the fire information to be obtained when the specific pattern is given. Storage means for storing weight values adjusted in advance so as to approximate each other, wherein the signal processing network uses the weight values stored in the storage means to correspond to each of the inputted detection information. The fire alarm device is characterized in that weighting of the fire alarm is performed.

[Operation] First, the adjusting unit adjusts the weighting value so as to minimize the error with respect to the input / output values indicated in the definition table, and thereby teaches the contents of the definition table to the signal processing network. I do. Once a signal processing network is formed in this way, the signal processing network can output desired output values for all input values,
It can also handle combinations of multiple patterns of time-series detection information that are not defined in the definition table, and indicate the value of desired fire information (fire accuracy, danger, smoke fire accuracy, etc.) . Thereby, detailed fire judgment can be performed based on the time-series detection information.

In this way, when defining the input / output relationship, it is not necessary to define all combinations of patterns, but it is sufficient to define each important point. Also, especially when it is necessary to describe in detail the singular point where the output value changes significantly due to a slight deviation of the input value, or the minimum point, the vicinity of the maximum point, define the surroundings in detail, and However, it can be roughly defined.

Also, if you want to change the relationship between input and output, there are cases where a different output value is defined for the input value defined so far, and a case where the definition is made for an undefined area so far. The definition can be easily changed by running the adjusting means (net structure creation program). That is, by changing the definition, accurate fire judgment, danger judgment, and the like can be performed.

EXAMPLES Examples of the present invention will be described below.

FIG. 1 shows a sensor level of an analog physical quantity based on a fire phenomenon detected by each fire detector, which is sent to a receiving means such as a receiver or a repeater, and the receiving means based on the collected sensor level. FIG. 2 is a block circuit diagram in a case where the present invention is applied to a so-called analog fire alarm device that makes a fire judgment. Of course, the present invention is also applicable to an on / off type fire alarm device in which each fire detector makes a fire judgment and sends only the result to the receiving means.

In Figure 1, RE is the fire receiver, DE ₁ ～DE _N, for example fire of N analog which is connected to the fire receiver RE through a transmission line L such as a pair of power supply and signal lines And the internal circuit of only one of them is shown in detail.

In the fire receiver RE, the MPU 1 is a microprocessor, the ROM 11 is a program storage area storing programs related to the operation of the present invention described later, and the ROM 12 is a table for various constants such as fire discrimination criteria for all fire detectors. Various constant table storage areas for storage, ROM 13 is a terminal address table storage area that stores the address of each fire sensor, RAM 11 is a work area, and RAM 12 is a definition table described later for all fire sensors. RAM13 is a storage area for weight values for storing signal line weight values for all fire detectors, and TRX1 is a serial / parallel converter or parallel / serial converter. DP is a display such as a CRT, KY is a numeric keypad for learning data input described later, IF11, IF12 and IF13 are interface Source.

In the fire detector DE ₁ , MPU2 is a microprocessor, ROM21 is a program storage area, ROM22 is a self-address storage area, RAM21 is a work area, and FS is heat, smoke, Alternatively, it is a fire phenomenon detecting means for detecting a physical quantity of gas or the like. In this embodiment, a scattered light type smoke sensor is used. The smoke sensor unit FS is
Although not shown, it has an amplifier, a sample and hold circuit, an analog / digital converter, and the like.

 TRX2 is a signal transmitting and receiving unit similar to TRX1, and IF21 and IF22 are interfaces.

Subsequently, the operation according to the embodiment of the present invention will be specifically described. Prior to that, the operation will be described first.

The present invention is intended to quickly and correctly perform various fire determinations such as fire accuracy and danger based on a plurality of time-series sensor levels from a sensor unit that detects a physical quantity of a fire phenomenon. As an example, sensor levels from the sensor unit sampled every 5 seconds are collected over 25 seconds, and a total of 6 sensor levels are input to the net structure as a pattern, and the fire accuracy is output as an output. The effects are listed first, and the action is first described in FIGS.
This will be described with reference to the drawings.

FIG. 2 shows a definition table that defines true or fairly accurate fire probabilities for 26 combinations or patterns of six sensor levels, and for each pattern number up to 26. , INPU above
T represents six chronological sensor levels. The six sensor levels correspond to the leftmost one sampled 25 seconds ago and represent new sampled data from left to right,
And the one on the far right is the most recently sampled sensor level. OU in the middle column for each pattern number
TPUT (T) has 6 sensors at INPUT in the upper column.
Fire accuracy according to the level is indicated by a value of 0 to 1.
The sensor level in the upper column is also converted to a value of 0 to 1,
In this case, as an example, 0 to 1 in the smoke sensor unit corresponds to a smoke concentration of 0 to 20% / m detected by the smoke sensor unit.
OUTPUT (R) in the lower column is an actually measured value of the fire accuracy actually obtained, which will be described later.

In the table of FIG. 2, the fire accuracy OU to be obtained when one pattern of six sensor levels is given
In this embodiment, TPUT (T) is derived based on the following concept.

That is, the sensor level converted to 0-1 is 0.3
If it is over and keeps a certain value or is on the rise, add 0.2 per section as fire accuracy,
If the sensor level is above 0.3 but is currently decreasing, add 0.1 for each section as fire accuracy, and 0 for other sections, and for all 5 sections with 6 sensor levels The sum of these fire accuracies is defined as the overall fire accuracy.

By expressing the above with an equation, a certain sensor level can be expressed as SL
Vn, the sensor level sampled after the next 5 seconds is SLVn + 1, and the fire probability in each section is Sm (1 ≦
When m ≦ 5), the value of Sm is SLVn ≧ 0.3 and SLVn ≦ SLVn + 1 according to the values of SLVn and SLVn + 1, Sm = 0.2 SLVn ≧ 0.3 and SLVn−1> SLVn, and Sm = 0.1 SLVn <0.3 and SLVn ≦ SLVn + 1, Sm = 0 SLVn <0.3 and SLVn> SLVn + 1 can be expressed as Sm = 0, and therefore, the fire probability S over the entire 5 sections is Becomes

The overall fire accuracy S obtained as described above forms the basis for deriving the value of OUTPUT (T) in the middle column of the definition table in FIG. The value as it is is not used as the value of OUTPUT (T), but is more accurate and closer to actuality in consideration of the influence of noise at each installation location and the like and the credibility of stochastic data. Also, pattern numbers 20 to 26
As can be seen in (1), the case where the sensor level does not change linearly is defined in the same way to provide redundancy, and can flexibly respond to the actual time-series sensor level pattern. It has become. For example, in the case of the pattern number 5, the OUTPUT (T) is 0.800, but based on the above concept, the OUTPUT (T) should be 0.7. This is because the relationship before and after the sensor level SLV ₅ has risen, and only the sensor level SLV ₅ has dropped extremely, so that SLV ₅ = 0.380 is considered to be the effect of noise, and SLV ₅ is actually the SLV _{5 4} <SLV ₅ <SLV ₆
It has become.

Such a definition table can be accurately created based on the above-described concept, by conducting an experiment for each characteristic or installation location of the fire detector, or by taking into account the credibility of data establishment. . But not just for 26 patterns of 6 sensor levels,
It is virtually impossible to create such a table for every pattern. According to the operation of the present invention described below, it is possible to obtain accurate fire accuracy including the filtering effect of noise and the like for all patterns based on the six sensor levels in time series. .

Now, assume a net structure as shown in FIG. 3 to explain the operation according to the present invention. The purpose of this network structure is to be obtained an accurate fire accuracy giving six sensor levels, assuming corresponding to each fire detector DE ₁ ~DE _N present in the fire receiver in RE Is what is done.
In the net structure shown in FIG. 3, IN _{1 to} IN ₆ on the left are input layers.
If we call IN and OT ₁ on the right the output layer OT,
In the present embodiment, the input layers IN _{1 to} IN ₆ are provided with six sensor levels converted to 0 to ₁ , respectively.
From the OT _1, the fire accuracy represented by 0 to ₁ in this embodiment is output. With four as an example, but the IM ₁ to IM _4, which is shown to be referred to as an intermediate layer, the intermediate layers IM ₁ to IM ₄ receives a signal from the respective input layers IN ₁ to IN _6, the output layer OT ₁ To output a signal. It is assumed that the signal travels from the input layer toward the output layer, that there is no signal coupling in the reverse direction or between the same layers, and that there is no direct signal coupling from the input layer to the output layer.
Therefore, as shown in FIG. 3, there are 24 signal lines from the input layer to the intermediate layer, and there are 4 signal lines from the intermediate layer to the output layer.

These signal lines shown in FIG. 3 have their weight values or coupling degrees changed according to the value to be output from the output layer in accordance with the signal input from each input layer. The signal on the signal line is better. The weights of a total of 28 signal lines, 24 between the input layer and the intermediate layer and 4 between the intermediate layer and the output layer, are first adjusted according to the relationship between the input and output, and The weighting values shown in the figure are stored in the respective fire detector areas in the storage area RAM13. The content of the weight value stored in this manner is used for the subsequent fire monitoring operation.

Specifically, the six values of the upper column INPUT in each pattern number of the definition table of FIG. 2 are respectively given to input values IN _{1 to} IN ₆ by a net structure creation program described later, and output based on those inputs. the value output from the layer OT _1, compared with the values of the fire probability as a teacher signal or learning data shown in oUTPUT (T) of the middle column of FIG. 2, the signal lines so that they error is minimized Change the weight value of. In this way, it is possible to instruct the net structure of FIG. 3 a very close approximation of the entire function of the definition table of FIG.

Now, the weight value between the input layer INi and the intermediate layer IMj is And the weight between the intermediate layer IMj and the output layer OTk is (I = 1 to I, j = 1 to J, k = 1 to
K, however, in this embodiment, I = 6, J = 4, K =
1), weight value Has positive, zero, and negative values, respectively. If the input values in the input layer INi are represented by INi, the total sum NET ₁ (j) of the inputs to the hidden layer IMj is This value NET ₁ (j) is converted to a value of 0 to ₁ by, for example, a sigmoid function, and is converted to IMj
If it is expressed by Becomes Similarly, the sum NET of inputs to the output layer OTk
₄ (k) When this value NET ₂ (k) is converted to a value of 0 to 1 by the sigmoid function and expressed as OTk, Becomes As described above, the relationship between the input values IN _{1 to} IN _I and the output value OT ₁ in the net structure shown in FIG. 3 is expressed by Equations 1 to 4 using weighting values. Here, gamma ₁ and gamma ₂ are adjustment factor sigmoidal curve, in this embodiment is selected properly in _{γ 1 = 1.0, γ = 1.2} . The slope of the sigmoid curve can be adjusted by these adjustment coefficients, thereby making it possible to adjust the convergence speed when the error is reduced.

In the net structure creation program, first, one of the six sensor level pattern combinations showing the 26 definition tables shown in FIG. 2 stored in the storage area RAM 12 is input to the input layer shown in FIG. When given to IN _{1 to} IN ₆ , the value OTk (k = 1 in this embodiment) calculated from the above-described equations 1 to 4 and output from the output layer is shown in the middle column of FIG. It is compared with the teacher signal output T ₁ of the as fire probability represented represent error Em in the output layer at that time (m = 1 to m, in the present embodiment = 26) with the following equation.

Here, OT ₁ is a value given by Equation 4 above. Error E
The value E obtained by summing m for M combinations of patterns, that is, for all 26 combinations of patterns in the table of FIG. Becomes

Finally, an operation is performed to adjust the weights of the signal lines one by one so that the value E in Equation 6 is minimized. The weight value stored in each fire detector area in the storage area RAM 13 is updated with these adjusted new weight values, and is used in a normal fire monitoring operation. Such adjustment of the weight of the signal line is performed for all the fire sensors in the fire alarm device.

When the education of the table shown in FIG. 2 for the net structure conceptually shown in FIG. 3 is completed, that is, when the adjustment of the weight value of each one is completed, at the time of actual fire monitoring, a net structure calculation program described later is used. , time series six sensor levels sampled over 25 seconds is given to the input layer of the network structure, determined by calculation a value obtained from the output layer OT ₁ with equations 1 4 above Then, a fire judgment is made by comparing the calculated values with a reference value of fire accuracy.

In the above description, the case where the number of pieces of information input from the input layer is six and the number of pieces of information output from the output layer is one is shown. Needless to say, it can be arbitrarily selected. The information output from the output layer includes various information such as the degree of danger, smoke density, and line-of-sight distance in addition to the fire accuracy.

Also, the case where the number of layers in the intermediate layer is one and there are four elements in one layer is shown, but the relationship between the number of elements in one intermediate layer and the number of input information and the number of output information is When the number of input information increases, the error can be further reduced by increasing the number of elements in the intermediate layer accordingly. Further, if the number of intermediate layers is increased, the accuracy is further improved.

Further, in the above description, the total sum NET ₁ (j) of the inputs to each element of the intermediate layer calculated by (Equation 1) is converted to a value of 0 to ₁ by a sigmoid function by (Equation 2), and is converted to (Equation 3) NET ₁ (j) in this way 0-1
May be used directly instead of IMj of (Equation 3) without converting to the value of IMj. Even in that case, the final output information is converted to 0 to ₁ by (Equation 4) and output from the output layer OT1.

In the above embodiment, there is no coupling between the elements of the intermediate layer and no coupling between the elements of the input layer and the output layer. Even in such a case, the weighting value is changed so as to reduce the error in principle. Thereby, the object of the present application can be achieved.

4 to 7 are flowcharts for explaining the operation of the present invention by the program stored in the storage area ROM 11 of FIG.

In FIG. 4, first, for each of the N fire sensors shown in FIG. 1, the net structure creation program is executed in order from the first fire sensor.

The operation of the net structure creation program in the nth fire detector (n = 1 to N) will be described. First, the six sensor levels in the upper column of the definition table described in FIG. The input is given as a teacher input or learning input from the learning data input numeric keypad KY (step 404). The definition table is prepared for each fire detector because the installation environment and individual characteristics of the fire detector itself are different for each fire detector, but if the environmental conditions and characteristic conditions are the same, Can, of course, use the same definition table for the same condition.

The content of the definition table for nth fire detector is numeric keypad KY
Is stored in the n-th fire detector area in the definition table storage area RAM 12 (Y in step 403), the net structure creation program shown in FIG.
Move on to run 600.

First, a total of 28 signals, 24 signals between the input layer and the intermediate layer and 4 signals between the intermediate layer and the output layer described in FIG. 3, which are stored in the n-th fire detector area of the storage area RAM13. Line weight Is set to a certain value (step 601). next,
Equations (1) to (6) based on the fixed weight value
, The actual output values for all M combinations (M = 26 in this embodiment) of the definition table in FIG.
OT and (the sum of the squares of the errors between the teacher output values T seek (E equation 6) it with E ₀ (step 602).

Next, the weights of the four signal lines between the intermediate layer and the output layer are first reduced by one so that the total value E ₀ of the errors becomes minimum when the same definition table input is given. The operation of adjusting one is performed (N in step 603). Since adjustment of the weighting value only between the intermediate layer and the output layer, the above-described equations 1 and 2 are used.
There is no change in the values up to. First, the weight of the first signal line (Step 604), and the same calculation of Expressions 3 to 6 is performed, and the final total value E of errors obtained by Expression 6 is set to E _S (Step 605). Then the E _S, compared with the total value E ₀ of the previous error changing the heavy bid (Step 60
6).

If E _S ≦ E ₀ (N in step 606), the E _S is set as a new E ₀ (step 609), and the changed weight value is set. Is stored in an appropriate position in the work area.

If E _S > E ₀ (Y in step 606), the original weight value is changed because the direction of changing the weight value is wrong. Change the weight value to the opposite side based on Using the values to calculate the E _S based on Equation 3 Equation 6 in the same manner as described above (step 607, 608), sets the value of the calculated E _S as a new E ₀ (step 609) , Changed weight value Is stored in an appropriate position in the work area.

Here, β is a coefficient proportional to | E _S −E ₀ |
S is variable depending on the number of times the weighting value is changed, and when the number of changes increases, S becomes a small value.

In steps 604-609, When the change adjustment for is completed, the weights of the remaining three signal lines are Are sequentially adjusted in steps 604 to 609 in the same manner.

In this way, the weight value of all signal lines between the intermediate layer and the output layer Is adjusted (Y in step 603), next, the weight of the signal line between the input layer and the intermediate layer Are also adjusted in steps 610 to 616 so that errors are similarly reduced based on all of the equations (1) to (6).

When the weights of all the signal lines have been adjusted (Y in step 610), E ₀ thus reduced is reduced.
Is compared to a predetermined value C, and if still greater than C (N in step 617), return to step 603 to further reduce the error, and remove the intermediate layer in steps 604-609.
The above-described process from adjusting the weight value between the output layers is repeated again. When E ₀ is equal to or less than the predetermined value C (Y in step 617), the process goes to step 406 in FIG. 4 and the weights of the 28 signal lines that have been changed and adjusted are adjusted. Are stored at the corresponding addresses of the n-th fire sensor area in the storage area RAM 13.

In the above operation, the values of S, α, β, C, etc. are stored in the storage area ROM 12 of the various constant tables.

Since the final error of E ₀ does not become 0, the adjustment of the weight of the signal line is terminated at an appropriate place, but when the value becomes equal to or less than the predetermined value C as shown in step 617. In addition to terminating the adjustment, the number of adjustments of the weighting value may be determined in advance and automatically terminated when the number of adjustments is reached.

The value of OUTPUT (R) in the lower column of each pattern number in FIG. Repeatedly to create a net structure, and with respect to the net structure thus created, the upper column IN of FIG.
Input ₆ sensor levels SLV _{1 to} SLV 6 shown on PUT
Indicates the fire accuracy output as OT when given as N. The fire accuracy OUTPUT (R) actually output from the net structure is the OU that was initially set as the teacher signal.
It can be seen from FIG. 2 that it is very close to TPUT (T). FIG. 8 shows the weights when the actual value OUTPUT (R) of the fire accuracy is obtained.

FIG. 9 shows the accuracy of the fire accuracy output from the net structure when not only the specific pattern of the six sensor levels but also the actual arbitrary value of the sensor level that changes every moment is input to the net structure. The abscissa indicates time Time, and the ordinate indicates the sensor level SLV, which changes every moment, and the fire accuracy F output from the net structure.

In this way, by defining the time-series input information of the six sensor levels and the fire accuracy as the teacher signal as 26 patterns, even if there is no combination of the input information in the definition table, the network between them is defined. Fill in the structure and output the optimal output as the answer. In the present embodiment, the case where the number of inputs to the net structure is six and the number of outputs is one is shown. However, it is possible to arbitrarily increase or decrease the number of inputs or the number of outputs. Those skilled in the art will readily understand. Outputs include fire accuracy, probability of non-fire,
Various combinations such as line-of-sight distance, walking speed, and probability of fire extinguishing are possible.

If the weighting of the signal line is adjusted for all N fire sensors in the fire alarm device (Y in step 407), and it is determined that there is no need for relearning ( In step 408, N), the fire monitoring operation is performed sequentially from the first fire detector.

The fire monitoring operation for the n-th fire detector DEn will be described. First, a data return command is sent from the signal transmitting / receiving unit TRX1 to the n-th fire detector DEn via the interface IF11 on the signal line L (step 411). .

When the nth fire detector DEn receives the data return command,
The fire detector DEn is detected by the sensor unit, that is, the fire phenomenon detection means FS according to the program stored in the program storage area ROM21, and is converted into a digital value by the built-in analog / digital converter (smoke related to the fire phenomenon,
The sensor level (based on a physical quantity such as heat or gas) is read via the interface IF21, and is returned from the signal transceiver TRX2 via the interface IF22.

If there is a return from the sensor unit of the n-th fire detector DEn (Y in step 412), the returned sensor level is stored in the work area RAM 11 (step 413).

The work area RAM 11 is assigned an area for storing a plurality of sensor levels for each fire detector,
The sensor level returned from each fire sensor at each polling is stored for a predetermined time, and the oldest data, that is, the sensor level is discarded. For example, stored in 1 polling cycle is 5 seconds for the fire detector DE ₁ ~DE _N of the fire receiver RE, if 25 seconds for a predetermined time, for each fire detector 6 times the polling of sensor level at all times Will be done.

When the sensor level returned from the n-th fire detector DEn is stored in the n-th fire detector area of the work area RAM 11 and the oldest data is discarded (step 413), the n-th fire detector is next detected. The six sensor levels stored in the dexterity area have values INi (i =
(1) to (6) and entered into the net structure calculation program (step 414), whereby the net structure calculation program 700 also shown in FIG. 7 is executed.

In the net structure calculation program 700, NET ₁ (j) is calculated according to the above equation 1 (step 703), and is converted into the value of IMj according to equation 2 (step 704).
When the value of all the IMj to _{IM 1 ~JM J (J = 4} ) is determined (Y of step 705), then calculate the NET ₂ (k) according to equation 3 above using the values of those IMj (Step 70
8), convert it to OTk value according to equation 4 (step
709). OTk (k = 1 in this embodiment), that is, the fire accuracy OT
_When the value of ₁ is determined (Y in step 710), the process returns to the flowchart of FIG.

Therefore, in FIG. 5, first, the value of OT ₁ is displayed as it is as the fire accuracy (step 415), and the value of OT ₁ is the reference value A of the fire accuracy read from the various constant table storage area ROM12. Are compared (step 416), OT ₁ ≧ A
If so, a fire display is performed (step 417).

Thus, the fire monitoring operation for the nth fire sensor is completed, and the same fire monitoring operation for the next fire sensor is performed.

In the above embodiment, the storage area RAM1 of the definition table is used.
Data is input to the storage area RAM 13 based on the data, and the weight value is stored in the storage area RAM 13 based on the data.
In the production stage at a factory or the like, a weighting value is obtained using a net structure creation program and stored in a ROM such as an EPROM, and this ROM is used. Can also.

Further, instead of the analog fire alarm device of the above embodiment, the present invention makes a fire judgment on each fire detector side, and sends only the result to a receiving means such as a fire receiver or a repeater. It can also be applied to an off-type fire alarm device, but in that case, the RO
M11, ROM12 and RAM14 were relocated to each fire detector side, and instead of RAM12 and RAM13, weighting values were stored in the above-mentioned calculation stage at the factory etc.
Advantageously, a ROM is provided for each fire detector. This is because there is no space in the fire detector for providing a numeric keypad or the like as shown in FIG. In this case, the fourth
Steps 401 to 408 in the figure are performed by a signal processing device provided in a factory or the like, and the weight value is stored in an EPROM in step 406 and mounted on a fire detector. Then, in the fire detector, steps 409 in FIG. 4 to step 418 in FIG. 5 are performed.

[Effects of the Invention] As described above, according to the present invention, a net structure, that is, a signal processing network is formed by weighting the combination of input / output values shown in the definition table so as to reduce the error, and At the time of monitoring, 6
Because it is configured to provide a combination of two time-series sensor levels, accurate and detailed fire information corresponding to a given combination of input information can be obtained, and therefore, a highly accurate fire judgment can be made. There is an effect that can be performed.

[Brief description of the drawings]

FIG. 1 is a block circuit diagram showing a fire alarm device according to an embodiment of the present invention, and FIG. 2 is a definition table of definition input INPUT versus definition fire information OUTPUT (T) used in the embodiment of the present invention, and a definition. FIG. 3 is a diagram showing actually measured fire information values OUTPUT (R) actually output from the net structure when an input INPUT is given. FIG. 3 is a diagram for conceptually explaining a signal processing network used in an embodiment of the present invention. Figures 4, 5 and 5
FIG. 6 is a flowchart for explaining the operation of FIG. 1, FIG. 6 is a flowchart for explaining the net structure creating program (weighting value adjusting means) shown in FIG. 4, and FIG. 8 is a flowchart for explaining the net structure calculation program shown in FIG. 8, FIG. 8 is a diagram showing each weight value when the measured fire information value of FIG. 2 is obtained, and FIG. 9 is an actual sensor level. FIG. 8 is a diagram showing fire probabilities output from the net structure with respect to transitions of FIG. In the figure, RE is a fire receiver, ROM 11 is a program storage area, RAM 11 is a work area, RAM 12 is a definition table storage area, RAM 13 is a weight value storage area, KY is a learning data input numeric keypad, and DE ₁ to DE _N is a fire detector, FS is a sensor (fire phenomenon detection means), Heavy bid, IN ₁ to IN ₆ input information (chronological six sensor levels), OT ₁ is fire information (fire probability) is.

Claims (2)

(57) [Claims] 1. A fire alarm device which performs signal processing on detection information outputted from a fire phenomenon detection means to obtain fire information, and makes various fire judgments based on the fire information. A table storing a set of a plurality of specific patterns of detection information to be obtained and fire information to be obtained when the specific pattern is given, and a plurality of pieces of detection information actually obtained in time series When input, according to the degree of contributing to the fire information, a corresponding weight is given to each of the plurality of detection information actually obtained in time series, and based on the weighted information, A signal processing network configured to calculate the fire information; and the fire information calculated when a specific pattern of a plurality of pieces of detection information in the table is given to the signal processing network. The fire information of Fire alarm device, characterized in that it and a adjusting means for adjusting the weighting so as to approximate. 2. A fire alarm device, wherein signal processing is performed on detection information outputted from a fire phenomenon detection means to obtain fire information, and various fire judgments are made based on the fire information. When the obtained plurality of pieces of detection information is input, a corresponding weight is given to each of the plurality of pieces of detection information actually obtained in time series according to the degree of contribution to the fire information, A signal processing network configured to calculate the fire information based on the weighted information, and when a specific pattern of a plurality of pieces of detection information to be obtained in time series is given to the signal processing network. Storage means for storing a weight value adjusted so as to approximate the calculated fire information to the fire information to be obtained when the specific pattern is given, wherein the signal processing network comprises: Stored in storage means It has to perform weighting corresponding to the detection information the is the input using the weighting value fire alarm device, characterized in that.