JPH08145775A - Earthquake judgment inference device - Google Patents

Abstract

PURPOSE: To perform a sampling at a practical speed, and extract a characteristic quantity followed by judgment and inference by extracting the characteristic quantity by an extracting means on the basis of the sampling data of ON/OFF for n-times obtained by a sampling means every lapse of the period n-times the sampling period. CONSTITUTION: This device is formed of a seismoscope 2 and a microcomputer 3, and the ON/OFF signals from the seismoscope 2 are sampled by a sampling means 6. Three characteristic quantities of ON period, ON time sum, and ON time maximum value are extracted by a characteristic quantity extracting means 4, and on the basis of these characteristic quantities, earthquake judgment is fuzzily inferred by an inferring means 5 every ON period. The characteristic quantities are based on the data sampled from the point of time when the first ON signal is inputted to each point of time when the ON signals are successively inputted, and the inference is performed on the basis of the corresponding characteristic quantity at each point of time. The seismoscope 2 takes the point where the vibration waveform exceeds a predetermined threshold as ON point, and outputs the signal.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an earthquake discrimination / inference apparatus which is attached to, for example, a household gas meter and is suitable for cutting off gas supply when an earthquake occurs.

[0002]

2. Description of the Related Art In recent years, as an earthquake discrimination reasoning device of this type, an ON / OFF switch from a seismic sensor built into a gas meter is used.
There is a method in which a signal is measured, a feature amount such as an ON time maximum value or an ON time ratio is extracted based on the signal, and whether or not an earthquake occurs is discriminated and inferred based on the feature amount. .

[0003]

In such a conventional example, in order to discriminate and infer at a practical speed, data is sampled at a relatively high speed, for example, about 10 msec, and the ON time is set for each sampling. It is necessary to measure such as to extract the feature amount, but a general 8-bit CPU has no problem in processing speed and can handle it.

However, a CPU used as a built-in gas meter is mainly a 4-bit CPU with a relatively low speed and low clock, and it is difficult to sample data at a high speed of about 10 msec to extract a feature amount. is there.

The present invention has been made in view of the above points. Even a CPU having a low processing speed can sample data at a practical speed and extract a feature amount for determination. It is an object of the present invention to provide an earthquake discrimination inference device that enables inference.

[0006]

In order to achieve the above-mentioned object, the present invention is constructed as follows.

The seismic discrimination reasoning apparatus of the present invention according to claim 1 is an ON / OFF signal output means for outputting an ON / OFF signal corresponding to a vibration waveform, and the ON / OFF signal is sampled at a predetermined sampling cycle. And the sampling means, and a feature quantity is extracted based on n times ON / OFF sampling data obtained by the sampling means every time a period of n times the sampling cycle (n is an integer of 2 or more) elapses. And an inference unit that infers whether or not there is an earthquake based on the feature amount extracted by the feature amount extraction unit.

According to a second aspect of the earthquake discrimination inference apparatus of the present invention, in the earthquake discrimination inference apparatus according to the first aspect, the feature amount extraction means is an ON cycle, an ON time sum total and O.
At least one of the three feature amounts of the N-hour maximum value is extracted.

According to a third aspect of the present invention, there is provided the seismic discrimination and inference apparatus according to the first or second aspect, wherein the feature quantity extracting means corresponds to the contents of the sampling data for the n times. The processing procedure of feature amount extraction is different.

[0010]

According to the first aspect of the present invention, the ON / OFF signal is only sampled in a predetermined sampling period, and the ON time or the like is measured for each sampling to extract the feature amount. Rather than every time, sampling data for n times is obtained, the ON time and the like are measured to extract the feature amount and the inference is performed as necessary. Therefore, the feature amount extraction and inference are performed for a plurality of new times. It suffices to use the interval between samplings until the sampling data is obtained. Therefore, even a CPU with a low processing speed can sample data at a practical speed, and based on the sampled data. Then, the feature amount can be extracted and the discriminative inference can be performed.

According to the second aspect of the present invention, whether or not an earthquake occurs is determined based on at least one of the three feature quantities of the ON period, the total ON time, and the maximum ON time. Since it is inferred, high-speed discrimination is possible, and by using the ON period corresponding to the vibration period as a feature amount, erroneous discrimination due to impacts other than earthquakes can be greatly reduced.

According to the third aspect of the present invention, the processing procedure of feature amount extraction is changed in accordance with the content of the sampling data for n times, that is, the processing is performed with the sampling data for n times as one unit. Since the processing is performed, the feature amount can be extracted more efficiently than when the processing is performed for each sampling data.

[0013]

Embodiments of the present invention will now be described in detail with reference to the drawings.

FIG. 1 is a block diagram of an embodiment of the present invention.

The seismic discrimination reasoning apparatus 1 of this embodiment comprises a seismoscope 2 as an ON / OFF signal output means for outputting an ON / OFF signal by vibration and a microcomputer 3, and the microcomputer 3 will be described later. like,
Sampling means 6 for sampling the ON / OFF signal from the seismoscope 2, the ON cycle, the total ON time and O
Based on the feature amount extraction means 4 for extracting the three feature amounts having the N-hour maximum value, it is determined whether or not there is an earthquake based on these feature amounts.
It has a function as an inference means 5 for performing fuzzy inference every N cycles.

Note that these feature quantities are based on data sampled from the time when the first ON signal is input to each time when the ON signals are sequentially input, and the corresponding feature quantity at each time. The reasoning is based on. That is, the feature amount is sequentially updated each time the ON signal is input, based on the data from the time when the first ON signal is input.

The seismoscope 2 is shown in FIG. 2 (B) with an ON point at a point exceeding a predetermined threshold with respect to the vibration waveform as shown in FIG. 2 (A). ON like
It outputs a / OFF signal.

The feature quantity extraction means 4 turns on the seismic sensor 2.
Based on the sampling data from the sampling means 6 obtained by sampling the / OFF signal as described later,
N cycles, total ON time (= ON time 1 + ON time 2 + O
N hours 3 + ...) and three ON time maximum values are extracted. In particular, in this embodiment, ON is used.
Regarding the cycle, the feature frequency is the occurrence frequency of the ON cycle (short, medium, long).

That is, in this embodiment, the following five feature quantities are extracted.

ON time total ON time maximum value ON cycle (short) occurrence frequency = ON cycle is 140 msec
Number of occurrences less than ON cycle (medium) Occurrence frequency = ON cycle is 140 msec
-Number of occurrences of less than 400 msec ON cycle (long) Occurrence frequency = ON cycle is 400 msec
The number of occurrences as described above The reason for extracting the total ON time (total of ON times), the maximum value of ON time, and the frequency of occurrence of ON cycles as feature amounts in this way is as follows. First, O
The total N time is an index representing the duration of vibration, and the ON time output from the seismoscope can be a speed equal to or higher than a predetermined acceleration. At this time, as shown in the following equation, The ON-time total has a correlation with the vibration energy, and thus the ON-time total is an index representing the strength of vibration.

E = (1/2) mV ² (E: Energy m: Mass V: Velocity) E∝V ^{2 Further} , the maximum ON time represents the strength of shaking per half cycle, which is directly below. In the case of a type earthquake, the duration of vibration is relatively short and the energy per half cycle is large, so it is a feature quantity.

Furthermore, it becomes possible to grasp the frequency band (period band) above a predetermined acceleration from the ON period output from the seismic sensor.

Since this ON cycle is 1/2 of the vibration cycle, the ON cycle is shorter than 140 msec (short).
Corresponds to a band with a vibration cycle of less than 0.28 seconds (frequency = 3.6 Hz or more). This band is a band in which seismic waves and shock waves partially overlap, and an ON cycle with an ON cycle of 140 msec or more and less than 400 msec ( Middle) is the vibration period =
0.28 seconds-0.8 seconds (frequency = 1.25Hz-3.5
Corresponding to the frequency band of 0.8 Hz), the dominant frequency of the seismic wave is the largest, so to speak, it is the center frequency band of the seismic wave. 25 Hz) and above,
It becomes the band of the response period (frequency) of the seismic wave response in the soft ground and the skyscraper.

The microcomputer 3 of this embodiment is
It is a relatively low speed 4-bit CPU that is often used for built-in gas meters.
It is difficult to infer by sampling the data and extracting the feature amount for each ms. Therefore, this embodiment has the following configuration.

First, the sampling means 6 for sampling the ON / OFF signal from the seismic sensor 2 is activated by a timer interrupt process of a predetermined sampling period, for example, 10 msec to sample the ON / OFF signal and turn it ON / OFF. Only the judgment is made and the time is not measured.

The feature amount extracting means 4 is n times the sampling period (n is an integer of 2 or more), and in this embodiment, 10 ms.
Every time a period of 4 times ec elapses, the sampling means 6
The above-described feature amount is extracted based on the ON / OFF sampling data for four times sampled by the inference means 5, and the inference means 5 performs fuzzy inference for each ON cycle based on the extracted feature amount as described later. Is to do.

The extraction of the feature quantity by the feature quantity extraction means 4 and the inference by the inference means 5 are performed within the next 40 ms period by utilizing the sampling interval of the sampling means 6.

FIG. 3 is a timing chart of the above sampling process, feature amount extraction process and inference process.

FIG. 3A shows an ON / OFF signal of the seismic sensor 2, FIG. 3B shows sampling processing and sampling data by the sampling means 6, and FIG. 3C shows extraction processing by the feature quantity extracting means 4. The inference processing by the inference means 5 is shown.

The sampling means 6 receives the ON / OFF signal from the seismic sensor 2 as shown in FIG.
Sampling is performed at a sampling cycle of sec. In this example, the sampling data for the first 40 msec period is “1100”, and the subsequent 40 msec sampling data is “1100”.

The feature amount extraction means 4 is arranged to make
Based on the sampling data for four times, the feature amount is extracted, and if the inference means 5 changes from OFF to ON and the ON cycle is confirmed within 40 msec, whether the earthquake is due to fuzzy inference. It is to determine whether or not.

In this example, based on the sampling data "1100" of the first 40 msec, the next 40 mse
The feature amount is extracted by using the period indicated by the diagonal line between the sampling processes, which is the period c.
In this example, since there is no change from OFF to ON in the first 40 msec period, fuzzy inference is not performed.

As described above, in sampling every 10 msec, only ON / OFF signals are sampled,
Since the extraction and inference of the feature amount based on the sampling data are performed in parallel during the period of 40 msec by utilizing the intervals of the sampling process, even the microcomputer 3 having a low processing speed can process the data at a practical speed. Sampling becomes possible, and discriminative inference can be performed by extracting the feature amount.

Further, in this embodiment, in order to efficiently extract the feature amount, the feature amount is not sequentially extracted based on the sampling data for every 10 msec, but 4
Interpret sampling data every 0 msec as a 4-bit binary value, and according to the value, 16 types of processing (work values 0 to 15 in the table below) are branched and time measurement is performed to extract the feature amount. Is.

The table below shows this 4-bit ON / OFF.
The sampling data, the determination content of the sampling data, and the content of the time measurement process for extracting the feature amount based on the sampling data are shown.

[0036]

[Table 1]

As shown in this table, for example, 40 m
If the sampling data of s is "0000", it is determined that OFF continues, and 40 is added to the period measured until then.

If the sampling data of 40 ms is, for example, "0001", it is determined that the cycle has changed from OFF to ON and a new cycle has started, and 30 is added to the cycle measured up to that point to be determined. The period is updated, the period information is updated to set a new period to 10, and the ON time is set to 10.

If the sampling data of 40 ms is, for example, “0010”, OFF → ON → OF
It is determined that the cycle has changed to F and a new cycle has started, and 20 is added to the cycle measured up to that point to make the fixed cycle,
While updating the cycle information to set the new cycle to 20,
Set ON time to 10.

If the 40 ms sampling data is, for example, "0011", it is determined that the cycle changes from OFF to ON and a new cycle is started, and 20 is added to the cycle measured up to that point to make a fixed cycle. , The cycle information is updated to set a new cycle to 20, and the ON time is set to 20.

If the sampling data of 40 ms is, for example, “0100”, then OFF → ON → OF
It is determined that the cycle has changed to F and a new cycle has started, and 10 is added to the cycle measured up to that point to make a fixed cycle,
While updating the cycle information to set the new cycle to 30,
Set ON time to 10.

If the sampling data of 40 ms is, for example, "0101", OFF → ON → OF
It is determined that the cycle has changed from F to ON and the new cycle has started within the 40 ms period (internal fixed cycle), and 10 is added to the cycle measured up to that point to make the fixed cycle. The determination cycle determines that the cycle (short) is 1 wave and the ON time is 10, and the cycle information is updated to set a new cycle to 10 and the ON time is set to 10.

In the same manner, the time measurement processing according to the contents of the 4-bit sampling data is performed.

Since the time measuring process is made different according to the 40 ms sampling data in this way, the ON cycle and the ON time can be calculated at high speed.

Next, fuzzy inference based on the extracted feature amount will be described.

FIG. 4 shows the membership functions of the above five feature quantities.

FIG. 7A shows the membership function of the total ON time, which is 3 for "short", "medium" and "long".
It has two labels. "Short" has a total ON time of 1
Corresponding to less than 40 msec, "medium" corresponds to a total ON time of 140 msec or more and less than 400 msec,
“Long” corresponds to a total ON time of 400 msec or more. For example, if the total ON time is less than 140 msec, the fitness of the label “short” is “1”, and the fitness of the labels “medium” and “long” is “0”.

FIG. 6B shows the membership function of the ON cycle (short) occurrence frequency, which has two labels of "small" and "many", and "small" is the ON cycle ( Short)
Corresponds to 3 times or less of the occurrence frequency, and “many” corresponds to 4 times or more of the occurrence frequency of the ON cycle (short). For example, ON
If the frequency of occurrence of the cycle (short) is 3 or less, the fitness of the label “small” is “1” and the fitness of the label “many” is “0”.

FIG. 7C shows the membership function of the occurrence frequency of the ON cycle (medium), which has two labels of "small" and "many", and "small" means the ON cycle ( During)
Corresponds to the occurrence frequency of 7 times or less, and “many” corresponds to the occurrence frequency of 8 times or more in the ON cycle (medium). For example, ON
If the frequency of occurrence of the cycle (medium) is a value of 7 times or less, the fitness of the label “small” is “1”, and the fitness of the label “many” is “0”.

FIG. 6D shows the membership function of the ON frequency (long) occurrence frequency, which is "small" and "large".
"Small" corresponds to the occurrence frequency of the ON cycle (long) of 3 times or less, and "Large" indicates ON.
It corresponds to four or more occurrences of the cycle (medium). For example,
Assuming that the frequency of occurrence of the ON cycle (long) is 3 or less, the fitness of the label “small” is “1” and the fitness of the label “many” is “0”.

FIG. 11E shows the membership function of the maximum ON time, which has two labels of "short" and "long". "Short" has the maximum ON time. 20
It corresponds to less than 0 msec, and "long" corresponds to the maximum ON time of 200 msec or more. For example, if the maximum ON time value is less than 200 msec, the conformity of the label “short” is “1” and the conformity of the label “long” is “0”.

FIG. 5 is a diagram showing inference rules for earthquake discrimination.

In this embodiment, it is determined whether or not there is an earthquake according to the following seven rules.

Rule 1. if Total ON time is short and ON period (short) frequency is low and ON period (medium) frequency is high then The earthquake rule 1. if Total ON time is short and ON period (short) occurrence frequency is low and ON period (long) occurrence frequency is high then earthquake rule 3. if ON total time is medium and ON period (medium) Occurrence frequency is then Earthquake rule 4. if ON sum total is medium and ON period (short) occurrence frequency is low and ON period (medium) occurrence frequency is high then earthquake rule 5. if ON time total is medium and ON cycle (short) frequency is low and ON cycle (long) frequency is high then the earthquake rule 6. if ON total time is long and ON period (long) frequency is high then the earthquake rule 7. if ON cycle (short) occurrence frequency is high and ON cycle (medium) occurrence frequency is high and ON time maximum value is long then earthquake That is, in this embodiment, the fitness of 1,0 obtained by the above-mentioned membership function When the (grade) is logically ANDed according to these rules and the result is 1, it is an earthquake, and when it is 0, it is not an earthquake.

Next, what kind of earthquake these rules are supposed to be defined will be described below.

First, rule 1 and rule 2 assume an earthquake on a relatively soft ground, and although the total ON time is short, the frequency of ON cycles (medium) is large, so a typical earthquake is generated. It is considered to be vibration with a frequency component. In addition, in order to avoid malfunction, the logical product is taken with the low occurrence frequency of the ON cycle (short).

Rule 3 is based on the assumption of an inland earthquake, has a low ON period (medium) frequency, and has a medium ON time total. Therefore, although the frequency is low, the ON time total is medium. Therefore, it is considered that shaking per unit cycle is strong and dangerous.

Rule 4 and Rule 5 assume a strong earthquake with a comparatively distant epicenter, and the vibration period is medium to long, and the total ON time is medium. And is considered a dangerous earthquake.
In addition, when the epicenter is far, the vibration becomes longer due to the strength of the reflected wave and shaking. Further, in order to avoid a malfunction, the logical product is taken with the small occurrence frequency of the ON cycle (short).

Rule 6 is based on the assumption of an earthquake on soft ground and an earthquake sway in a high-rise building. Since the total ON time is long and the ON cycle (long) occurrence frequency is large, it is a very large energy vibration. it is conceivable that.

Rule 7 is based on the assumption of an earthquake on a solid ground and an extremely damaging earthquake. Since the vibration fluctuation component occurs over a wide band and the maximum ON time is large, It is thought to be a big earthquake.

Next, the operation of the seismic discrimination reasoning apparatus having the above configuration will be described based on the flowcharts of FIGS. 6 to 8.

FIG. 6 is a flow chart of the sampling process of the output of the seismic sensor 2. First, it is judged whether or not an ON signal is input from the seismic sensor 2 (step n1). For example, start a timer of 2.5 seconds. (Step n2) This timer is updated each time an ON signal is input, and it is judged whether or not this timer has timed out (step n3).
Return to

10m when the time is not up
The ON / OFF signal is sampled by interrupt processing every sec (step n4), and it is determined whether or not four samplings have been completed (step n5). When four samplings have been completed, four samplings are performed. The sampling data is written in the area for the earthquake determination process of the memory (step n6), the flag for activating the period calculation process for extracting the feature amount is set, and the process ends (step n7).

FIG. 7 is a flow chart of the cycle calculation processing. First, it is judged whether or not the start flag of the cycle calculation processing is set (step n1). When it is set, the sampling data for four times is obtained. Assuming that the sampling data has been obtained, it is judged whether or not the sampling data is data having a definite period, and further whether or not the data has an internal definite period (step n2). , Work values 0, 8, 1 in the above table
If it is data corresponding to 5, the cycle is added, the flag is reset, and the process returns to step n1 (step n).
3, 4) Data with a fixed period, that is, the work values 1, 2, 3, 4, 6, 7, 9, 10, 11, 1 in the above table.
If the data corresponds to 2, 13, and 14, the definite period is written in the processing area of the memory (step n5),
A new cycle is calculated (step n6), the determination process start flag is set (step n7), the cycle calculation process start flag is reset, and the process returns to step n1 (step n).
8).

When the data has the internal fixed period, that is, the data corresponding to the work value 5 in the above table,
Write the fixed cycle to the processing area of the memory (step n
9), the internal confirmation cycle is calculated (step n10), a new cycle is calculated (step n11), the determination process start flag is set (step n12), the cycle calculation process start flag is reset, and the process returns to step n1. (Step n1
3).

FIG. 8 is a flow chart of the earthquake determination process. First, it is determined whether or not the determination process start flag is set (step n1). When it is set, the cycle calculated by the cycle calculation process is determined. Based on the characteristic amount (step n2), based on the characteristic amount, fuzzy inference is performed as to whether it is an earthquake (step n3), it is determined whether it is an earthquake (step n4), If it is, the judgment processing start flag is reset (step n5) and an output corresponding to the earthquake is given, and if it is not an earthquake, the judgment processing start flag is reset (step n5).
6) Return to step n1. The shutoff valve is closed and the gas supply is shut off based on the determination output indicating that the earthquake.

As described above, in the seismic discrimination inference apparatus, the feature quantity is newly updated based on the data from the time when the first ON signal is input from the seismic sensor 2, and based on the updated feature quantity. It is reasoning.

In this embodiment, attention is paid to the predominant period of the seismic wave / shock wave, and the ON period corresponding to the vibration period is used as the characteristic amount, so that misidentification due to impacts other than earthquakes can be reduced, and vibration energy ON in proportion to
O corresponding to the total time and the vibration intensity per half cycle
Since the maximum value for N hours is used as the feature amount, dangerous vibration can be identified.

For the evaluation of this embodiment, as a result of changing the pipe length of a gas meter having a built-in seismic discrimination reasoning device and hitting a steel ball to give an impact or knocking down a gas cylinder to give an impact, It was confirmed that this seismic discrimination reasoning device is extremely effective, because it does not make a misjudgment due to a shock other than an earthquake.

In the above-mentioned embodiment, the inference is made based on the three feature quantities of the ON period, the total ON time and the maximum ON time, but the present invention uses at least one of these feature quantities, for example, An earthquake may be determined when the maximum ON time becomes long.

In the above-described embodiment, the time measuring process is branched according to the contents of the 4-bit sampling data. However, as another embodiment of the present invention, the time measuring process is performed for each bit of the sampling data. Of course you can go.

[0072]

As described above, according to the present invention, only the ON / OFF signal is sampled in a predetermined sampling cycle, and the ON time and the like are measured every time sampling data of n times is obtained. Then, the feature amount is extracted and inference is performed as necessary. Therefore, the feature amount extraction and inference can be performed in parallel by using sampling intervals during the period until new sampling data for n times is obtained. Therefore, for example, even a CPU with a low processing speed used in a gas meter or the like can sample data at a practical speed, and extract a feature quantity based on the sampled data to extract an earthquake. It is possible to discriminate and infer whether or not.

[Brief description of drawings]

FIG. 1 is a block diagram of one embodiment of the present invention.

FIG. 2 is a diagram showing an output of a seismic sensor corresponding to a vibration waveform.

FIG. 3 is a timing chart of sampling processing, feature amount extraction processing, and inference processing.

FIG. 4 is a diagram showing a membership function of each feature amount.

FIG. 5 is a diagram showing inference rules.

FIG. 6 is a flowchart of sampling processing.

FIG. 7 is a flowchart of a cycle calculation process.

FIG. 8 is a flowchart of inference processing.

[Explanation of symbols]

 2 seismic sensor 3 microcomputer 4 feature extraction means 5 inference means 6 sampling means

Claims (3)

[Claims] 1. An ON / OFF signal output unit for outputting an ON / OFF signal corresponding to a vibration waveform, a sampling unit for sampling the ON / OFF signal at a predetermined sampling period, and n times (n) the sampling period. Is an integer greater than or equal to 2)
Based on the ON / OFF sampling data for each batch,
An earthquake discrimination inference device, comprising: a feature amount extraction unit that extracts a feature amount; and an inference unit that infers whether or not there is an earthquake based on the feature amount extracted by the feature amount extraction unit. . 2. The feature amount extraction means is ON cycle, ON
The earthquake discrimination inference apparatus according to claim 1, wherein at least one of the three feature quantities of the time sum and the maximum ON time is extracted. 3. The seismic discrimination inference apparatus according to claim 1, wherein the feature quantity extraction means changes the processing procedure of feature quantity extraction according to the content of the sampling data for the n times.