JP2783753B2 - Detecting new leak sources in a micro leak gas atmosphere - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention detects a gaseous substance such as a gas or a vapor leaked from a plant or equipment in a plant which handles flammable or toxic substances in a gaseous, liquid or solid state. The present invention relates to a method for detecting a newly generated abnormal leak with a low false alarm rate and a low alarm rate even in a micro leak gas (BGL) environment unique to a plant.

[0002]

2. Description of the Related Art In the "High Pressure Gas Control Law" which regulates industries such as the chemical industry and petroleum refining, gas sensors are installed at the outer edge of flammable gas handling facilities at intervals of 20 m or less, and the gas sensors explode. It specifies that it has a detection performance capable of detecting a flammable gas having a concentration of one-fourth or more of the limit, and a contact combustion type flammable gas detector has been generally used as a corresponding detector. Incidentally, the legally set concentration of isobutane, which is a standard indicator substance of a combustible gas detector, is 4500 ppm, and 675 for ethylene.
It is 0 ppm. In recent years, a hot wire semiconductor type having higher detection sensitivity than a contact combustion type has been used as a kind of combustible gas detector. The semiconductor flammable gas detector can confirm the change of the indicated value at a concentration of about 1 ppm with respect to flammable gas, for example, ethylene and isobutane, but other background concentrations are high in an actual plant yard. For this reason, while having a highly sensitive measurement capability, it has not been used for finding small leaks of about several ppm.

The cause of the background density is as follows.
(I) Among the plants that have been operating for many years after construction, there is always a small leak that does not exceed the legally set concentration, and it fluctuates on the order of several PPM. 5 to 1 due to changes in atmospheric temperature / humidity even when almost no reactive gas is present.
This is because there is a "zero point" fluctuation corresponding to 0 ppm.

[0004] Due to the above problems, there has been no early detection method for minute leaks using an industrial gas detector, except for expensive devices such as infrared spectrophotometry and gas chromatography. Was. Accordingly, there is no published method for identifying abnormal leakage in a normal small leakage gas atmosphere.

[0005]

In a plant that has passed a long time since its construction, the concentration is lower than the gas detector emission standard (1/4 of the lower explosive limit concentration) defined by laws and regulations, There is a small amount of leak gas (hereinafter referred to as BGL: background leak). Under such a BGL environment, an increase in gas concentration due to abnormal leakage newly occurring due to breakage of a gasket in a pipe flange or a valve gland does not clearly appear, so that it is highly likely that this will not be detected until a large leak develops. Was.

In general, the false alarm rate and the missing alarm rate have mutually contradictory properties. If the alarm setting is lowered to prevent the false alarm, the false alarm is generated, and if the alarm setting is increased to prevent the false alarm, the false alarm is lost. The information will increase respectively. An object of the present invention is to balance both the false alarm rate and the missing alarm rate rationally,
Identifies new abnormal leaks that occur even in the environment and issues an alarm. The BGL concentration range to be detected by the gas detection system is directly affected by zero point drift due to changes in atmospheric temperature and humidity of the high-sensitivity gas sensor. Therefore, an object of the present invention is to provide a method for correcting the influence and improving the accuracy of the sensor.

[0007]

The object of the present invention is to detect a gas leaking from a plant in a plant handling a flammable or harmful substance in which a very small amount of leaked gas (BGL) is always present. a high sensitive gas sensor group, and wind-wind vane, using a system composed of a data processing system that processes these sensor signals, when certain of detecting a new leak-free normal and new leakage
Time-series data of the detection values of each of the sensor groups and
From the time series data of wind direction and wind speed, the BGL total amount and the sensor
-Average BGL concentration of group and average BGL concentration of each sensor
The degree is estimated respectively, and what is described in the following (1) to (3)
The problem is solved by providing a method for detecting a new leak source in a micro leak gas atmosphere, which is characterized by making a judgment . (1) Average BGL concentration of the sensor group at normal and specific times
If the difference exceeds the threshold defined by the false alarm rate and the missing alarm rate,
In this case, it is judged as a new leak. (2) Average BGL concentration for each sensor at normal and specific times
Error rate exceeds the threshold determined from the false alarm rate and the missing alarm rate.
If the number of sensors exceeds the number determined from the false alarm rate and the
In this case, it is judged as a new leak. (3) The time series data of the BGL total amount at normal time is referred to as data 2.
Then, the time series data of the BGL total amount at a specific time is referred to as data 1.
And non-parametric tests on Data 1 and Data 2
At the significance level considering the false alarm rate and the missing alarm rate
And that the value of data 1 belongs to the group of data 2.
Is rejected at the above significance level, it is determined to be a new leak.
You. In the sensor group of the present invention described above, the following (1),
(2) and (3) procedures for changing the temperature and humidity
It is preferable to compensate for zero point fluctuation of the sensor.
No. (1) Known gas concentration under specific temperature and humidity conditions and
And show the correlation with the indicated value of the sensor.
Create, (2) record the sensor indicated value, (3) gas is the same temperature and humidity as and not present on the real (2)
Under the conditions used in (1) and (2) above
The indicated value of the gas sensor independent of the
Using this, the sensor reading of (2) is corrected, and after correction
Is applied to the reference calibration curve to determine the corrected gas concentration.
You.

There are three methods for statistically processing the time-series data of the gas sensor and extracting the characteristics of BGL as follows. (A) Macromass balance method One of the reasons why the BGL concentration fluctuates irregularly is that gas stays in the yard in a windless state or a light wind of 0.5 m / sec or less within a certain period of time, It is considered that moving in the time section of, for example, has an effect. This method attempts to identify a new leak by defining the total amount of BGL in the yard, which changes every moment due to a change in wind direction and speed, so as to reflect past BGL concentration data.

[0009] Assuming that there is no new springing from the outside or intrusion from the outside, the rate at which the gas present in the target yard decreases is the gas abundance itself η at that time.
(M ³ ), the wind speed v (m / sec), and the average moving distance (pseudo yard radius) L (m) of the gas to reach the outside of the yard can be expressed as follows.

Dη / dt = − (v / L) η (1) From equation (1), reference time t = 0 Assuming that the gas volume at η is η ₀ , η = η ₀ e- ^{(v / L) t} ..... (2) On the other hand, the amount of gas newly generated in the time section from t = 0 to t = T is φ (m ³ / sec), and the gas generated at this time is at the time of t = T The remaining amount in (L / v) φ (1-e- ^{(v / L) T} ) ........ (3) Therefore, considering the data at intervals of T seconds, η _N (m ³ )
After T seconds, the gas becomes η _{n + 1} using (2) and (3).

Η _{n + 1} = η _n e− ^{(v / L) T} + (L / v) φ (1−e− ^{(v / L) T} ) (4) The pseudo yard radius is a plane For a typical plant, L =
(S / π) 1/2, for a three-dimensional plant L =
(3V / 4π) 1/3. Here, S indicates a plant area, and V indicates a plant volume. For a time average over a sufficiently large T, equation (4) can be expressed as:

Η _AV = (L / v _AV ) φ _AV ... (4) ′ The sensor average BGL concentration C _AV is defined as C _AV = μ (L / v _AV V) φ _AV using the above definition. ... (5) μ: expressed as a correction coefficient for determination. Expected BGL concentration from the average value of all sensors actually observed

[0013]

[Outside 1] By monitoring the corrected sensor data after subtracting, a new abnormal leak is detected. That is,

[0014]

(Equation 1) Then, there is no new leak.

[0015]

(Equation 2) Then a new leak occurs.

ΔC is an appropriately set threshold parameter. Here, how to obtain the average time T is determined according to a change in weather conditions specific to the installation location of the system. (B) Spatial correlation method BGL of each gas sensor at a certain sample time interval
Examination of the correlation between the time-averaged values revealed that there was a certain relationship that did not depend on the time, which represented the spatial characteristics unique to the target plant. In the short term, the BGL gas constantly floating in the plant changes in accordance with the wind direction and wind speed data at that time. However, when a time average over a certain period of time is taken, the relationship between the BGL gas sensor data and the number of devices is large. It is thought to represent some pattern associated with the gas movement path in a dense plant.

Now, the total leak amount in the target yard is φ (m3)
Then, assuming that the average BGL concentration Cm of all the sensors is proportional to φ, Cm = Kφ. (6) Therefore, the sum of the BGL concentrations of the N sensors is constant.

[0018]

(Equation 3) Here, if C ₁ changes to 0.5 * C ₁ , it may be possible that adjacent C ₂ and C ₃ simultaneously change to 1.2 * C ₂ and 1.3 * C ₃ , respectively. Therefore, an appropriate weighting coefficient a =
Even if (a ₁ , a ₂ ,..., A _N ) is introduced, it is considered that the constant summation of equation (7) holds.

[A _i ] [C _j ] ^T = constant (Σa _i * C _i = K) (8) [a _i ] is a space that affects the pattern of the BGL distribution Considered a vector. From expression (8), the i-th BGL value is

[0020]

(Equation 4) If you rewrite this,

[0021]

(Equation 5) The BGL concentration of the sensor i can be expressed by a linear sum of the BGL concentrations of the other sensors plus a constant term. Here, the coefficient a _j ′ is a coefficient representing a spatial correlation existing between the sensor i and the sensor j. BGL collected in the past
If there is time series data of concentration, regression analysis will be performed using this data to obtain the estimated BGL value of each sensor.

[0022]

[Outside 2] Equation (9) ′ is obtained. And the concentration C _i of the newly observed sensor i, if focusing on the difference between the sensor i, which is estimated using spatial correlation coefficient far in the following manner it is possible to detect a new leakage.

[0023]

(Equation 6) Then there is no new leak.

[0024]

(Equation 7) Then there is a new leak. Here, k is an integer from 1 to n. φ is 1
It is a binary variable that takes a value of or 0.

[0025]

(Equation 8) Δ is an appropriate threshold. (C) Statistical test method The matching between the fluctuation pattern of the BGL time history measured by the gas sensor (reference data block) and the latest data time series (current block) collected in a certain time interval is statistically performed. (Non-parametric test), for example, run test or Man-Whitney
U test.

Here, a method of determining the presence or absence of a new leak using the U test will be described. Now, the BGL total amount η defined by the previous macromass balance method is calculated at each time,
Create time series data of η . The total number of BGLs in the current block is N ₁ , the BG of the reference block is
It is assumed that the number of L total amount time series data is N ₂ . A total of N ₁ + N ₂ pieces of BGL total amount data are arranged in descending order, and the order of the BGL total amount data belonging to the current block is summed up, and this is set to R ₁ . In _{this, U = N 1 N 2 +} N 1 (N 1 +1) / 2-R 1 E = N 1 N 2/2, V = N 1 N 2 (N 1 + N 2 +1) / 12 Z = (U- E) If Z obtained by V / 2 continuously exceeds the threshold value of the U test more than the set number of times, it is determined that there is leakage.

The determination threshold for Z is determined according to the significance level of the test. For example, in the actual verification test of the inventors, the following is determined. If the amount of new leakage is more than twice the total amount of BGL, the detection probability within 20 minutes is 95% or more, and the false alarm rate is 15% or less. Met. New leak discrimination cycle: 1 minute Number of consecutive tests required for leak detection: 20 Threshold value of U test: 90% significance level Z = 1.65 95% significance level Z = 1.96 99% significance level Z = 2 .57 99.8% significance level Z = 3.09 The correction of the influence of changes in atmospheric temperature and humidity of the highly sensitive gas detector described in claim 2 is as follows. Know the relationship between the concentration of flammable gas in the atmosphere measured under a certain temperature and humidity condition and the indicated value of the sensor (hereinafter referred to as “reference calibration curve”). A sensor indication when the flammable gas concentration is "zero" under different temperature and humidity conditions (hereinafter referred to as "zero point"),
The difference between the “zero point” and the “zero point” of the “standard calibration curve” is subtracted from the sensor instruction at the time of measuring the flammable gas under the condition. (Hereinafter referred to as “correction instruction value”) “correction instruction value”
To the “standard calibration curve” to obtain a correction value of the measured value of the combustible gas. In a flammable gas detection target facility, a sensor installed at a fixed point may have some leakage at the time of measurement, and therefore does not always indicate a “zero point”. For this reason, it is necessary to measure the “zero point” under substantially the same temperature and humidity conditions as the measurement point and under the condition that there is almost no flammable gas. However, it is difficult to place a sensor at a fixed point under these conditions frequently, following day-to-day "zero" fluctuations. Therefore, an independent sensor (hereinafter referred to as a "correction sensor") for constantly measuring "zero point" on behalf of the sensors in the facility is installed, and the fluctuation of the "zero point" is used to correct the fluctuation of other sensors. Used.

A location near the flammable gas detection target facility where the temperature and humidity conditions are close to each other and where flammable gas hardly comes is selected, a high-sensitivity sensor is installed, and a "correction sensor" is used. Although there is no ideal place that completely satisfies these conditions, even under imperfect conditions, the correction effect of the present invention is not completely eliminated. To compensate for these conditions, a method of covering the sensor with a special resin film that selectively allows only moisture in the atmosphere to permeate, and adsorption that captures only flammable gas in the atmosphere that passes through the gas detection unit of the sensor The use of agents is also conceivable.

Since the "correction sensor" serves as a reference for other sensors, it is desirable to install a standby redundant sensor in consideration of failure and reliability of measured values within the limits of cost. However, the information of the zero point drift from the correction sensor is applied to all other sensors, so that only one device is required in terms of function. The accuracy of detection of new leaks by these means is evaluated based on the occurrence rates of false reports (reports when there is no new leak) and missing reports (do not report when new leaks). Assuming that the total BGL amount is Qm3 / hr, the present invention reduces the amount of new leakage by 1.5 times
It can be found with a false alarm rate and a missing alarm rate of 0% or less (defined by the time ratio between the new leakage duration time and the alarm time). However, the target limit false alarm rate and target alarm rate can be adjusted according to the psychological reaction of the plant operator.

That is, in the former macro mass balance method, the threshold value (corrected BG
In the latter spatial correlation method, adjustment can be made by taking a parameter k (when at least k of n sensors exceed a threshold value, it is regarded as a new leak). is there. As the threshold value is increased in the former case and as the parameter k is increased in the latter case, false alarms are reduced, but missing alarms are also increased. Therefore, the method of alarm generation is not treated in a uniform manner. First, a parameter is set so that there is a low false alarm rate even if there are some false alarms, and a warning is issued. It is also a realistic countermeasure to use a two-stage method in which an alarm is issued based on a parameter with a lower false alarm.

[0031]

[Function] Plants that have been operating for many years since construction are highly likely to cause fire or explosion due to flammable gas leakage from aged components such as pipe gaskets and pump shaft seals. There is a need to detect small amounts of gas diffusion. For this reason, industrial gas sensors such as hot-wire semiconductor sensors that can detect a low concentration region far below 1/4 of the lower explosive limit concentration, which is the alarm set value by law, have come to be marketed. High-sensitivity sensors are replacing older sensors whose life has passed.

However, in an actual existing plant, since a very small BGL is always present, even if a highly sensitive gas sensor is installed, a signal indicating a new abnormal leak is generated by the BGL. It was buried inside, and the above needs could not be achieved. In addition, when detecting a new leak in such a BGL environment, an error that is not issued when it should be originally issued (missing report) and an error that is issued when it should not be issued (false report). In other words, we face the problem of how to rationally trade off the two opposing error modes.

The present invention provides a time series data of BGL concentration measured by a plurality of high sensitivity sensors installed in a plant,
In addition, the time series data of the wind speed and direction at that time are statistically processed, the temporal and spatial properties of the BGL specific to the plant are extracted, and the gas concentration value estimated from the statistical properties is calculated as It is possible to detect a new leak by monitoring a time series subtracted from gas sensor data collected from time to time, or by monitoring the statistical properties of BGL itself. Through a simulated leak test in an actual plant, it has been confirmed that the discovery method proposed in the present invention can rationally manage the above-mentioned trade-off between the false alarm rate and the missed alarm rate, and both reduce the trade-off until it can be sufficiently used. .

Further, in order to effectively utilize the discovery of a new leak in a BGL environment, the present invention proposes a method of always automatically correcting a zero-point drift due to a change in temperature and humidity of the gas sensor itself. This makes it possible to more prominently identify a sign of a new leak from a slight change in gas concentration.

[0035]

Next, a leak detection system for gas, steam, etc.
An example in which the present invention is applied to equipment shown in the example of "Wind direction and wind speed measuring device (Japanese Patent Laid-Open No. 05-231979)" will be described with reference to the drawings. However, the present invention is not limited to the embodiment using the multi-point sampling module described below, and a case where a high-sensitivity sensor is used in a small leak gas atmosphere even in a conventional point monitoring type gas detection system. Is applicable. Target Plant: FIG. 1 is a plan view of a plant in which a leak point and a gas leak detection system according to an embodiment of the present invention are installed. This embodiment is an embodiment applied to a liquefied ethylene production plant, and as shown in FIG. 1, L1XL2 = 80 mX
Eight module sets M ₁ , M ₂ , M ₃ , M ₄ , M ₅ , M ₆ , M ₇ , including a sampling module 10 and a sensor module 20 in a 50 m plant yard 40.
M ₈ is installed. Sampling module 10,
And the sensor module 20 are disclosed in
The same one as 79 was used.

The sampling module 10 is located on the ground 2-3.
m, screws are attached to the side of the gas pipe at 2m intervals, 20 air conduits open at both ends, connected to the air conduit, perforated plate at the tip to prevent rainwater intrusion Consists of a conical air sampler,
The gas pipe is connected to the suction pump from the center via the sensor module.

The air collected by the sampling module 10 is introduced into the housing from the sample air inlet nozzle, enters one half of the housing, ascends the space separated by the baffle plate, and opens the cylindrical side of the punch plate. A plurality of holes enter the inside of the sensor, flow around the gas sensor, descend one half of the housing, are drawn by the suction pump from the sample air outlet nozzle, and are discharged to the atmosphere from the pump outlet. The gas sensor in this case is a sensor for detecting ethylene gas. In general, when the present invention is applied to a plant, a sensor of a substance handled in the plant is used. At present, the most desirable model is a hot-wire semiconductor sensor, and various types are commercially available. The higher the detection sensitivity, the better, but preferably about 1-10 ppm.
The detected signal is sent to the data processing device via a signal line. The gas concentration in the atmosphere detected by each sensor module 20 and collected by each sampling module 10 and the data measured by the wind direction and wind speed measurement device 30 are transmitted to a data processing device. FIG. 2 shows a calibration curve for ethylene of a hot-wire semiconductor gas sensor. From this, it can be seen that the sensor indicated value may be reversed depending on the difference in the relative humidity, but has a positive correlation with the absolute humidity. Therefore, if a certain error is allowed, the indicated value can be corrected using the absolute humidity. FIG. 3 shows the principle of the measurement error correction method for the hot-wire semiconductor gas sensor according to the present invention.

The present inventors have studied to approximate the “calibration curve” by an algebraic expression. As a result, the following three equations were obtained. 1) When the ethylene concentration is 5 ppm or less I = kC + I ₀ ..................... (10) 2) When the ethylene concentration is 5 ppm or more I = a (log C) ⁿ + c. ... (11) or I = a ′ (logC) ² + b ′ (logC) + c ′... (12) Here, each character indicates the following.

I: indicated value of sensor (% relative to upper limit) I ₀ : indicated value of “zero point” of sensor (% relative to upper limit) C: ethylene concentration (ppm) k, a, n, c, a ′, b ', C': each coefficient of the algebraic formula Furthermore, since the "calibration curves" having different temperature and humidity conditions are almost parallel in the range where the ethylene concentration is low as shown in FIG. 2, the semiconductor gas sensor due to the difference in temperature / humidity conditions. When the "calibration curves" are moved in the direction of the vertical axis, it was found that they substantially overlap each other.

The relationship between the flammable gas concentration in the atmosphere measured under a certain temperature and humidity condition and the indicated value of the sensor—a “reference calibration curve” has been obtained. Changes the calibration curve. Therefore, if the variation ΔI of the “zero point” due to the influence of temperature and humidity is known from the “correction sensor” described above, the sensor indication value I is replaced with (I−ΔI) and substituted into the calibration curve.
In a region where the concentration is 10 ppm or less, the gas concentration after temperature and humidity correction can be obtained as a fairly good approximation. 1) When the ethylene concentration is 5 ppm or less I−ΔI = kC + I0 (10) ′ 2) Ethylene concentration Is 5 ppm or more I−ΔI = a (logC) n + c (11) ′ or I−ΔI = a ′ (logC) 2 + b ′ (LogC) + c ′ .. (12) ′ Although it is desirable that the frequency of sensor correction by the above method be as frequent as possible, one minute to one minute in consideration of the capability and price of the device used for communication and calculation of measurement data. A range of once per hour is reasonable. The range of 5 to 20 minutes is desirable in Japan.

In the plant shown in FIG.
One sensor of the same type was installed outside the plant side of the office at a distance of m from the office, and it was used as a "correction sensor."
In the vicinity of this "correction sensor", the atmosphere is collected in a polyurethane resin gas sample bag, and within 20 minutes after collection, the "zero point" of all sensors is collected in this atmosphere in order to correctly evaluate the effect of temperature / humidity. It was measured. For confirmation, the atmosphere was analyzed by gas chromatography, and it was confirmed that hydrogen and hydrocarbons were not mixed in the gas. Table 1 shows the "zero" readings measured under different atmospheric conditions. Next, the same 50 liters of air were collected in a gas sample bag, and the ethylene concentration adjusted by injecting and diluting 99.9% or more of ethylene was 0,
Using 5, 20, 50, and 200 ppm of sample gas, the indicated value of each sensor is recorded, and a calibration curve is determined. Table 2
Are the coefficients of the above-mentioned equations (11) and (12) obtained as a result. As shown in Tables 3 and 4, the difference in measured density between when the above correction was not performed and when the correction was performed was 10 pp.
It can be seen that it is considerably reduced in the low concentration region of less than m. New leak detection under BGL environment: By monitoring the corrected sensor data, BGL of 10 ppm or less existing in the plant can be detected. When a simulated leak is artificially generated in such a BGL environment, the data observed by the sampling module indicates, as shown in FIG. 4, when the new leak started and ended. Does not give clear information.

Therefore, the above-mentioned two methods (the macro mass balance method or the spatial correlation method) were used to discriminate new leakage in a BGL environment. For safety reasons, it is not possible to increase the simulated leak rate without limit in a real plant during operation, so sensor data observed in an experiment with a leak rate of 5 m ³ / hr conducted during periodic repair (in an environment without BGL) Was proportionally multiplied to prepare data of a simulated leak test of 5 m ³ / hr or more (an environment without BGL). By overlaying these data with BGL data on a day when the weather condition is similar to that at the time of regular repair, it is possible to create simulated leak test data in a BGL environment.

With respect to these simulated leak test data, the above two methods have what degree of new leak detection accuracy,
Alternatively, in order to evaluate the minimum amount of new leakage that can be detected in a BGL environment, the false report rate and the missing report rate are defined as follows.

False alarm rate: The ratio of the alarm time to the time when there is no new leak within the data collection time.

Missed report rate: The proportion of time during which no report was made to the new leak duration within the data collection time.

When the method proposed in the present invention was evaluated with such a definition, the results were as shown in FIG. 5 (spatial correlation method) and FIG. 6 (macro mass balance method). Improving the false alarm rate and the missed alarm rate will not avoid the tendency of the other to worsen. Of course, the larger the new leakage amount, the clearer the judgment. Now, assuming that the allowable criterion for the determination is “10% or less for both the false alarm rate and the missing alarm rate”, the detection results of the two methods are as follows. At this time, the total amount of BGL is about 5 to 10 m ³ / hr.

Spatial Correlation Method Macro Mass Balance Method (1) Minimum New Leakage Volume 12m ³ / hr 6-10m ³ / hr (2) Recommended Reporting Criteria 6 Threshold 4.0-4.5ppm Parameters in All Modules or 7 Is issued.

The allowable criterion for the judgment assumed here is “false alarm rate,
The validity of "both 10% or less of missed reports" may vary depending on the subjective opinion of the plant operator. Although it is difficult to generalize all the results of the embodiments described here, it is possible to judge a new leak of about 15 m ³ / hr with a considerably low false alarm and missing alarm rate.

[0049]

[Table 1] 5 is a table showing "zero point" readings of a hot-wire semiconductor gas sensor measured under different atmospheric conditions.

[0050]

[Table 2] It is a table | surface which shows the coefficient of the "reference calibration curve" of a hot-wire semiconductor gas sensor.

[0051]

[Table 3]

[0052]

[Table 4]

[0053]

According to the present invention, the statistical properties of the BGL fluctuations are extracted, and the estimated BGL obtained by using the extracted statistical properties is monitored from the gas sensor, and the following effects are obtained. Although it is lower than the reporting standard defined by laws and regulations, it is possible to easily report a new leak even in a BGL environment caused by a small leak that always exists in an actual plant yard. A new leak can be determined with a low false alarm rate and a low false alarm rate even in a BGL environment due to a minute leak.

[Brief description of the drawings]

FIG. 1 is a plan view of a plant to which a detection system according to an embodiment of the present invention is applied.

FIG. 2 is a diagram showing a calibration curve for ethylene gas of a hot-wire semiconductor gas sensor.

FIG. 3 is a diagram illustrating the principle of a measurement error correction method for a hot-wire semiconductor gas sensor.

FIG. 4 is a diagram showing an example of a gas sensor indicated value at the time of a new leak in a BGL environment.

FIG. 5 is a diagram showing a false alarm rate and a false alarm rate of new leak detection in a BGL environment by a spatial correlation method.

FIG. 6 is a diagram showing a false report and a missing report rate of new leak detection in a BGL environment by a macro mass balance method. 9 is a table showing the effect of temperature and humidity correction of a hot-wire semiconductor gas sensor.

[Explanation of symbols]

M ₁ ~M ₇ module 10 sampling module 20 Sensor module 30 wind direction and wind velocity measuring device 40 plant yard

──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Saburo Takahashi 1-6-6 Takasago, Takaishi City, Osaka Prefecture Inside Mitsui East Pressure Chemical Co., Ltd. (72) Inventor Masahiko Tsuchiya 2-8-1 Akanehama, Narashino-shi, Chiba Toyo Engineer Inside the Ring Co., Ltd. (72) Inventor Ippeita Kishiguchi 2-8-1 Akanehama, Narashino-shi, Chiba Toyo Engineering Ring Co., Ltd. (72) Inventor Yosio Kawauchi 2-8-1 Akanehama, Narashino-shi, Chiba Toyo Engineering Co., Ltd. (56) References JP-A-5-231979 (JP, A) (58) Fields investigated (Int. Cl. ⁶ , DB name) G01M 3/00-3/04 G01M 3/20 -3/22

Claims (2)

(57) [Claims] 1. A plant for handling flammable or harmful substances in which a very small amount of leaked gas (BGL) is always present, a group of high-sensitivity sensors for detecting gas leaking from the plant, an anemometer, It uses a system consisting of a data processing system that processes sensor signals and uses a system that detects normal leaks without new leaks and that detects new leaks.
Time series data of the detection values of each of the above sensor groups at a fixed time
From the time series data of wind direction and wind speed,
Average BGL concentration of each sensor group and average BG of each sensor
The L concentration was estimated and described in (1) to (3) below.
Method of detecting new leak source under slight leakage gas atmosphere and performing one of the determination. (1) Average BGL concentration of the sensor group at normal and specific times
If the difference exceeds the threshold defined by the false alarm rate and the missing alarm rate,
In this case, it is judged as a new leak. (2) Average BGL concentration for each sensor at normal and specific times
Error rate exceeds the threshold determined from the false alarm rate and the missing alarm rate.
If the number of sensors exceeds the number determined from the false alarm rate and the
In this case, it is judged as a new leak. (3) The time series data of the BGL total amount at normal time is referred to as data 2.
Then, the time series data of the BGL total amount at a specific time is referred to as data 1.
And non-parametric tests on Data 1 and Data 2
At the significance level considering the false alarm rate and the missing alarm rate
And that the value of data 1 belongs to the group of data 2.
Is rejected at the above significance level, it is determined to be a new leak.
You. 2. A sensor group, the following (1),
(2) and (3) , wherein a correction is made for zero point fluctuation of the gas sensor due to temperature and humidity changes.
2. A new leak under a micro leak gas atmosphere according to claim 1.
How to detect leak sources. (1) Correlation between a known gas concentration under a specific temperature / humidity condition and a sensor reading corresponding thereto is created to create a reference calibration curve, (2) the sensor reading is recorded, and (3) the gas is substantially It is not present and is placed under the same temperature and humidity conditions as in (2) , and the indicated value of the gas sensor independent of that used in (1) and (2) above is taken as the correction zero point ,
Using this, the sensor reading of (2) is corrected, and after correction
To give a gas concentration corrected by applying the indicated value to the reference calibration curve
You.