JPS5856822B2 - Fluid leak detection device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a device for detecting leakage of limbs from pipelines, oil tanks, etc.

Conventional methods for detecting liquid leakage include TDR (Time Dom), which observes pulse reflections as shown in Figure 1.
A in Ref 1ecto, ry) method has been announced.

As a detection sensor, a leakage detection cable 1 whose characteristic impedance ZO changes due to liquid impregnation is used, a termination resistor 2 is connected to its terminal for impedance matching, and the cable is disposed along a pipeline or a tank.

Then, an incident wave of a rectangular pulse is sent from the observation point PO, and changes in the reflected wave are observed with an oscilloscope A to measure the leakage position distance and leakage width.

If leakage occurs between PL and P2 of the cable, the incident wave at PO, P
The reflected wave at l and the reflected wave at P2 become a, b, and c, respectively, and their composite wave d is observed on the onroscope.

Since the phase velocity Ve of the incident wave is known by measuring the times TI and T2 from this observed waveform, the leakage position distance 11 and the leakage width 12 can be determined as follows.

However, the phase velocity itself is actually an ambiguous value.

Phase velocity is the speed at which the signal travels within the cable,
Strictly speaking, it is the speed at which a signal having only one frequency component travels within the cable.

The same cable has different phase velocities depending on the frequency.

Therefore, when the incident wave is a rectangular wave pulse, the pulse contains many frequency components, so the phase velocity is not uniquely determined.

Therefore, the above equations (1) and (2) cannot be defined strictly.

Equations (1) and (2) are taken on the assumption that the phase velocity is constant for all frequency components.

For example, in a cable whose phase velocity is said to be Ve=0.2 m/nS, there is a 1% error in the phase velocity and T1.
= 10 μsec, then l 1 Niha Also, the phase velocity is different between the leakage section and the non-leakage section,
From this point as well, 11 and 12 cannot be determined accurately.

Furthermore, since the attenuation rate of a signal in a cable differs depending on the frequency of the signal, when a rectangular wave pulse is used as an incident wave, the incident wave is attenuated according to each frequency component of the incident wave, and the incident wave progresses in a distorted manner.

The effect extends to reflected waves.

Therefore, as the leakage position becomes farther away from the observation point, the distortion becomes larger and larger, as shown in Figure 2e.

From these distorted waveforms, the exact leakage point P1. P2
Therefore, TI and T2 cannot be measured accurately.

For example, if the phase velocity is 0.2 m/n S and the time T
If there is an error of 100 nS in l, the voltage to 11 will be several mV.
The S/N ratio is low at ~ several tens of mV, but if the leakage width is small, the reflected waves at 41 points and the reflected waves at 42 points overlap, and as a result, the overall reflected voltage and S/N ratio are low. N
As the ratio became smaller and smaller, it became practically impossible to measure even small leaks over long distances.

In addition, a method has been announced in which the reflected waveform at each point across all cables is memorized in advance, and the leakage position is calculated by subtracting the reflected waveform at the time of leakage from the reflected waveform. Not only does this method have the disadvantage of requiring a large capacity storage device, but it has also been found that when this method is used in practice, it is not possible to obtain accurate values as in theory.

As described above, the conventional detection method has the disadvantage that only the approximate leak location and leak width can be known, but it is not possible to do this at long distances.

The present invention overcomes these drawbacks of the conventional methods, and aims to provide a liquid leakage detection method that can accurately detect the position and width of even a small leakage occurring over a long distance.

The general idea is to divide the leakage detection cable into multiple small sections, find the voltage reflectance in each small section from the sample values of the incident wave and reflected wave at the input end of the incident wave, and then find the maximum reflectance. The leakage start position is determined from the position where the reflectance becomes the minimum or the position where the leakage starts, the phase velocity in the leakage section is calculated from the numerical value of the reflectance, and the leakage width is detected from the phase velocity.

The present invention will be described in detail below. First, the principle of the detection method will be clarified, and then one embodiment of the detection method will be described.

Characteristic impedance Z for cable with characteristic impedance Zo
When a cable of r is connected, reflection occurs at the connected part and the ratio of the reflected voltage to the incident voltage, that is, the voltage reflectance r (
It is known that the reflectance (hereinafter abbreviated as reflectance) is as follows.

Therefore, in the case of FIG. 1, the interval POPI.

If the characteristic impedances of PIP2 and P2P3 are Zo, Zr, and Zo, respectively, then the section POPI (point P1
(excluding point PIP2), section PIP2 (excluding point PIP2), section P
2P3 (excluding point P2), the reflectance is zero, and the reflectances rP1 and rP2 at points PL and P2 are as follows.

Therefore, calculate the reflectance at each point from the incident voltage and reflected voltage at the incident point, and if you find the points where the reflectance is not zero, that is, the points where the characteristic 1 impedance changes, you can find the liquid leakage point.
The leakage section will be found.

Therefore, the cable in Fig. 3a is divided into a large section POQ1 and one small section QIQ2゜Q2Q3, as shown in Fig. 3.
..., QIQI+1, and large section QI+IP3, a total of I+2, are considered to be connected in series.

Here, it is assumed that points PL and P2 are divided so as to coincide with any one of points Q1 to QI+1.

In the large section POQI, QI+IP3, the reflection inside the cable is negligible, but because the section is long, the reduction rate phase velocity changes depending on the frequency.The small section QIQ2.・・・・・・
It is assumed that QIQI+1 is an ideal cable, that is, internal reflection is zero within the section, the phase velocity is constant in all frequency bands, and the attenuation rate is zero.

This assumption can be considered to be valid for small intervals.

At this time, if the characteristic impedance of each section is different, the reflection will occur at points Ql, Q2, . ...It will occur in Q I -1-1 etc.

Thinking like this leads to the following two relationships.

(4) Relationship between sample values of incident voltage and reflected voltage at point PO and equivalent reflection coefficient at point PO.

(B) Relationship between the equivalent reflection coefficient at point po and the reflectance at each point.

Here, the sample values of the incident voltage and reflected voltage refer to the values obtained by temporally sampling the incident voltage and reflected voltage, respectively, and the equivalent reflection coefficient at point PO means each point Ql,
Q2. . . . is the reflectance at point PO when it is assumed that the reflection at QI+1 occurs at point PO.

Of the above two relationships, (4) is generally expressed by ARM
A) It is expressed by a relational expression called a model, and can be obtained by applying, for example, the principle of a Kalman filter, as described later.

As will be described later, (B) can be approximately replaced with an algebraic equation by developing Z (Z) conversion theory.

Furthermore, if the reflectance at each point is known, the phase velocity in the leakage section can be determined, and the leakage position and leakage width can be determined from this phase velocity and sample time.

The above-mentioned matters can be summarized as follows.

(a) Cable large section POQI, I small sections QIQ2
,Q2Q3 ,...,QIQI+1, large interval Q
Divided into I+IP3, each point Ql, Q2 .・・・・・・Q
Let the reflectance at I+1 be an unknown quantity.

Here, the point Q1 should be a point closer to the input terminal PO than the leakage start point P1, and the point QI+1 should be a point closer to the terminal P3 than the leakage end point P2.
It is necessary to choose a point close to .

When points PI and P2 are unknown, point Q1 becomes point PO and point QI+
Divide it so that 1 becomes point P3, roughly point PL, P
2, and then set the point Ql, so that it meets the above conditions.
After determining QI+1, points PL and P2 are determined again using the same method.

(b) Observe sample values of incident voltage and reflected voltage,
For example, a Kalman filter is applied to find the equivalent reflection coefficient at point PO.

(C) Calculate the reflectance at each point from the equivalent reflection coefficient at point PO.

(d) Extract points that are not zero among the reflectance at each point.

This point corresponds to a point where the impedance changes.

(e) Calculate the phase velocity in each small section from the reflectance at each point.

(f) Calculate the leakage width from the non-zero points of the phase velocity and reflectance, and the sample time.

Through the above procedure, the length of the section POQI is known, and (
Since the leakage width is determined by f), the distance from the point PO to the impedance change point, that is, the leakage position and leakage width are calculated.

Note that the Z transformation is generally called a sample value transformation, and is a Laplace transformation of a system in which observed values are given as sample values.

The Laplace transform gives a complete description of the target system.

In other words, while it is an expression for all frequency domains, in Z-transform, if the sample time interval is T, the target system can only be expressed up to the frequency band tear, but when the sample time interval T is set to zero, Since it matches the usual Laplace transform, it can be considered that if T is chosen sufficiently small, it is possible to provide an almost complete description of the target system.

Therefore, calculation errors caused by changes in phase velocity and attenuation rate due to frequency, which were a problem with the conventional TDR method, are almost eliminated.

Furthermore, it is known that when noise is superimposed on the observed value, the output of the Kalman filter gives an estimated value in the sense of least squares error, and even if noise is added to the observed values of the incident voltage and reflected voltage, The most probable value can be obtained as the equivalent reflectance of point PO using a Kalman filter.

Therefore, even if there is a slight leak over a long distance, the reflected voltage is small, and the observation noise is large, the leak location can be accurately determined.
The leakage width is calculated.

Now, let's go back to Figure 3 and explain in more detail the relationship between characteristic impedance and reflectance at each point on the cable.

As shown in Figure 3, the cable is divided into (1+2) sections POQI, QIQ2, ......, QIQI.
+1, each characteristic impedance at QI+IP3 is WO,
When Wl , W2 . . . , WI+1, the reflectance r1 . r2. ..., rI+1 is.

For example, if a cable with a characteristic impedance of 50β impregnates the limb from point Q3 to point Q6 and the characteristic impedance becomes 40Q, then WOW1=W2=50!2. W3=W
4=W5=4CI.

Since W6=W7-...-WI+1=50Q, substitute this value into equation (6) and get rl=r2=
0 +r3--0.11. r4=r5=o, r6-+
0゜11, r 7= r 8== r I+1 =
It becomes 0.

When this is expressed graphically as shown in FIG. 4, it is clearly seen that the reflectance has a minimum value at the leakage start point Q3 and a maximum value at the leakage end point Q6.

Therefore, if the reflectance of each point is calculated, then the minimum value and maximum value are found, and the corresponding point is found.
The leak start point and leak end point are determined.

In addition, the characteristic impedance of each point can be calculated by transforming equation (6). More sought after.

Next, the relationship between the equivalent reflection coefficient at point PO and the reflectance ri at each point is determined.

Focusing on the section POPI, the incident voltage at point PO is e(t)
At this time, let the incident voltage at point Q1 be el(t).

el(t) is the voltage across the point Q1 when it is terminated with a characteristic impedance.

The Laplace transform of e(t), e1(t) is converted into E(S), E1(S
) and the transfer function of both is W O (S), then the following holds true.

Here, WO(S) is determined by the cable-specific characteristics of section POQ1.

When formula (8) is Z-transformed, E 1 (Z) = WO (Z) E (Z) ...... (
9) Here, Z is the Z (Z) operator, E 1 (Z),
WO(Z).

E(Z) are E 1 (S) and WO(S) respectively
, represents the Z transformation of E(S).

Similarly for the reflected voltage, the reflected voltage Vl(t) at point Q1
If that becomes V(t) at point PO, then v 1
(t), the Laplace transform of V(t) as v 1(S)
, can be written as V(S).

In this case the transfer function is the same as that of the incident voltage.

When formula (10) is Z-transformed, here V(Z), V
1 (Z) are the Z transformations of V(S) and V 1 (S), respectively.

The input impedance at point Q1 (the value obtained by dividing the input voltage at Ql by the current) is ZQl, and its Laplace transform is Z Q
1 (S), WO is the characteristic impedance of the section POQ1, and in the equation that can be written as It is a display.

The incident voltage and reflected voltage at point Q1 are E 1 (S), V
1(S), this holds true.

Similarly, the incident voltage, reflected voltage, and equivalent reflection coefficient at point Q2 are expressed by Laplace transform as E 2 (S), respectively.
When r 2 (S) and F (2), holds true.

By the way, since the reflectance at point Q1 is rl, reflection rel(t) occurs at point Q1.

Furthermore, the reflected voltage at point Q2 is assumed to be V2(t), and when this voltage reaches point Q1, if section QIQ2 is an ideal cable, there will be a delay of a certain time τ, V2(t-τ).

Part of the voltage that has reached point Ql -r12 (t-7) is reflected again, and the remaining (1-r1)V2 (t-τ) passes through.

Therefore, the reflected voltage at point Q1 is V1(t)-r1e1(t)+(1-rl
)V2(t-τ)·(15).

Next, when the voltage el(t) incident on point Q1 passes through point Q1, it becomes (1+r 1 ) e 1(t), and the sum of this and the re-reflection rlV2(t-τ) is delayed by a certain time τ. It will appear at point Q2.

Therefore, if the incident voltage at point Q2 is e2(t), (15
) and (16) are respectively Laplace transformed, (1
3) Obtain from equations (14), (17), and (18).

According to the Z conversion unit, the sample time interval T2τ of equation (19)
The Z transformation of is expressed as .

z-1 is the inverse Z-transform. Therefore, from equation (13), Fl(
Let F 1 (Z) be the Z transformation of S).

Fl(Z) is the Z-transformed representation of the equivalent reflection coefficient at point Q1.

Equation (21) shows that, in general, the equivalent reflection coefficient Fi(Z) at point Qi is as follows.

Here, the input impedance ZQi+1 at point Qi+1
+1 (since it is a force, substitute Fi+1(Z) for the citizen and note that 1 ril < 1, IFil <, (
21) Approximating the equation to a quadratic term results in the following.

Next, it is known that the relationship between the incident voltage e(t) at point PO and the incident voltage el(t) at point Q1 is approximately expressed by where T is the sampling time interval.

Here, K is the number of samples, Gj+1 j=0, 1
,..., J is a scalar quantity called an impulse response, which is information specific to the cable, and includes information on the attenuation rate and phase velocity depending on the frequency.

J is the order of the impulse response.

Also, since the Z transformation in equation (23) is represented by *, WO(Z) in equations (9) and (11) is It is expressed as

On the other hand, from equations (9), (11), (20), and (21), it can be expressed as F(Z) F 1 (Z) (WO(Z) ) 2 .

Here, F(Z) is the equivalent reflection coefficient at the point of incidence PO.

moreover, From (23), (25), and (27), Let us express this as follows.

That is, al, -aN+1, b2. −, b I+
The numerical value shall be such that 14 & 1 holds true.

Then, equation (28) becomes, and equation (29) and (32) become the point of incidence PO.
Expressions (30) and (31) express the relationship between the equivalent reflection coefficient and the reflectance at each point.

Next, let us find the equivalent reflection coefficient at point PO from the sample values of the incident voltage and reflected voltage.

It is known that when the Z-transformer performs inverse Z-transformation of equation (32), the following equation is obtained.

Here, T represents the sample time interval and K represents the Kth sample.

Equation (33) is generally a relational expression called an AI (MA model).

For simplicity, v(KT) = vk, e((k−j
)T)-ek-j, N+1-1- I-dimensional vector Φ, fk to Φ'-[al,...
aN+1. b2.・・・−・bl+1), fk'=
(ek, -, e k-N, v k,...v
k-1].

However, ′ represents the transposed vector.

At this time, equation (33) becomes.

Generally, in a vector equation such as equation (34), vk,
When Φk is known, the method for finding fk' is as follows:
There are methods such as steepest gradient method, learning identification method, Kalman filter, etc., but here we will describe the method using Kalman filter.

Since vk and ek j are the observed values sampled at the th and k-jth, respectively, they must be considered to contain noise.

Therefore, equation (34) adds noise nk, It is expressed as

Equation (35) is called the observation equation. On the other hand, (30),
From equation (31), it is assumed that the reflectance does not change over time (actually, the sample time interval is several ns to several tens of ns, and KT is on the order of several tens of microseconds, so the reflectance does not change within this time. ) and a 1 r
a N + 1 + b 2 +...bl+
This is because 1 is a constant.

Equation (36) is called a system equation. (35),
When the system equation and observation equation are given as difference equations, as in equation (36), a discrete-time Kalman filter can be applied, and Φ can be estimated in the sense of least square error using the following series of recurrence equations. It is known that

Ak is the (N+1+I)XI Kalman filter gain matrix Φ is the estimated value of Φ, and Ck is (N+1+I)X(N+
1+1), the covariance matrix, R, of the estimation error has a noise n
It is the variance of k.

(37) (39) Formula 8
To find Φk, k--1
1 of the estimated value in and the covariance matrix of the estimation error C-
1, is assumed.

Assuming that the reflectance at each point is initially zero for platform-1, a 1 s...aN-1-1, b2
...bI-1-1 is zero, so platform-1 is 0
It is appropriate to use it as a vector, and the estimation error for this is given by equation (39), rl + """ + r I + 1
, gl , --, g I+1 is smaller than 1, so al
, , −−゛・aN+1. b2.・・・・・・b1+1
It is good to consider the estimation error to be 1 or more, so C
-1, the diagonal elements are 1 or more (N+1+I)X(N+
1+1)? It would be appropriate to take Trix.

Assuming Q-1, CI in this way, when = 1 (3
From formula 7), A1゜(38) A-ori Ql, (39)
Calculate C1 from the formula. When K2, substitute C1 into formula (37) to find A2, and then sequentially find Ak, Gk, and Ck.

When calculated in this way, if each element of Ck becomes sufficiently small, Ck is a covariance matrix of estimation errors, so ek at this time means that the estimation error is sufficiently small. Therefore, Qk can be calculated by Ck. Each element of ek is found within the given error range, al,...8aN+1, b2,...b Estimated value of I+1, a
l.

Eight eight eight...+ a N + 1. r b 2 ,・
...b I+1 will be found.

Among the elements of ek, al...aN+188 is 11, r2, ---, r
The estimated value to I+188 is rl, r2+...rJ+1
Since there is a relationship between hl and h2.・・・・・・
, hL+1. L2J+1 (gl+g2Z '+...
+g J+I Z') 2=h1+h2Z-'+・
・・・・・・+hL+IZ L・・・・・・(41) If it is an impressive numerical value, rl, r2.・・・・・・rI+1
are in the order, i.e. (37), (38), (39)
The equivalent reflection coefficient of the point P6 is determined by the formula, and the reflectance of each point is determined by the formula (42).

Also, without finding the equivalent reflection coefficient at point PO, (28
), the reflectance at each point can be directly determined from equation (29), so we will describe the method.

From equations (28), (29), (31), and (41), the following *main equation is obtained by inverse Z-transforming equation (43).

In the above equation, v(KT)-vk, (2I+1)-dimensional vector Φk is set to Φ'-[rl, r2. ......, r1+1
゜b2.・・・b I+1 ), (21+1) dimensional vector fkL
Lfk′−[Σ hl+1e((k−#)T),
−, Σ hzA=0
1=0+xe((K-1-l)T), v((K-1)
T), -.

By setting v((KI)T)) and adding noise nk, the observation equation is derived.

Since each element in Φ′ is a constant, the system equation can be set as follows.

Since equations (45) and (46) are the same relational equations as equations (35) and (36), the Kalman filter adaptation equation (3
7) Using , (38) and (39), the estimated value of Φ must be found, but (1+1 in 79)
) to the elements are rllr21...+ r I +
1, so in this case, the reflectance of each point can be directly determined.

Finally, the leakage position is determined from the reflectance at the connection point of each section.
We will explain how to calculate the leakage width.

Here, we will discuss the case of coaxial cable.

It is known that the characteristic impedance ZO and the phase velocity Ve are expressed as follows, where the outer conductor diameter is d2, the inner conductor diameter is dl, the magnetic permeability of the conductor is μ, and the dielectric constant of the insulator is ε.

Therefore, even if the dielectric constant changes with the same shape, becomes constant.

Now, the reflectance at each point takes a minimum value at the point Qil, and that value is taken as a maximum value at the 12 point Qi2, and that value is defined as Vi2.

! 1 < 12.

At other points, the reflectance is set to zero or a value close to zero.

Until this time interval QIQi 1, there is no change in the characteristic impedance, and W],=W2Wil-1=ZO.

ZO is a value given by equation (47).

In the section QilQi2, the characteristic impedance changes and Wi
1 = Wi 1-1-1-...Wi 2-
It is 1\20.

Similar to equation (7), It is expressed as

Therefore, from equation (49), the phase velocity V in the interval QilQi2
e(QilQi2) becomes.

On the other hand, since the signal delay time in section QIQ2 is τ, the length QIQ2 of section QIQ2 is because the sample time interval is 2τ when the phase velocity is Ve.

Expressed as a general formula, the interval Q i I Q i 2, from equation (51) becomes.

Therefore, the distance 11 from the point PO to the leakage start point Q i 1 (Pl in FIG. 1) and the distance 12 from the point Qi 1 to the leakage end point Q i 2 (P2 in FIG. 1) are calculated as follows.

If the reflectance is only a very thick value or a very small value, it means a cable break or a cable short circuit, respectively, and 12 has no meaning.

In addition, to provide a supplementary explanation about the sampling time interval T, in general, to sample a signal with a frequency band of fHz or less,
It is necessary that the sample time interval be upward and downward.

On the other hand, from equation (53), the length of the small section when the sampling time interval T is VeT, so the leakage position detection accuracy ■eT
2 is as follows.

2- Therefore, how to select T is determined by the frequency band of the incident voltage reflected voltage and the detection accuracy of the leakage position.

For example, when Ve=0.2m/nS, the frequency band of the signal is 2
0MHz1 Assuming that the detection accuracy of the leakage position is 3m1
2A, then T<w7-30f nS, and T should be selected to be 30 nS or less.

FIG. 5 shows an example of a configuration for realizing the detection method of the present invention.

1 is a leakage detection cable, and 2 and 3 are terminal resistors, respectively.

4 is a pulse generator, 5 is a receiving amplifier, 6 is an AD converter, I
is a timing generator that issues a pulse generation command to 4 and a sample command signal at a fixed time interval to 6. This sampling time interval is necessary for the frequency band of the voltage at the input point and leakage detection as described above. It is determined by the positional accuracy, etc.

8 is an arithmetic control unit composed of a microprocessor;
9 is a device that stores a program for operating 8 according to a predetermined procedure; 10 is a device that stores the impulse response of the cable; 11 is a monitor display device in the event of a leak; and 12 is a central device that collects leak information. It is a control device.

In this configuration, for example, the arithmetic and control device 8 starts the leakage detection operation in response to a command from the central control device 12.

First, 8 issues a pulse voltage sending command to the timing generator 7, 7 sends a pulse generation command signal shown in FIG. 6g to the pulse generator 4, and at the same time, the A/D converter 6
As shown in the figure, sample command signals are sent out at fixed time intervals T.

4, when a pulsed voltage is applied to the cable in accordance with a pulse generation command signal, an incident voltage and a reflected voltage appear at the input point as shown in C in the figure.

A receiving amplifier 5 amplifies the voltage at the point of incidence, and 6 samples the voltage in accordance with a sample command signal, converts it into a digital quantity, and inputs it to 8.

The sample value of the voltage at the incident point obtained in this manner is shown in d of the same figure.

8 is the sample value of d in the same figure e. As shown in f, it is separated into a sample value of the incident voltage and a sample value of the reflected voltage.

Next, connect the cable to the large section POQI as shown in Figure 3.
, small interval QIQ2 , Q2Q3 , ...QIQ
1+1, divided into large section QI → -IP3, and each point Ql,・
...Reflectance at Ql+1 rl, ...r
Assume l+1.

Here, the leakage section PIP2 needs to fall within the section QIQI+1.

When such points Ql and QI+1 are unknown, the method described later can be used.

Since the impulse response of the large interval POQI is stored in the storage device 10, this impulse response and FIG.
When using the Kalman filter, the observation equation (45) and the system equation (46A) can be constructed using sample values of the incident voltage and reflected voltage as input information.
When equations (37), (38), and (39) are solved using the arithmetic program stored in 9, as shown in Figure 6g,
Find the reflectance of each point.

In the example of g, the opposite car has Q3 at the minimum value r3Q6 and the maximum value r
6, the leakage position is from equation (56): 1=POQ
1 is found.

This information is displayed on the monitor display device 11 or sent to the central control device 12.

Now, when the points Ql and Ql+1 are unknown in the above-mentioned cable division, the following method can be used.

Let us divide the cable so that the points Ql and QI+1 coincide with the points PO and P3, that is, divide the cable into small sections, and calculate the reflectance of each point using the same method as described above.

Assume that the result is shown in Figure 6i. The reflectance obtained at this time is the reflectance at each point assuming that the cable is ideal.

Therefore, take the minimum value ail at point P1 and point P
Even if it takes an extremely thick value at 2', the interval P 1'P 2'
Although it cannot be guaranteed that is an accurate leakage interval, the interval x
(il-1), the length of section PIP2' is POP2'
■Since eT(12-1), point Q1. As a way to take QI+1, POQI<POPI'.

It can be seen that POQI+1>PIP2' must be satisfied.

Therefore, we empirically calculated the calculation error △11°△12 and obtained POQ1=POPIL△11. POQI+1-POP2
If the points Ql and QI+1 are determined so that
2 can be entered.

As described above, the detection method of the present invention considers a cable as having multiple small sections connected, observes sample values of incident pulses and reflected pulses at fixed time intervals, and uses these sample values as input information. By calculating with a microprocessor, the positions where the reflectance and reflectance at each connection point of the above-mentioned small section are minimum and maximum are calculated, the leakage start position is determined from the position where the reflectance is minimum or maximum, and the reflectance is calculated. The phase velocity in the leakage section is calculated from the value of , and the leakage width is determined from the phase velocity.

Therefore, it has the following advantages compared to conventionally known detection methods.

■ As a calculation method, for example, Kalman filter can be used, so even if noise is added to the observed values, it can be detected reliably.

■ When considering the cable as a large section and several small sections, the impulse response of the cable is taken into account for the large section, which reduces the ambiguity of the calculation results as in the conventional TI)R method. .

■ When compared to the TDR method, which stores the reflected waveform when there is no leakage as a base data over the entire length of the cable, this method can be used because it is only necessary to store the impulse response specific to the cable regardless of the cable length. Requires less memory capacity.

■ The leakage section can be accurately determined by calculating the phase velocity in the leakage section, which was impossible with the conventional TDR method.

■ The incident/glucose does not have to be an exact pulse waveform,
Since it is only necessary to accurately measure the voltage waveform at the point of incidence, the circuit design of the pulse generator is simplified.

■ Even if the reflectance of the cable changes continuously and gradually, the reflectance at each point, especially the impedance, can be determined.

(2) Due to the effect of (2), if the characteristic impedance due to impregnation with various liquids is known in advance, the type of leaked liquid can be determined.

[Brief explanation of drawings]

Figure 1 is a diagram of the measurement principle of the conventional example, Figure 2 is a diagram of each pulse waveform obtained by the operation of Figure 1, Figure 3 is a conceptual diagram of the structure of the cable used in the method of the present invention, and Figure 4 is the diagram of Figure 3. 5 is a basic configuration diagram showing an example of implementation of the detection method according to the present invention, and FIG. 6 is obtained by operating the embodiment of FIG. 5. Each waveform diagram. DESCRIPTION OF SYMBOLS 1... Leakage detection cable, 2... Upper dance matching paper end resistor, 4... Incident pulse generator, 5... Receiving amplifier, 7 ... Timing generator that outputs sample command signal, 8...
Microprocessor, 9...Program storage device, 10...Cable impulse response storage device.

Claims (1)

[Claims] 1 Disposed along the transport path of the fluid to be detected,
a leak detection cable whose terminal end is impedance matched and whose characteristic impedance changes due to penetration of the fluid to be detected; a pulse generator that applies an incident pulse to the leak detection cable; and a reflected pulse provided at the starting end of the leak detection cable. a receiving device that receives the incident pulse; a timing generator that instructs the pulse generator to send out the incident pulse and samples the output of the receiving device at regular intervals; an AD converter that converts the output of the receiving device into AI) and outputs it; a reflectance calculation means that calculates the reflectance of the leakage detection cable from a change in the AD conversion output of the AD converter; detection phase velocity means for detecting a change in the phase velocity of the leakage detection cable based on the output of the means; extreme value detection means for detecting a minimum or thick value of the reflectance based on the output of the reflectance calculation means; A fluid leakage detection device comprising a leakage point calculation means for calculating a leakage start position based on the output from the value detection means and a leakage width based on the output from the phase velocity detection means.