JP3036217B2 - Flowmeter - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a flow meter for measuring a flow rate of a fluid such as a gas or a liquid such as city gas or LPG gas, and more particularly to a flow meter having a high-precision arithmetic function.

[0002]

2. Description of the Related Art Conventionally, this type of flow meter has a configuration as shown in FIGS. 3 and 4, for example, as disclosed in Japanese Patent Application Laid-Open No. 3-95420.

That is, in the conventional flow meter of FIG. 3, 1 is a flow meter, 2 is a gas pipe, and 3 is a fluidic oscillation element, which generates fluid oscillation by utilizing kinetic energy of the fluid. A sensor 4 detects the frequency of fluid oscillation. Reference numeral 5 denotes a shutoff valve which shuts off gas supply when an abnormal use state is detected. Reference numeral 6 denotes a control device, an example of which is shown in FIG.

In FIG. 4, reference numeral 7 denotes an analog amplifier which amplifies a flow signal detected by the sensor 4. A waveform shaping circuit 8 converts the amplified signal into a pulse signal. Reference numeral 9 denotes a rising point detection circuit which detects the rising of the flow pulse signal. Numeral 10 is a cycle measuring means for measuring the time from the rising point of the flow pulse to the next rising point, that is, the cycle. Numeral 11 is a storage circuit, and the relationship between the pulse constant and the flow rate or the cycle is approximated by n polygonal lines. Unit time storage means 13 for storing
And a correction unit amount means 14 for storing a correction unit amount α of the pulse constant, and a constant storage means 15 for storing a constant term a. These storage means are provided n by n in correspondence with the division of the n polygonal lines. Reference numeral 16 denotes an adder circuit for obtaining and adding a pulse constant K = a + Σα indicating a flow rate per pulse for each cycle. Reference numeral 17 denotes an integrating circuit for integrating the obtained flow rates. Reference numeral 18 denotes a display, which displays the integrated result.

Next, the above-mentioned conventional operation will be described with reference to FIGS. When any kind of gas appliance is used, the gas enters the fluid oscillation element 3 to generate fluid oscillation, and the sensor 4 detects a change in the flow rate. The output signal is amplified by an analog amplifier 7 and converted into a pulse signal by a waveform shaping circuit 8.

The relationship between the flow rate and the oscillation frequency is Q = a · F + b
Given by This is calculated by the equation K for calculating the flow rate per pulse.
= Q / F = a + b · T. Here, K is called a pulse constant, and indicates a flow value per pulse. a and b are coefficients. Since the relationship between the pulse constant and the flow rate or the oscillation frequency is not constant as shown in FIG.
If the period of the flow pulse exceeds the boundary of the polygonal line approximation, the coefficient is changed to calculate the pulse constant. Therefore, the coefficient is set for each line segment. In this case, the pulse constant K is obtained by performing the addition processing without performing the multiplication processing of b · T.

The contents will be described with reference to FIG. First, the rising point detection circuit 9 detects the rising of the flow rate pulse output from the waveform shaping circuit 8. When the rise is detected, the cycle measuring means 10 starts measuring the cycle of the flow pulse. At the same time, the following processing is performed by the addition circuit 16. Instead of performing the calculation of b · T, a unit correction amount α that is much smaller than the value of b · T is added and obtained. The addition is the period T of the flow pulse.
This is performed for each unit correction time t, which is a relatively short time. Therefore, it can be said that α = bt. Therefore, the unit correction amount α is continuously added each time the unit time t elapses from the rising point to the detection of the next rising point in one cycle of the flow pulse. The resulting K = a + Σα is the flow rate per pulse, that is, the pulse constant. Since the relationship between the pulse constant and the cycle of the flow pulse approximates a polygonal line, each polygonal segment has a coefficient a, a unit correction amount α, and a unit correction time t. In FIG. 6, α1 and t1 change at the boundary period T1 to T2, and α2 and t2 change at T2 to T3 at the boundary T2. Therefore, the addition circuit 16 determines whether the period measured by the period measuring unit 10 has reached the boundary period of the polygonal line segment (the period measuring unit 10 measures the pulse period and also measures the boundary period), When the processing enters the area of the next polygonal line section, the coefficient a, the unit correction amount α, and the unit correction time t are changed and the above processing is continued.

The flow rate obtained in this manner is integrated with the integrating circuit 17.
By adding in, the used integrated value is obtained. This integrated value is displayed on the display 18.

[0009]

However, in the above-described conventional configuration, since the pulse constant indicating the relationship between the flow rate and the vibration frequency (or period) is linearly approximated, an error becomes large especially near the boundary of the polygonal line, and the flow rate is reduced. There has been a problem that accurate measurement cannot be performed and the integrated flow rate value is greatly affected.

An object of the present invention is to solve the above-mentioned problems and to provide a flow meter capable of performing accurate flow measurement.

[0011]

In order to achieve the above object, the present invention provides a flow rate detecting means for detecting a fluid flow rate, a flow rate input means previously detected and inputted, and a flow rate detecting means for detecting the fluid flow rate. A neural network model for estimating a function indicating a non-linear characteristic with an output signal and obtaining a flow rate value.

[0012]

According to the present invention, a function for obtaining the flow rate is first estimated by the neural network model means from the output signal of the flow rate detection means for detecting the fluid flow rate and the flow rate value of the flow rate input means which has been inputted in advance. The instantaneous flow rate is calculated from the function obtained in (1) and the output signal of the flow rate detection means, and the integrated flow rate is further obtained.

As described above, a function indicating a non-linear function of the fluid flow rate and the output signal of the flow rate detection means is inferred using the neural network model means, and the flow rate is further inferred from the function and the output signal of the flow rate detection means. The error is extremely small because
In addition, high accuracy can be obtained. As a result, the integrated value, which is the amount of gas used, can also be accurately measured.

[0014]

DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to FIGS. 1, the same components as those in FIG. 3 are denoted by the same reference numerals.

FIG. 1 is a block diagram of a flow meter according to the present invention. In FIG. 1, reference numeral 19 denotes a flow rate detecting means, which generates fluid oscillation using, for example, a fluidic oscillation element 3, and changes the oscillation frequency of the fluid as a pressure-voltage change and a heat-resistance change using, for example, a piezoelectric sensor or a thermistor. Detect, or determine the flow velocity with a hot wire sensor. Reference numeral 20 denotes a flow rate input means which inputs a flow rate corresponding to a signal of the flow rate detection means 19 in advance or simultaneously.

Reference numeral 21 denotes a neural network model, and FIG. 2 shows the configuration of the neural network model. The neural network model means includes a pseudo-synaptic coupling converter 20a, an adder 20b for adding the output from the pseudo-synaptic coupling converter 20a, and a non-linear function converter for nonlinearly converting the output of the adder 20b using, for example, a sigmoid function. 20c and error calculating means 20
d and correction means 20e. The outputs of the flow rate detecting means 19 and the flow rate inputting means 20 enter the neural network model means 21, where a flow rate function indicating the relationship between the flow rate and the flow rate detecting means is inferred, and then the inferred function and the output signal of the flow rate detecting means 19 are output. And obtain the flow rate at that time. 22 is a flow rate integrating means for integrating the determined flow rates to determine an integrated value.

Next, the operation of the above configuration will be described. When the gas or the like starts to be used, the fluid flow rate is detected by the flow rate detecting means 19 in the form of a signal such as a voltage signal, and the detected signal and the fluid flow rate at this time are input from the flow rate input means 20 to the neural network model means 21. . Neural network model 2
In this case, 1 receives two inputs and outputs one output, and the inside of the block has a multilayer perceptron configuration. The neural network schematic means first enters the pseudo-synaptic coupler 20a, and the pseudo-synaptic coupler 20a outputs the weight coupling coefficient ω
A product operation ωi · fi of i and the input signal fi is performed. All signals output from the pseudo-synaptic coupler 20a are added to the adder 2
0b, the addition processing y = Σωi · fi is performed, and the result is output to the non-linear converter 20c and becomes the final signal. Nonlinear converter 2
0 is composed of a nonlinear function called a sigmoid function,
A signal output from the adder 20b is input. The sigmoid function is

[0018]

(Equation 1)

Is defined by Next, error calculation means 2 is calculated from the signal of the flow rate input means and the final signal of the non-linear converter 20c.
At 0d, the correction amount, that is, the error signal (evaluated by the least mean square error method) is expressed by the following equation.

[0020]

(Equation 2)

Calculate from the following. In the correcting means 20e, the pseudo synapse coupler 20a is adjusted so that the error signal ε (f) is minimized.
The correction amount of the weight coupling coefficient ωi is obtained. The correction amount is

[0022]

(Equation 3)

## EQU1 ## By repeating the above processing, the flow rate Q
Showing the relationship between the output signal f and the output signal f
= Δ · f (δ is a coefficient having a non-linear characteristic) is learned, and the output signal Q of the flow rate input means 20 is adjusted so that the value of δ · f matches.

After optimizing the weighted coupling coefficient ωi in this way, the signal of the flow rate detecting means 19 is input, and the flow rate at that time is obtained by the neural network model means 21. The obtained flow rate is integrated by the flow integration means 22, and the integrated value is displayed by the display means 1.
Indicated by 8.

According to the configuration of this embodiment, the coefficient indicating the relationship between the flow rate and the detection signal of the flow rate detecting means 19 can be estimated to the inherent nonlinear characteristic by using the neural network model means 21, so that high accuracy is achieved. Flow rate calculation, and as a result, a flow rate measurement such as an integrated flow rate can be accurately performed.

Although the present invention has been described with respect to a fluidic flow meter for measuring a fluid flow rate, the above description can be applied to other flow meters.

[0027]

As described above, the flow meter according to the present invention has the following features.
Flow rate detecting means for detecting a fluid flow rate, and a neural network schematic means for estimating a function indicating a non-linear characteristic of a flow rate input means, a flow rate inputting means and an output signal of the flow rate detecting means which are detected and input in advance and obtaining a flow rate value And a function indicating the relationship between the flow rate and the detection signal in the neural network model means by the output signals of the flow rate input means and the flow rate detecting means, the output signal of the flow rate detecting means and the function estimated by the neural network model means. , That is, by learning the weighted coupling coefficient of the neural network model means and optimizing the function, then inferring based on the signal of the flow rate detecting means to obtain the flow rate, and then calculating the integrated flow rate from the obtained flow rate Is calculated, the error can be made extremely small as compared with the case of linear approximation, and the flow rate calculation can be performed with high accuracy, so that there is an effect that the flow rate measurement such as the integrated flow rate can be accurately performed.

[Brief description of the drawings]

FIG. 1 is a control block diagram of a flow meter according to an embodiment of the present invention.

FIG. 2 is a block diagram of a neural network model used in the control device of the flow meter.

FIG. 3 is a system diagram of a conventional flow meter.

FIG. 4 is a control block diagram of the flow meter.

FIG. 5 is a characteristic diagram of a control device of the flow meter.

FIG. 6 is a detailed characteristic diagram of a control device of the flow meter.

[Explanation of symbols]

 19 flow detection means 20 flow input means 21 neural network schematic means

Continuation of the front page (56) References JP-A-3-44512 (JP, A) JP-A-5-264317 (JP, A) JP-A-58-66820 (JP, A) JP-A-3-274403 (JP, A) , A) JP-A-63-210713 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G01F 1/00 G01F 1/20 G01F 15/075

Claims (1)

(57) [Claims] 1. A flow rate detecting means for detecting a flow rate of a fluid, a flow rate input means previously detected and input, and a function indicating a nonlinear characteristic between an output signal of the flow rate input means and an output signal of the flow rate detecting means are estimated. Flowmeter comprising a neural network schematic means for determining