JP2021043165A - Flow pattern evaluation device for gas-liquid mixed phase flow, flow pattern evaluation method, and gas production system - Google Patents

Abstract

To accurately evaluate a flow pattern in a gas-liquid mixed phase flow of a liquid and a gaseous matter flowing in a pipe.SOLUTION: A flow pattern evaluation device includes: an imaging unit for imaging a gas-liquid mixed phase flow passing a prescribed observation position of a pipe; an image acquisition unit for acquiring an image of the gas-liquid mixed phase flow when the gas-liquid mixed phase flow passes the observation position, from the photographed image captured by the imaging unit; a machine learning unit comprising an evaluation module in which a plurality of mixed phase flow images, which are obtained by variously changing formation conditions of a gas-liquid mixed phase flow of the liquid and the gaseous matter, are utilized as input data for learning, and in which classification results of flow patterns of a gas-liquid mixed phase flow in the plurality of mixed phase flow images are utilized as output data for learning, in order to machine-learn the relationships between the mixed phase flow images and the classification results of the flow patterns; and an evaluation unit which utilizes the images of the gas-liquid mixed phase flow acquired by the image acquisition unit, as images for evaluation, in order to evaluate the flow patterns of the gas-liquid mixed phase flow when passing the observation position, from the images for evaluation, by using the evaluation module.SELECTED DRAWING: Figure 1

Description

The present invention relates to a flow mode evaluation device for evaluating a flow mode in a gas-liquid mixed phase flow of a liquid and a gas flowing in a pipe, a flow mode evaluation method, and a gas production system using the flow mode evaluation device.

Generally, in a gas-liquid mixed phase flow of liquid and gas flowing in a pipe, the ratio of the flow rate of liquid and gas in the pipe, the moving speed of liquid and gas, the pressure in the pipe, the density and viscosity of liquid and gas, and the type of liquid. The flow mode changes depending on the factors. The flow mode is, for example, a bubble flow mode in which bubbles are composed of only fine bubbles in a liquid and fine bubbles are dispersed in a continuous liquid, or a liquid portion containing large bubbles and fine bubbles spread over a cross section of a flow path. Includes the style of slag flow that flows alternately along the flow path.
For example, the amount of liquid discharged from the pump is changed by such a difference in the flow mode. Therefore, in order to control the amount discharged by the pump within an appropriate range, the flow mode of the gas-liquid multiphase flow in the pipe should be evaluated. Is important.

For example, the flow pattern of gas-liquid two-phase flow in a pipe can be objectively discriminated from the outside with a high accuracy rate (accuracy) without affecting the pipe and its flow. A flow pattern discrimination method is known (Patent Document 1).
In this discrimination method, the sound of gas-liquid two-phase flow or the strain of the pipe generated by gas-liquid two-phase flow is continuously recorded as data from the outside of the pipe through which the gas-liquid two-phase flow flows. The recorded data is chaotically analyzed to obtain the fractal dimension D and the cormogoloventropy K, and the flow pattern of the gas-liquid two-phase flow is discriminated from the sizes of the obtained fractal dimension D and the cormogoloventropy K.

Japanese Unexamined Patent Publication No. 2003-344363

However, with the above discrimination method, it is difficult to accurately acquire data on the sound of gas-liquid two-phase flow caused by the noise of the external environment, and the piping that occurs due to changes in the external environment surrounding the piping. It is difficult to accurately obtain data on piping strain generated by gas-liquid two-phase flow due to the influence of the strain of gas-liquid two-phase flow, and as a result, it is difficult to accurately discriminate the flow pattern of gas-liquid two-phase flow. ..
In particular, in the case of gas-liquid multiphase flow in a pipe in a gas production system that decomposes gas hydrate in the seabed to produce gas, noise such as noise due to the ocean current near the seabed and distortion of the pipe due to changes in water pressure due to tidal current. Since the components are included in the sound and distortion data used for discriminating the flow pattern of the gas-liquid two-phase flow, it is not possible to accurately discriminate the flow pattern of the gas-liquid two-phase flow in the pipe.

Therefore, the present invention provides a gas-liquid mixed phase flow flow mode evaluation device, a flow mode evaluation method, and a gas production system capable of accurately evaluating the flow mode in a gas-liquid mixed phase flow of liquid and gas flowing in a pipe. The purpose is to do.

One aspect of the present invention is a flow mode evaluation device that evaluates the flow mode of a gas-liquid mixed phase flow of a liquid and a gas flowing in a pipe. The flow mode evaluation device is
An imaging unit that captures the gas-liquid multiphase flow passing through a predetermined observation position on the pipe,
An image acquisition unit configured by a computer and acquiring an image of the gas-liquid multiphase flow when the gas-liquid multiphase flow passes through the observation position from an image captured by the image pickup unit.
A plurality of mixed-phase flow images formed by a computer and obtained by variously changing the conditions for generating the gas-liquid mixed-phase flow of the liquid and the gas are used as learning input data, and the flow of the gas-liquid multiphase flow in the plurality of mixed-phase flow images. A machine learning unit including an evaluation module in which the relationship between the multiphase flow image and the classification result of the flow mode is machine-learned using the format classification result as learning output data.
The flow mode of the gas-liquid multiphase flow when passing through the observation position using the evaluation module using the image of the gas-liquid multiphase flow configured by the computer and acquired by the image acquisition unit as an evaluation image. With an evaluation unit that evaluates from the evaluation image,
To be equipped.

The image acquisition unit captures a plurality of images of the gas-liquid multiphase flow passing through the observation position in time series at a predetermined frame rate, and images a plurality of images of the gas-liquid multiphase flow at the observation position in chronological order. By joining them together, it is preferable to create one still composite image of the gas-liquid multiphase flow as it passes through the observation position as the evaluation image.

Each of the multiphase flow images used as the input data for learning captures the gas-liquid multiphase flow passing through the observation region having the same size as the observation position in time series at the frame rate, and the said in the observation region. It is preferable that the image is one static composite image of the gas-liquid multiphase flow created by connecting a plurality of images of the gas-liquid multiphase flow in the order of time series.

When the superficial velocity of the gas in the pipe is Vg [m / sec], the frame rate is R [frame / sec], and the length of the observation position along the pipe is L [m]. vg / (R · L) is 1.42 × ^{10 -3} or more 1.05 × 10 ⁴ or less, it is preferable.

The image of the gas-liquid multiphase flow acquired by the image acquisition unit is preferably an image of an ascending flow rising in the pipe.

The evaluation module displays a plurality of images in which at least one of the moving speed of bubbles in the gas-liquid multiphase flow, the size of bubbles in the gas-liquid multiphase flow, and the density of the bubbles is changed. It is preferable that the module is machine-learned by using it as input data for learning.

In addition to the mixed-phase flow image, the evaluation module learns at least one numerical information of the moving speed of bubbles in the gas-liquid multiphase flow, the size of bubbles in the gas-liquid multiphase flow, and the density of the bubbles. It is preferable that the input data is a machine-learned module.

In addition to the multiphase flow image, the evaluation module performs machine learning using at least one numerical information of the superficial velocity of the gas in the pipe and the superficial velocity of the liquid in the pipe as input data for learning. It is preferably a module.

Another aspect of the present invention is a gas production system equipped with the flow mode evaluation device, which decomposes a gas hydrate in the ground to produce gas. The gas production system
A long tube having a tip configured to be buried in the ground, a gas hide outside the tube using the pressure generated by the liquid in the tube extending upward from the tip. By reducing the pressure acting on the rate, a gas-liquid mixture containing bubbles generated by decomposition from the gas hydrate is provided at the tip of the pipe and outside the pipe so as to be taken into the liquid in the pipe. A riser tube with an open hole and
A gas-liquid separator provided in the riser tube to separate the bubbles from the liquid in the riser tube that has taken in the gas-liquid mixture.
A gas generation line including a gas generation pipe for taking out the gas generated from the bubbles as a gas to be produced from the riser pipe, and a gas generation line.
A pump that sucks up the liquid in order to discharge the liquid from which the gas has been separated by the gas-liquid separator from the riser pipe.
A liquid discharge line including a liquid discharge pipe that takes out the liquid sucked up by the pump from the riser pipe, and
The conduit serving as the path of the liquid leading from the gas-liquid separation device to the pump is the pipe in the flow mode evaluation device, and the bubbles remaining in the liquid and the liquid are provided in the pipe as a gas-liquid mixed phase flow. The image acquisition unit that acquires an image of the gas-liquid multiphase flow when the gas-liquid multiphase flow passes through the observation position by imaging the gas-liquid multiphase flow that passes through the observed position, and the flow mode. The machine learning unit of the evaluation device and the evaluation unit of the flow mode evaluation device are provided so that the liquid level of the liquid is formed in the gas-liquid separation device and the decomposition of the gas hydrate can be controlled. In addition, the first control that controls the rotation of the pump according to the pressure at the tip portion, and the rotation speed of the pump that is controlled by the first control based on the evaluation result of the flow mode by the evaluation unit. A control device for performing a second control for further controlling the above.

Yet another aspect of the present invention is a flow mode evaluation method for evaluating a flow mode in a gas-liquid mixed phase flow of a liquid and a gas flowing in a pipe. The flow style evaluation method is
A step of acquiring an image of the gas-liquid multiphase flow when the gas-liquid multiphase flow passes through the observation position by imaging a gas-liquid multiphase flow passing through a predetermined observation position in the pipe.
A plurality of mixed-phase flow images obtained by changing the generation conditions of the gas-liquid mixed-phase flow of the liquid and the gas are used as learning input data, and the classification results of the flow modes of the plurality of mixed-phase flow images are used as learning output data. A step of preparing an evaluation module in which the relationship between the multiphase flow image and the classification result of the flow mode is machine-learned, and
Using the acquired image of the gas-liquid multiphase flow as an evaluation image, the evaluation module is provided with a step of evaluating the flow mode of the gas-liquid multiphase flow when passing through the observation position.

In the step of acquiring an image of the gas-liquid multiphase flow, the gas-liquid multiphase flow passing through the observation position is sequentially imaged in time series at a predetermined frame rate, and a plurality of the gas-liquid multiphase flows at the observation position are sequentially imaged. It is preferable to include a process of creating one still composite image of the gas-liquid multiphase flow as it passes through the observation position by stitching the images in chronological order.

Each of the mixed-phase flow images used as the learning input data sequentially images the gas-liquid multiphase flow passing through a region having the same size as the observation position in time series at the frame rate, and the gas in the region. It is preferable that the image is one static composite image of the gas-liquid multiphase flow created by connecting a plurality of images of the liquid-mixed multiphase flow in chronological order.

According to the above-mentioned flow mode evaluation device and flow mode evaluation method, it is possible to accurately evaluate the flow mode in the gas-liquid mixed phase flow of liquid and gas flowing in the pipe.

FIGS. (A) to (D) are diagrams for explaining an example of a flow mode of a gas-liquid mixed phase flow of a liquid and a gas flowing in a pipe. It is a figure explaining the evaluation module used in one Embodiment. It is a figure which shows an example of the structure of the flow style evaluation apparatus of one Embodiment. (A) is a diagram showing an example of an image and an observation position captured by the imaging unit of the flow mode evaluation device of one embodiment, and (b) is a still composite image obtained by the flow mode evaluation device of one embodiment. It is a figure which shows an example. It is a figure which shows an example of the structure of the gas production system of one Embodiment. It is a figure which shows an example of the structure of the gas production system of one Embodiment.

Hereinafter, the flow mode evaluation device, the flow mode evaluation method, and the gas production system of one embodiment will be described with reference to the drawings. 1 (a) to 1 (d) are diagrams for explaining an example of a flow mode of a gas-liquid mixed phase flow of a liquid and a gas flowing in a pipe.

As shown in FIGS. 1 (a) to 1 (d), the flow modes are a bubble flow form (flow mode 1 shown in FIG. 1 (a)) and a slug flow form (flow mode 3 shown in FIG. 1 (b)). , A form of churn flow (flow mode 5 shown in FIG. 1 (c)) and a form of circular flow (flow mode 7 shown in FIG. 1 (d)). The form of the bubble flow is a form in which the bubbles are composed of only fine bubbles and the fine bubbles are dispersed in a continuous liquid. The form of the slag flow is a form in which large bubbles (gas slag) that spread over the entire cross section of the flow path and a liquid portion (liquid slug) containing fine bubbles flow alternately along the flow path. The churn flow is a flow in which the gas slag is long and the interface is pulsating, and the liquid slag contains a large number of bubbles, and the boundary between the gas slag and the liquid slag is unclear. The form of the circular flow is a form in which a liquid film is present on the tube wall and droplets are scattered in a large number of gases at the center of the cross section of the flow path.

In addition to such flow modes 1, 3, 5, and 7, a flow mode 2 showing an intermediate form between the flow modes 1 and a flow mode 3, and an intermediate form between the flow modes 3 and the flow mode 5 are shown. The flow mode 4, and the flow mode 6, which shows an intermediate form between the flow mode 5 and the flow mode 7, are set as the items to be predicted by the evaluation module 1 described below.

FIG. 2 is a diagram illustrating an evaluation module 1 used in one embodiment.
The evaluation module 1 uses a plurality of mixed-phase flow images obtained by variously changing the generation conditions of a liquid-gas gas-liquid multiphase flow as input data for learning, and classifies the flow modes of the gas-liquid mixed-phase flow in the plurality of mixed-phase flow images. This module uses the results (fluid modes 1 to 7) as training output data and machine-learns the relationship between the multiphase flow image and the classification result of the flow mode. Such an evaluation module 1 is configured by a computer.
The evaluation module 1 that has been machine-learned using the learning input data and the learning output data is
The flow mode of the gas-liquid multiphase flow is evaluated using the image of the gas-liquid multiphase flow acquired by the image acquisition unit 18c (see FIG. 3) described later as an evaluation image.

Evaluation module 1 is a well-known random forest that "classifies" or "regresses" using a deep neural network represented by well-known deep learning, preferably a model using a convolutional neural network, and a plurality of decision trees. Includes models using the method and models using LASSO regression. Further, in the model of the evaluation module 1, a non-linear function using a polynomial, kriging, or RBF (Radial Base Function) can also be used.

FIG. 3 is a diagram showing an example of the configuration of the flow mode evaluation device 10 of one embodiment.
The flow mode evaluation device 10 is a device that evaluates the flow mode of the gas-liquid mixed phase flow of liquid and gas flowing in the pipe 11. The pipe 11 is provided with an observation window 11a and an illumination window 11b.
The flow mode evaluation device 10 includes a light source 12, an imaging unit 14, a computer 18, an input operation unit 20, and a monitor 22.
The light source 12 emits illumination light toward the illumination window 11b to illuminate the gas-liquid multiphase flow. For the plateau 12, for example, an LED light source is used.
The image pickup unit 14 takes an image of a gas-liquid multiphase flow using the illumination light as the backlight light. In the image, the liquid part becomes dark and the gas part such as bubbles becomes bright. Therefore, the gas part can be clearly distinguished from the liquid part.
The imaging unit 14 continuously images at a constant frame rate, for example, 24 [frames / second] to 70,000 [frames / second]. The image of the gas-liquid multiphase flow obtained by the imaging unit 14 is sent to the computer 18.

The computer 18 includes a CPU 18a and a memory 18b including a ROM and a RAM, and is connected to an input operation device 20 including a keyboard or a mouse and a monitor 22. The computer 18 reads the computer program from the memory 18b and starts the computer program when executing various processes. Computer programs perform various processes. For example, the computer 18 forms an image acquisition unit 18c, a machine learning unit 18d, and an evaluation unit 18e as units for executing various processes. That is, the image acquisition unit 18c, the machine learning unit 18d, and the evaluation unit 18e are software modules formed on the computer 18.

The image acquisition unit 18c is configured by the computer 18 and acquires an image of the gas-liquid multiphase flow when the gas-liquid multiphase flow passes through the observation position from the image captured by the imaging unit 14.
The machine learning unit 18d is composed of a computer 18, and uses a plurality of mixed-phase flow images obtained by variously changing the generation conditions of a liquid-gas gas-liquid multiphase flow as learning input data, and the gas-liquid in the plurality of mixed-phase flow images. The evaluation module 1 (see FIG. 2) in which the relationship between the multiphase flow image and the classification result of the flow mode is machine-learned is included as the output data for learning the classification result of the flow mode of the multiphase flow.
The evaluation unit 18e is composed of a computer 18, and the flow of the gas-liquid multiphase flow when passing through the observation position using the evaluation module 1 using the image of the gas-liquid multiphase flow acquired by the image acquisition unit 18c as an evaluation image. Evaluate the style. The evaluation of the flow mode means to determine which of the fluid modes 1 to 7 the flow mode of the gas-liquid mixed flow to be evaluated is.

The input operation device 20 is used to set various conditions to be set in the processing performed by the image acquisition unit 18c, the machine learning unit 18d, and the evaluation unit 18e.
The monitor 22 displays a condition setting screen for setting the above conditions and a screen of the processing result when each processing is performed.

In one embodiment, the observation position of the gas-liquid multiphase flow imaged by the image pickup unit 14 is a part within the visual field range imaged through the observation window 11a.
FIG. 4A is a diagram showing an example of an image and an observation position imaged by the image pickup unit 14 of the flow mode evaluation device 10 of one embodiment. As shown in FIG. 4A, the captured image includes an image of the gas-liquid multiphase flow around the observation window 11a and through the observation window 11a. Here, the observation position is a part of the area of the observation window 11a. In FIG. 4A, the observation position W is shown in a rectangular area. It is preferable that a part of the image of the gas-liquid multiphase flow obtained through the observation window 11a, that is, the image of the gas-liquid multiphase flow passing through the image at the observation position W is used as the evaluation image. The gas-liquid multiphase flow is often a flow in an unsteady state in which the flow mode changes depending on the location of the pipe 11 in the longitudinal direction of the pipe, and the flow mode tends to change depending on the location. In particular, when the pipe 11 is extended in the height direction to form an ascending gas-liquid multiphase flow in the pipe 11, the flow mode differs greatly depending on the location in the height direction of the pipe due to the influence of gravity. .. Therefore, it is preferable to evaluate the flow mode at the observation position W without making a mistake. For example, in a pump that sucks in liquid and sends it upward, the sending characteristics of the pump change according to the flow mode immediately before the suction port of the pump. Therefore, in order to control the sending characteristics of the pump to appropriate characteristics, It is preferable to evaluate the flow mode without error and control the drive of the pump based on the evaluation result. From this point, in the present embodiment, an image of the gas-liquid multiphase flow when the gas-liquid multiphase flow passes through the observation position W is acquired from the image captured by the imaging unit 14. The length F in the longitudinal direction of the tube at the observation position W is preferably, for example, a ratio F / Q with the diameter Q of the tube is 5.71 × ^{10-5 to} 20, preferably 0.005 to 1.2. Is more preferable.
As a result, the evaluation unit 18e uses the acquired image of the gas-liquid multiphase flow as an evaluation image, and uses the evaluation module 1 of the machine learning unit 18d to determine the flow mode of the gas-liquid multiphase flow when passing through the observation position W. Can be evaluated.

Further, considering the case where the flow mode of the gas-liquid multiphase flow changes depending on the location of the pipe 10 in the pipe longitudinal direction, it is preferable that the range of the observation position W in the pipe longitudinal direction is narrow. However, as shown in FIGS. 1 (b) and 1 (c), gas slag and liquid slag containing fine bubbles flow alternately in the flow mode, so a narrow observation position W is used for the slag flow or churn flow. Therefore, it is difficult to evaluate without error, and it is preferable to obtain an image by increasing the range of the observation position W.
That is, if the range of the observation position W is widened, the flow mode differs depending on the location of the pipe 11, so that it is easy to evaluate the erroneous flow mode. On the other hand, if the range of the observation position W is narrowed, the slag flow or the churn flow is erroneous. It is difficult to evaluate without.
Therefore, according to one embodiment, the image acquisition unit 18c images the gas-liquid multiphase flow passing through the observation position W in time series at a predetermined frame rate, and observes the gas-liquid multiphase flow obtained by this imaging. It is preferable to create one still composite image of the gas-liquid multiphase flow when passing through the observation position W as an evaluation image by connecting a plurality of images at the position W in the order of time series.

FIG. 4B is a diagram showing an example of a still composite image I obtained by the flow mode evaluation device 10 of one embodiment. The static composite image I is an image of a gas-liquid multiphase flow when passing through the observation position W. This image shows a plurality of images at the observation position W of the gas-liquid multiphase flow obtained when the gas-liquid multiphase flow passing through the observation position W is continuously imaged in time series by the imaging unit 18 at a predetermined frame rate. , It was created by connecting them in chronological order.
Therefore, the still composite image I is often different from one image of the gas-liquid multiphase flow obtained by expanding the observation window 11a and performing one imaging.

According to one embodiment, each of the multiphase flow images used as the input data for learning has the same size as the observation position W, as in the still composite image I, so that the error in the evaluation of the flow mode in the evaluation module 1 is reduced. Created by imaging the gas-liquid multiphase flow passing through the observation region in time series at the same frame rate as the above frame rate, and connecting multiple images of the gas-liquid multiphase flow in the observation region in the order of time series. It is preferable that it is one still composite image of the gas-liquid multiphase flow.

The mixed-phase flow image used as such learning input data is created by supplying a set amount of liquid and gas from the lower end of a pipe having the same diameter as the pipe 11. Since the flow mode of the gas-liquid multiphase flow changes variously depending on the height from the lower end of the pipe, the gas-liquid multiphase flow is imaged at a plurality of locations different in the longitudinal direction of the pipe, and the air in the observation region. It is preferable to create one still composite image by connecting a plurality of images of the liquid multiphase flow in the order of time series. As a result, an appropriate number of static composite images of gas-liquid mixed-phase flow of various flow modes can be acquired, and the evaluation accuracy of the evaluation module 1 can be improved.

According to one embodiment, the superficial velocity of gas in the pipe 11 is Vg [m / sec], the frame rate of imaging in the imaging unit 14 is R [frame / sec], and the line is along the pipe 11 at the observation position W. when the length was L [m], Vg / ( R · L) is 1.42 × ^{10 -3} or more 1.05 × 10 ⁴ or less, it is preferable. More preferably, Vg / (RL) is 0.003 or more and 36 or less. It is preferable that the images in which the gas-liquid multiphase flows are connected are included in the still composite image without interruption. The superficial velocity is a value obtained by dividing the volumetric flow rate of gas by the cross-sectional area of the pipe 11, and can be said to be the average flow velocity of the gas flowing through the pipe 11. The velocity and the moving velocity of the bubble passing through the observation position W are not the same. Therefore, Vg / (R · L), by a 1.42 × ^{10 -3} or more 1.05 × 10 ⁴ or less, it is possible to prevent the image to Awa connecting the gas-liquid multiphase flow is interrupted. Even if the same part of the image to be connected (for example, the part of the same bubble) partially overlaps, it does not significantly affect the evaluation result of the flow mode, but in order to prevent the same part in the image from overlapping, By image processing the images to be joined, identifying the same part in this image, and removing this same part from the image, the images of gas-liquid multiphase flow are not interrupted and do not overlap. You may create a still composite image in.

As shown in FIGS. 3 and 4, a plurality of images at the observation position W of the gas-liquid multiphase flow obtained when images are continuously imaged in time series by the image pickup unit 18 at a predetermined frame rate are connected in chronological order. According to the above-described embodiment in which the still composite image I created by combining is used as the evaluation image, the image of the gas-liquid multiphase flow acquired by the image acquisition unit 18 can be used as the image of the ascending flow rising in the pipe 11. it can. Since the image of the ascending flow often shows a different flow mode depending on the location of the pipe 11 due to the influence of gravity, by connecting the images of the gas-liquid multiphase flow passing through the observation position W, the image of the gas-liquid multiphase flow is connected at the observation position W. The flow mode can be evaluated appropriately.

The evaluation module 1 uses input data for learning a plurality of images in which at least one of the moving speed of bubbles in the gas-liquid multiphase flow, the size of bubbles in the gas-liquid multiphase flow, and the density of bubbles is changed. It is preferable that the module is machine-learned by using the above. The gas-liquid multiphase flow used as an evaluation image often differs in the size and density of bubbles and the moving speed of bubbles, and in order to obtain a stable evaluation, the moving speed of bubbles and the moving speed of the gas-liquid mixed flow are included. It is preferable to let the evaluation module 1 perform machine learning using a plurality of images in which at least one of the bubble size and the bubble density is changed.

According to one embodiment, the evaluation module 1 has at least one numerical information of the moving speed of bubbles in the gas-liquid multiphase flow, the size of the bubbles in the gas-liquid multiphase flow, and the density of the bubbles, in addition to the mixed-phase flow image. Is preferably a machine-learned module using the above as input data for learning. In such an evaluation module 1, in addition to the evaluation image, at least one numerical information of the moving speed of the bubbles in the gas-liquid multiphase flow, the size of the bubbles in the gas-liquid multiphase flow, and the density of the bubbles is input. Therefore, the flow mode can be evaluated more accurately.

Further, according to another embodiment, the evaluation module 1 uses at least one numerical information of the superficial velocity in the gas pipe and the superficial velocity in the liquid pipe as input data for learning in addition to the multiphase flow image. It is also preferable that the module is machine-learned. In such an evaluation module 1, in addition to the evaluation image, at least one numerical information of the superficial velocity in the gas pipe and the superficial velocity in the liquid pipe is input to more accurately evaluate the flow mode. It can be carried out.

According to one embodiment, the flow mode evaluation device 10 can be applied to a gas production system that decomposes a gas hydrate in the ground to produce gas.
In the gas production system, the gas produced by decomposing the gas hydrate in the ground by the decompression method is taken out to the ground using a riser pipe. Gas hydrate is often buried in the ground on the seabed, and when gas is taken out, water and seawater are also taken out at the same time. In particular, in the decompression method, the lower part of the riser pipe is decompressed and the gas hydrate in the lower part is decomposed by assembling water or seawater on the ground by a pump. When the gas obtained by decomposing the gas hydrate is taken out, water or seawater and gas are in a gas-liquid multiphase flow, so the gas and water or seawater are separated by using a gas-liquid separator. However, if this separation is not sufficient, when the pump tries to suck up water or seawater, the gas will be mixed into the water or seawater and sucked into the pump. For this reason, the drainage characteristics of the pump become unstable, and it may not be possible to stably control the decomposition of gas hydrate by the decompression method. Therefore, in order to stably control the delivery characteristics of the pump, it is necessary to appropriately evaluate the form of the gas mixed with the water sucked up by the pump and the seawater, that is, the flow mode of the gas-liquid multiphase flow.

5 and 6 are diagrams showing an example of the configuration of the gas production system (hereinafter, simply referred to as a system) 30 of one embodiment. Hereinafter, natural gas hydrate will be described as an example of gas hydrate.
The system 30 is a system that decomposes natural gas hydrate in the ground to generate natural gas and takes it out to the ground in a riser pipe 40 extending from the drillship 33 on the sea to the ground via the seabed. ..
The system 1 mainly includes a riser pipe 40, a gas-liquid separation device 50, a pump 53, a gas generation line 42, a liquid discharge line 43, and a control device 70.

The riser pipe 40 is a long pipe having a tip portion 40a configured to be buried in the ground. In the example shown in FIG. 5, the riser pipe 40 extends vertically downward from the drillship 33, and the tip portion 40a is buried in the well 37 on the seabed. The well 37 is a hole provided by excavation, and in the example shown in FIG. 5, it penetrates the upper layer 34 including the seabed 32 and is closed in the hydrate layer 35 located in the lower layer. The upper layer 34 is, for example, a mud layer containing a large amount of mud. The hydrate layer 35 is, for example, a layer called a sand-mud alternating layer containing a large amount of mud and sand. The hydrate layer 35 has a sandy layer spread in the lateral direction in which natural gas hydrate is incorporated into sand or mud. The boundary between the upper layer 34 and the hydrate layer 35 is, for example, at a position several hundred meters below the seabed, and the seabed 32 is, for example, at a depth of 300 meters to a thousand and several hundred meters.

The riser tube 40 includes a tube body 41 and a screen 49 (see FIG. 6).
A gas-liquid separation device 50, a pump 53, a gas generation line 42, a liquid discharge line 43, and a heater 56 are provided in the pipe body 41.

The pipe body 41 is a member that isolates the inner space from water or seawater, except for the hole 48a of the portion 48 that functions as a pickup pipe, which will be described later. In the example shown in FIG. 5, the pipe body 41 is provided with partition walls 47a and 47b and partition walls 47c that partition the inner space vertically. The portion of the pipe body 41 extending from the partition wall 47c to the tip of the riser tube 40 is a portion 48 in which the gas-liquid mixture taken into the liquid from the hydrate layer 35 flows upward together with the liquid (hereinafter, this portion is collected. In the example shown in FIG. 5, the pipe diameter is smaller than that of the pipe body 41 above the partition wall 47c. The pickup pipe portion 48 is located in the hydrate layer 35.

The screen 49 (see FIG. 6) is provided in the lift pipe portion 48 so as to cover the hole 48a opened to the outside of the riser pipe 40. The hole 48a is provided at the tip of the riser tube 40 extending to the deepest position in the ground. The hole 48a is provided in the lift pipe portion 48 at a depth position in contact with the sandy layer in the hydrate layer 35.
The screen 49 provided in the hole 48a is a member that takes in air bubbles and water generated by decomposition of natural gas hydrate, as well as seawater, and separates and removes sand and mud. The screen 49 has a function of allowing air bubbles, water, and seawater to pass through, but not sand and mud. The screen 49 is, for example, a sheet-like or plate-like structure having a large number of holes, and is composed of a plurality of structures having different hole sizes and shapes. Specific examples of combinations of a plurality of structures include Johnson screens, meshes, and gratings. The Johnson screen is well known as a wire mesh structure manufactured by Johnson Screen. Grating is a member in which steel materials are assembled in a grid pattern. The Johnson screen, mesh, and grating are arranged on the lift pipe portion 48 in this order from the side of the lift pipe portion 48 toward the side of the hydrate 35 layer.
That is, the riser tube 40 includes a hole 48a provided at the tip portion 40a and opened to the outside of the tube. The hole 48a is decomposed from the gas hydrate by reducing the pressure acting on the gas hydrate outside the riser tube 40 by using the pressure generated by the liquid in the riser tube 40 extending upward from the tip portion 40a. It functions to take the gas-liquid mixture containing the generated bubbles into the liquid in the riser tube 40.

As shown in FIG. 6, the lift pipe portion 48 is provided with a plurality of holes 48a along the depth direction for taking in the gas-liquid mixture that has passed through the screen 49. The hole 48a penetrates the wall portion of the lift pipe portion 48 and opens to the outside of the lift pipe portion 48. By providing the riser pipe 40 with the hole 48a, the pressure acting on the natural gas hydrate using the bottom pressure is reduced, whereby the gas-liquid mixture can be taken into the riser pipe 40.
The bottom pressure is the sum of the pressure generated at the tip 40a in the riser pipe 40 by the liquid in the riser pipe 40 extending upward from the tip 40a of the riser pipe 40 and the pressure in the gas phase space G described later, which will be described later. It is a pressure substantially determined by the head pressure received by the lower end of the riser pipe 40 by the liquid below the liquid level S. The lower end of the riser pipe 40 is located at substantially the same height (depth from the sea surface) as the hole bottom (pit bottom) of the well 37. The tip portion 40a is located at substantially the same height as the lower end (depth from the sea surface), and a hole 48a is provided in the tip portion 40a.
In the liquid in the riser tube 40, a gas-liquid mixture produced by decomposition from natural gas hydrate is taken in, and water and seawater that have entered through the hole 48a are taken in. Since the gas-liquid mixture contains bubbles, the liquid in the riser tube 40 contains bubbles. Water and seawater originate from water and seawater contained in the hydrate layer 35, and water and seawater contained in other strata in contact with the hydrate layer 5.

The riser pipe 40 further includes a pressure gauge 61 for measuring the bottom pressure, which is provided at the tip portion 40a of the lift pipe portion 48, specifically, at the lower end of the riser pipe 40. The pressure gauge 61 is connected to the control device 70 (see FIG. 5), and outputs a measurement signal of the bottom pressure toward the control device 70.

As shown in FIG. 6, the gas-liquid separation device 50, the pump 53, and the heater 56 are provided in the space 45b of the riser pipe 40 partitioned by the partition walls 47b and 47c. In the space 45b, in the example shown in FIG. 6, a gas phase space G into which the gas separated from the liquid by the gas-liquid separation device 50 flows is formed above the liquid surface S of the liquid. The gas phase space G is preferably formed in the riser tube 40 so as to be located above the seabed 32.

The gas-liquid separation device 50 is a device that separates at least a part of bubbles in the gas-liquid mixture taken into the liquid in the lift pipe portion 48. The gas in the separated bubbles is the gas produced. According to one embodiment, the gas-liquid separation device 50 includes a surrounding container 51 and a centrifuge 52.

The surrounding container 51 is a cup-shaped member that surrounds the lower end of the liquid transport pipe 44 (described later) constituting the liquid discharge line 43 from the outside. A centrifuge 52 and a pump 53 are provided in the liquid transport pipe 44. In the example shown in FIG. 6, the surrounding container 51 has a tubular side wall 51a arranged with a gap from the inner wall of the pipe body 41, and a bottom wall 51b that closes the lower end of the side wall 51a. The upper end of the side wall 51a is located above the lower end of the liquid transport pipe 44. As a result, the liquid containing the gas-liquid mixture rises in the gap between the pipe body 41 and the side wall 51a along the thin arrow shown in FIG. 6, and then the liquid descends in the gap between the side wall 51a and the liquid transport pipe 44. Then, it flows into the liquid transport pipe 44. In this process, since the relatively large bubbles have a high floating speed, they separate from the liquid flow and float on the liquid surface S by the time the liquid flows into the liquid transport pipe 44, and the gas is released into the gas phase space G. Will be done. In FIG. 6, the flow of bubbles is indicated by thick arrows.
That is, the gas-liquid separation device 50 is connected to the ascending path in which the liquid flow path faces upward and a part of the bubbles from the liquid, and changes the liquid flow path from the upper side to the lower side. It has a descending path. Such a gas-liquid separation method is called a gravity separation method because it separates gas and liquid by using gravity (specific gravity).

The centrifuge 52 is a device that separates relatively small bubbles that still remain in the liquid flowing into the liquid transport pipe 44 from the liquid. In the example shown in FIG. 6, the centrifuge 52 has a rotating body 52a provided in the liquid transport pipe 44 and rotating around a rotation center line extending in the vertical direction. The rotating body 52a is driven by a motor 54 described later. When the liquid containing bubbles approaches the rotating body 52a, it flows along a swirling flow created by the rotation of the rotating body 52a. At this time, a centrifugal force acts on the bubble and the liquid, and since the liquid has a higher specific gravity than the bubble, the liquid moves away from the rotation center line, and the bubble is collected closer to the rotation center line than the liquid. At this time, the collected and enlarged bubbles are discharged from the holes provided in the liquid transport pipe 44 that open to the outside of the liquid transport pipe 44. As a result, it floats on the liquid surface S and is released into the gas phase space G. On the other hand, the minute bubbles that are not released from the holes of the liquid transport pipe 44 rise in the liquid transport pipe 44 together with the liquid. Such a method of separating using centrifugal force is called a centrifugal separation method.
As described above, the gas-liquid separation device 50 uses both the gravity separation method and the centrifugal separation method, but according to one embodiment, the gas-liquid separation can be performed only by the gravity separation method. Further, according to one embodiment, gas-liquid separation can be performed only by the centrifugal separation method.

The pump 53 draws the liquid into the liquid transport pipe 44 and discharges it from the riser pipe 40. That is, the pump 53 sucks up the liquid in order to discharge the liquid from which the gas has been separated by the gas-liquid separation device 50 from the riser pipe 40.
The pump 53 of the example shown in FIG. 6 is an auger pump arranged in a liquid transport pipe 44 and having a motor 54 and a screw 55 driven by the motor 54. The screw 55 has a shaft extending in the vertical direction and blades extending spirally around the shaft, and has a function of feeding the liquid in the liquid transport pipe 54 upward while pressurizing by the rotation of the screw 55. .. The motor 54 is electrically connected to the control device 70 of the drillship 33. The motor 54 receives the signal output from the control device 70 and is controlled to drive at a set frequency or an adjusted frequency. The motor 54 is arranged in the liquid transport pipe 44 so as to form a gap in the liquid transport pipe 44 that serves as a flow path for the liquid. Since seawater or water continues to flow into the riser pipe 40 through the hole 48a during the operation of the system 30, the pump 53 is normally maintained in an operating state.
An observation window 11a and an illumination window 11b are provided near the inlet of the screw 55 of the liquid transport pipe 44, and further, a light source 12 of the flow mode evaluation device 10 shown in FIG. 3 is located at the position of the observation window 11a and the illumination window 11b. And an image pickup unit 14 is provided. In the imaging unit 14, the liquid transport pipe (conduit) 44, which is the path of the liquid leading from the gas-liquid separation device 50 to the pump 53, is used as the pipe 11 in the flow mode evaluation device 10 shown in FIG. Taking a liquid and a gas-liquid mixed phase flow as a gas-liquid mixed phase flow, an image is taken of the gas-liquid mixed phase flow passing through an observation position provided in a liquid transport pipe (conduit) 44. The captured image is sent to the control device 70.

The heater 56 is a device that heats the liquid that has flowed into the space 45b. The heater 56 is connected to the control device 70. Since the decomposition reaction of natural gas hydrate is an endothermic reaction, the temperature of the gas-liquid mixture taken into the liquid drops and the natural gas hydrate is regenerated, for example, the lower end of the liquid transport pipe 44 may be blocked. is there. The heater 56 heats the liquid continuously or intermittently during the operation of the system 30 to suppress the regeneration of natural gas hydrate. Further, when it is determined that the natural gas hydrate has been regenerated, it is controlled to be driven by receiving a signal output from the control device 70, and the regenerated natural gas hydrate is heated by heating the liquid. And disassemble.

The gas generation line 42 includes a gas generation pipe 42a and a gas flow rate measuring device 42b.
The gas generation pipe 42a takes out the gas generated from the bubbles floating on the liquid surface S and flowing into the gas phase space G from the riser pipe 40 as the natural gas to be produced. The gas generation pipe 42a carries the gas in the gas phase space G toward the drillship 33 as natural gas for producing the gas. The gas generation pipe 42a is arranged in the pipe body 41 above the liquid level S, and the lower end of the gas generation pipe 42a is connected to the gas phase space G. The gas flow rate measuring device 42b is provided on a portion of the tip of the gas generation pipe 42a on the drillship 33, and measures the amount of natural gas produced per unit time. The measured information on the amount of natural gas produced is transmitted to the control device 70, which will be described later.
Further, the upper tip of the gas generation pipe 42a is connected to, for example, a storage tank (not shown) provided on the drillship 33 or another vessel, or a pipeline extending to a storage system on land. The natural gas stored in the storage tank is appropriately liquefied and transported over the sea by a drillship 3 or another vessel.

The liquid discharge line 43 includes a liquid discharge pipe 43a that carries the liquid separated from the natural gas in the pipe body 41 to the drillship 33, and a liquid flow rate measuring device 43b.
In the example shown in FIG. 5, the liquid discharge pipe 43a includes a liquid transport pipe 44 extending from the gas-liquid separation device 50 to the space 45a, and a pipe 46 branching from the pipe body 41 and extending from the space 45a to the drillship 33. Have. The liquid discharge pipe 43a is a pipe that takes out the liquid sucked up by the pump 53 from the riser pipe 40. The space 45a is a space partitioned by partition walls 47a and 47b.
The liquid flow rate measuring device 43b is provided on a portion on the drillship 33 at the upper end of the pipe 46, and measures the discharge amount of the liquid discharged by the pump 53 per unit time. The measured liquid discharge amount information is transmitted to the control device 70, which will be described later.
The discharged liquid is discarded at sea from, for example, the drillship 33.

The gas phase space G in the riser tube 40 is provided with a pressure gauge 60 for measuring the pressure in the gas phase space G. The pressure gauge 60 is connected to the control device 70, and information on the measurement result of the pressure in the gas phase space G is transmitted to the control device 70. In the example shown in FIG. 6, the pressure gauge 60 is provided on the wall surface surrounding the space 45a.
Further, the tip portion 42a of the riser pipe 40 is provided with a pressure gauge 61 for measuring the pressure (pit bottom pressure) at the tip portion 40a of the riser pipe 40. The pressure gauge 61 is connected to the control device 70, and information on the measurement result of the pressure (pit pressure) at the tip portion 40a is transmitted to the control device 70. The pressure at the tip portion 40a is determined by the position of the liquid level S of the liquid and the pressure of the gas phase space G, and is a pressure that affects the rate of decomposition of the natural gas hydrate in the hydrate layer 35.

The control device 70 is composed of a computer including a CPU, a memory, and the like. The control device 70 measures the rotation of the pump 53 with the pressure gauge 61 so that the liquid level S of the liquid is formed in the gas-liquid separation device 20 and the decomposition of the natural gas hydrate can be controlled, that is, the tip. The first control is controlled according to the information of the pressure measurement result in the unit 40a, and the first control is performed by using at least the evaluation result of the flow mode of the gas-liquid mixed phase flow evaluated by using at least the image captured by the imaging unit 14. The second control is configured to further control the rotation speed of the pump 53 controlled by the above. The control device 70 includes the computer 18 shown in FIG. That is, the control device 70 includes an image acquisition unit 18c, a machine learning unit 18d, and an evaluation unit 18e. The machine learning unit 18d includes an evaluation module 1 that has been machine-learned in advance using the above-mentioned learning input data and learning output data.

Here, the evaluation of the flow mode of the gas-liquid multiphase flow uses at least one numerical information of the superficial velocity of the natural gas in the liquid discharge pipe 43a and the superficial velocity of the liquid liquid discharge pipe 43a such as water or seawater. The numerical information of the gas production amount per unit time of the gas taken out from the gas generation line 42, the numerical information of the discharge amount of the liquid taken out from the liquid discharge line 43 per unit time, and the liquid level. Of the numerical information of the gas phase space pressure of the gas phase space G above S, at least two numerical information may be used for evaluation. The numerical information of the gas production amount, the numerical information of the liquid discharge amount, and the numerical information of the gas phase space pressure are the information transmitted from the gas flow rate measuring device 42b, the liquid flow rate measuring device 43b, and the pressure gauge 60, respectively. is there.

In the system 30, for example, the material to be the riser tube 40, the computer 18, the light source 12, the imaging unit 14, the pressure gauges 60 and 61, the control device 70, and the like are loaded on the drillship 3 and transported to a predetermined position on the sea. Assembled. The well 37 is pre-drilled before assembling the system 30.

The system 30 may include a fixed or floating offshore platform instead of the drillship 33. In this case, it is preferable to provide a pipeline that connects the offshore platform and the land and transports natural gas from the offshore platform to the land.

The control device 70 controls the pressure acting on the natural gas hydrate according to the information of the pressure measurement result at the tip portion 40a measured by the pressure gauge 61 to control the decomposition rate of the natural gas hydrate. Thereby, the amount of gas-liquid mixture produced is adjusted. As described above, normally, when the position of the liquid level S is high, the pressure at the tip 40a (bottom pressure) is high, and when the pressure at the tip 40a (bottom pressure) is high, it acts on the natural gas hydrate. The pressure to do is also high. Therefore, in order to control the pressure (pit pressure) at the tip portion 40a, the rotation speed of the pump 53 is controlled to control the position of the liquid level S of the liquid.

Specifically, the control device 70 holds information on the range of the target pressure with respect to the pressure (pit pressure) at the tip portion 40a. The range of the target pressure is the range of the pressure (bottom pressure) at the tip portion 40a corresponding to the pressure for decomposing the natural gas hydrate. If the pressure at the tip 40a (bottom pressure) is within the target pressure range, the rate of decomposition of methane hydrate can be controlled within a predetermined range.
The control device 70 determines at regular time intervals whether or not the information of the pressure measurement result at the tip portion 40a is within the range of the target pressure. In this determination, if the information of the pressure measurement result at the tip portion 40a is within the range of the target pressure, the rotation speed of the pump 53 is maintained. On the other hand, when the information of the pressure measurement result at the tip 40a is higher than the target pressure range, the rate of decomposition of the natural gas hydrate is not within the predetermined range, so that the pressure at the tip 40a is low. Control is performed to increase the rotation speed of the pump 53 so as to be. This can accelerate the rate of decomposition of natural gas hydrate.
Further, when the pressure measurement result at the tip portion 40a, that is, the information of the measurement result of the pressure gauge 61 is lower than the target pressure range, the decomposition speed of the natural gas hydrate is not within the predetermined range. Control is performed to reduce the rotation speed of the pump 53 so that the pressure at the tip portion 40a becomes high. As a result, the rate of decomposition of natural gas hydrate can be suppressed.
The above is the first control performed by the control device 70.

In such a first control, the rotation speed of the pump 53 is controlled to control the amount of liquid discharged by the pump 53, but for the liquid sucked up by the pump 53, bubbles are separated from the liquid by the gas-liquid separator 50. However, the bubbles cannot be sufficiently removed, and fine bubbles may be mixed in the liquid. In some cases, large bubbles may be mixed with the liquid and sucked into the pump 53 in the form of a slug flow (see FIG. 1B). In such a case, the amount of liquid discharged from the pump 53 tends to decrease due to the influence of fine bubbles and large bubbles (gas slag). Since the amount of fine bubbles mixed in the liquid sucked up by the pump 53 fluctuates with time, the discharge amount of the pump 53 also fluctuates with time, and the height of the liquid level S also fluctuates with time. As a result, the pressure at the tip portion 40a also fluctuates with time. Therefore, the rotation speed of the pump 53 is controlled by the first pressure control according to the time fluctuation of the pressure. Therefore, since the rotation speed of the pump 53 fluctuates with time, the decomposition of natural gas hydrate tends to fluctuate with time. As described above, the first control may be an unstable control.
Therefore, in the system 30, in addition to the first control, the second control is performed to control the form of the gas mixed in the liquid sucked into the pump 53.
That is, the control device 70 is provided with the computer 18, and the rotation of the pump 53 is set to the pressure at the tip portion 40a so that the liquid level of the liquid is formed in the gas-liquid separation device 50 and the decomposition of the gas hydrate can be controlled. Based on the first control that is controlled accordingly and the evaluation result of the flow mode by the evaluation unit 18e, the second control that further controls the rotation speed of the pump 53 controlled by the first control is performed.

The control device 70 evaluates the flow mode, which is the form of the gas contained in the liquid sucked into the pump 53, at the inlet of the pump 53 in the riser pipe 40 shown in FIG. 6, and the evaluation result is the stability of the pump 53. If the form of the gas is unfavorable for securing the discharged amount, the gas mixed in the liquid sucked into the pump 53 can be obtained by increasing or decreasing the rotation speed of the pump 53 according to the form of the gas. Control the flow mode, which is a form. This form of control is the second control.
For the evaluation of such a flow mode, the control device 70 of the system 30 includes a flow mode evaluation device 10 shown in FIG. 3, which functions as a flow mode evaluation device. That is, the control device 70 includes an image acquisition unit 18c, a machine learning unit 18d, and an evaluation unit 18e.

In such a flow style evaluation device 10, the following flow style evaluation is performed.
(1) Air and liquid passing through a predetermined observation position on the downstream side of the lift pipe portion 48 in the pipe 11, for example, an observation position provided in the intermediate portion 72 that moves from the lift pipe portion 48 to the gas and liquid separation device 50. By imaging the mixed-phase flow with the imaging unit 14, an image of the gas-liquid mixed-phase flow when the gas-liquid mixed-phase flow passes through the observation position W is acquired.
(2) A plurality of mixed-phase flow images obtained by variously changing the conditions for generating a gas-liquid mixed-phase flow of liquid and gas are used as learning input data, and the classification results of the flow modes of the plurality of mixed-phase flow images are used as learning output data. Prepare an evaluation module 1 that machine-learns the relationship between the multiphase flow image and the classification result of the flow mode.
(3) Using the acquired image of the gas-liquid multiphase flow as an evaluation image, the evaluation module 1 is used to evaluate the flow mode of the gas-liquid multiphase flow when passing through the observation position W.

At this time, according to one embodiment, when an image of the gas-liquid multiphase flow is acquired, the gas-liquid multiphase flow passing through the observation position W is imaged in time series at a predetermined frame rate, and the air at the observation position W is captured. It is preferable to include a process of creating one still composite image of the gas-liquid multiphase flow when passing through the observation position W by connecting a plurality of images of the liquid mixed phase flow in the order of time series. Since the flow mode of the gas-liquid multiphase flow differs depending on the location of the pipe 11, it is not preferable to set the observation position W in a wide range in the longitudinal direction of the pipe 11 from the viewpoint of highly accurate evaluation. On the other hand, it is not preferable to evaluate the flow mode of the gas-liquid multiphase flow with the observation position W in a narrow range in the longitudinal direction of the pipe 11 in terms of evaluating the slag flow as shown in FIG. 1 (b). Therefore, by sequentially capturing a plurality of images of the gas-liquid multiphase flow passing through the observation position W, which is a narrow range, as a moving image and connecting them in the order of time series, the gas-liquid multiphase flow when passing through the observation position W is performed. It is preferable to create one still composite image of the flow and use this image as an evaluation image.
Each of the multiphase flow images used as learning input data for machine learning in the evaluation module 1 also sequentially captures the gas-liquid multiphase flow passing through a region of the same size as the observation position W as a moving image at the same frame rate. For the above reasons, it is preferable that the image is one static composite image of the gas-liquid multiphase flow created by connecting a plurality of images of the gas-liquid multiphase flow in this region in chronological order.

In order to confirm the effect of this embodiment, an experiment was conducted using the laboratory test of the flow mode evaluation device shown in FIG.
The height of the pipe 11 was 29 m, and bubbles and salt water having a specific gravity of 1.04 were supplied from the lower end of the pipe 11 at a controlled supply amount and supply speed.
Using the images obtained at the observation positions W at five positions in the height direction of the pipe 11 (each position of 0.6 m, 3.0 m, 11 m, 17 m, and 25.3 m from the lower end), the input data for learning 4460 mixed-phase flow images were prepared. This image showed various flow modes as shown in FIG. 1 because the supply rate and the supply amount of the bubbles and the liquid were variously changed. As for the flow style, in addition to the styles 1, 3, 5, and 7, the flow style between the styles 1 and 3 is set as the style 2, and the flow style between the styles 2 and 5 is set as the flow style. Form 4 was used, and the flow mode intermediate between Form 5 and Form 7 was designated as Form 6, and was classified into a total of 7 flow modes at the discretion of the evaluator. This classification result was used as learning output data.
The learning input data and the learning output data were machine-learned by a model using a convolutional neural network (manufactured by Google). The number of intermediate layers was set to 9 as a parameter of the model of the convolutional neural network for machine learning.
At a later date, 100 images of the gas-liquid mixed phase flow when passing through the observation position W of 0.8 m in length are created and evaluated from the images obtained at the observation position W at an arbitrary position in the height direction of the pipe 11. It was used as an image. When this evaluation image was evaluated by the evaluation module 1, the degree of agreement between the evaluation result of the flow mode and the judgment result of the flow mode by the evaluator was 85%.

On the other hand, as the evaluation image, not the image of the gas-liquid multiphase flow when passing through the observation position W, but the captured image in the range of 0.03 m corresponding to the height centered on the observation position W was used. In this case, the degree of agreement of the flow patterns in the machine-learned prediction module 1 was 3%.

From this, the evaluation accuracy of the flow mode can be improved by evaluating with the prediction module 1 using the image of the gas-liquid multiphase flow when the gas-liquid multiphase flow passes through the observation position W from the image captured by the imaging unit 14. You can see that it improves.

The flow mode evaluation device for the gas-liquid mixed phase flow of the present invention and the flow mode evaluation method have been described in detail above, but the present invention is not limited to the above-described embodiment, and various types are described within the scope of the gist of the present invention. Of course, it may be improved or changed.

1 Evaluation module 10 Flow mode evaluation device 11 Piping 11a Observation window 11b Lighting window 12 Light source 14 Imaging unit 18 Computer 18a CPU
18b Memory 18c Image acquisition unit 18d Machine learning unit 18e Evaluation unit 20 Input operation device 22 Monitor 30 Gas production system 32 Submarine 33 Excavator 34 Upper layer 35 Hydrate layer 37 Well 40 Riser pipe 40a Tip 41 Pipe body 42 Gas generation Line 43 Liquid discharge line 43a Liquid discharge pipe 44 Liquid transport pipe 45a Space 46 Pipe 47a, 47b, 47c Partition 48 Lifting pipe part 50 Gas-liquid separator 51 Enclosing container 51a Side wall 51b Bottom wall 52 Centrifugal separator 53 Pump 54 Motor 55 Screw 56 Heater 60, 61 Pressure gauge 70 Control device 72 Intermediate part

One aspect of the present invention is a flow mode evaluation device that evaluates the flow mode of a gas-liquid mixed phase flow of a liquid and a gas flowing in a pipe. The flow mode evaluation device is
An imaging unit that captures a gas-liquid multiphase flow passing through a predetermined observation position provided in a pipe in a time series at a predetermined frame rate at the observation position.
An image composed of a computer and connecting the images obtained by the imaging unit in chronological order to acquire one still composite image of the gas-liquid multiphase flow when passing through the observation position at the observation position. Acquisition unit and
A plurality of mixed-phase flow images formed by a computer and obtained by variously changing the conditions for generating the gas-liquid multiphase flow of the liquid and the gas are used as learning input data, and the flow of the gas-liquid multiphase flow in the plurality of mixed-phase flow images. A machine learning unit including an evaluation module in which the relationship between the plurality of multiphase flow images and the classification result of the flow mode is machine-learned using the format classification result as learning output data.
When one static composite image of the gas-liquid multiphase flow acquired by the computer and acquired by the image acquisition unit at the observation position is used as an evaluation image and the evaluation module is used to pass through the observation position. An evaluation unit for evaluating the flow mode of the gas-liquid multiphase flow from the static composite image is provided.
The machine learning unit transmits a gas-liquid multiphase flow that passes through an observation region having the same size as the observation position, which is provided at a plurality of locations along the flow of the gas-liquid multiphase flow of the liquid and the gas. A plurality of static composite images in the plurality of observation regions created by capturing images in time series at a rate and connecting a plurality of images of the gas-liquid multiphase flow in the observation region in the order of the time series are obtained by the plurality of mixed phases. Used as a flow image.

Another aspect of the present invention is a gas that produces gas by decomposing a gas hydrate in the ground provided with a flow mode evaluation device for evaluating the flow mode of a gas-liquid mixed phase flow of a liquid and a gas flowing in a pipe. It is a production system. The gas production system
The above-mentioned when the gas-liquid multiphase flow passes through the observation position from an image captured by the imaging unit, which is composed of an imaging unit for imaging a gas-liquid multiphase flow passing through a predetermined observation position of a pipe and a computer. A plurality of mixed-phase flow images obtained by variously changing the conditions for generating the gas-liquid multiphase flow of the liquid and the gas, which are composed of an image acquisition unit for acquiring an image of the gas-liquid multiphase flow and a computer, are used as input data for learning. , A machine learning unit including an evaluation module in which the classification result of the flow mode of the gas-liquid multiphase flow in the plurality of multiphase flow images is used as learning output data, and the relationship between the mixed phase flow image and the classification result of the flow mode is machine-learned. And the image of the gas-liquid multiphase flow acquired by the image acquisition unit, which is configured by the computer, is used as an evaluation image, and the evaluation module is used to pass the gas-liquid multiphase flow when passing through the observation position. A flow mode evaluation device including an evaluation unit that evaluates the flow mode from the evaluation image, and
A long tube having a tip configured to be buried in the ground, a gas hide outside the tube using the pressure generated by the liquid in the tube extending upward from the tip. By reducing the pressure acting on the rate, a gas-liquid mixture containing bubbles generated by decomposition from the gas hydrate is provided at the tip of the pipe and outside the pipe so as to be taken into the liquid in the pipe. A riser tube with an open hole and
A gas-liquid separator provided in the riser tube to separate the bubbles from the liquid in the riser tube that has taken in the gas-liquid mixture.
A gas generation line including a gas generation pipe for taking out the gas generated from the bubbles as a gas to be produced from the riser pipe, and a gas generation line.
A pump that sucks up the liquid in order to discharge the liquid from which the gas has been separated by the gas-liquid separator from the riser pipe.
A liquid discharge line including a liquid discharge pipe that takes out the liquid sucked up by the pump from the riser pipe , and
The conduit serving as the path of the liquid leading from the gas-liquid separation device to the pump is the pipe in the flow mode evaluation device, and the bubbles remaining in the liquid and the liquid are provided in the pipe as a gas-liquid mixed phase flow. The image acquisition unit that acquires an image of the gas-liquid multiphase flow when the gas-liquid multiphase flow passes through the observation position by imaging the gas-liquid multiphase flow that passes through the observed position, and the flow mode. The machine learning unit of the evaluation device and the evaluation unit of the flow mode evaluation device are provided so that the liquid level of the liquid is formed in the gas-liquid separation device and the decomposition of the gas hydrate can be controlled. In addition, the first control that controls the rotation of the pump according to the pressure at the tip portion, and the rotation speed of the pump that is controlled by the first control based on the evaluation result of the flow mode by the evaluation unit. A control device for performing a second control for further controlling the above.

Yet another aspect of the present invention is a flow mode evaluation method for evaluating a flow mode in a gas-liquid mixed phase flow of a liquid and a gas flowing in a pipe. The flow style evaluation method is
By imaging the gas-liquid multiphase flow passing through the predetermined observation position provided in the pipe in time series at the observation position at a predetermined frame rate, the gas-liquid mixed phase flow when the gas-liquid multiphase flow passes through the observation position is taken. A step of acquiring one still composite image of the gas-liquid multiphase flow when passing through the observation position, in which images of the liquid-mixed multiphase flow are connected in chronological order, and
A plurality of mixed-phase flow images obtained by changing the generation conditions of the gas-liquid multiphase flow are used as learning input data, and the classification results of the flow modes of the plurality of mixed-phase flow images are used as learning output data with the mixed-phase flow image. A step of preparing an evaluation module in which the relationship with the classification result of the flow mode is machine-learned, and
Using the evaluation module as an evaluation image of the acquired still composite image of the gas-liquid multiphase flow, a step of evaluating the flow mode of the gas-liquid multiphase flow when passing through the observation position. Be prepared.
The plurality of mixed-phase flow images used as the learning input data pass through observation regions of the same size as the observation positions provided at a plurality of locations along the flow of the gas-liquid multiphase flow of the liquid and the gas. A plurality of images in the plurality of observation regions created by imaging the gas-liquid multiphase flow in time series at the frame rate and connecting a plurality of images of the gas-liquid multiphase flow in the observation region in the order of time series. It is a still composite image.

Claims (12)

A flow mode evaluation device that evaluates the flow mode of a gas-liquid mixed phase flow of liquid and gas flowing in a pipe.
An imaging unit that captures the gas-liquid multiphase flow passing through a predetermined observation position on the pipe,
An image acquisition unit configured by a computer and acquiring an image of the gas-liquid multiphase flow when the gas-liquid multiphase flow passes through the observation position from an image captured by the image pickup unit.
A plurality of mixed-phase flow images formed by a computer and obtained by variously changing the conditions for generating the gas-liquid mixed-phase flow of the liquid and the gas are used as learning input data, and the flow of the gas-liquid multiphase flow in the plurality of mixed-phase flow images. A machine learning unit including an evaluation module in which the relationship between the multiphase flow image and the classification result of the flow mode is machine-learned using the format classification result as learning output data.
The flow mode of the gas-liquid multiphase flow when passing through the observation position using the evaluation module using the image of the gas-liquid multiphase flow configured by the computer and acquired by the image acquisition unit as an evaluation image. With an evaluation unit that evaluates from the evaluation image,
A flow mode evaluation device characterized by comprising. The image acquisition unit captures a plurality of images of the gas-liquid multiphase flow passing through the observation position in time series at a predetermined frame rate, and images a plurality of images of the gas-liquid multiphase flow at the observation position in chronological order. The flow mode evaluation device according to claim 1, wherein one static composite image of the gas-liquid multiphase flow when passing through the observation position is created as the evaluation image by connecting the images. Each of the multiphase flow images used as the input data for learning captures the gas-liquid multiphase flow passing through the observation region having the same size as the observation position in time series at the frame rate, and the said in the observation region. The flow mode evaluation device according to claim 2, which is one static composite image of the gas-liquid multiphase flow created by connecting a plurality of images of the gas-liquid multiphase flow in chronological order. When the superficial velocity of the gas in the pipe is Vg [m / sec], the frame rate is R [frame / sec], and the length of the observation position along the pipe is L [m]. vg / (R · L) is, 1.42 × ^{10 -3} or more 1.05 × 10 ⁴ or less, flow regime evaluation apparatus according to claim 2 or 3. The flow mode evaluation device according to any one of claims 2 to 4, wherein the image of the gas-liquid multiphase flow acquired by the image acquisition unit is an image of an ascending flow rising in the pipe. The evaluation module displays a plurality of images in which at least one of the moving speed of bubbles in the gas-liquid multiphase flow, the size of bubbles in the gas-liquid multiphase flow, and the density of the bubbles is changed. The flow mode evaluation device according to any one of claims 1 to 4, which is a module machine-learned using as input data for learning. In addition to the mixed-phase flow image, the evaluation module learns at least one numerical information of the moving speed of bubbles in the gas-liquid multiphase flow, the size of bubbles in the gas-liquid multiphase flow, and the density of the bubbles. The flow mode evaluation device according to any one of claims 1 to 6, which is a machine-learned module as input data for use. In addition to the multiphase flow image, the evaluation module performs machine learning using at least one numerical information of the superficial velocity of the gas in the pipe and the superficial velocity of the liquid in the pipe as input data for learning. The flow mode evaluation device according to any one of claims 1 to 7, which is a module. A gas production system for producing gas by decomposing a gas hydrate in the ground, provided with the flow mode evaluation device according to any one of claims 1 to 8.
A long tube having a tip configured to be buried in the ground, a gas hide outside the tube using the pressure generated by the liquid in the tube extending upward from the tip. By reducing the pressure acting on the rate, a gas-liquid mixture containing bubbles generated by decomposition from the gas hydrate is provided at the tip of the pipe and outside the pipe so as to be taken into the liquid in the pipe. A riser tube with an open hole and
A gas-liquid separator provided in the riser tube to separate the bubbles from the liquid in the riser tube that has taken in the gas-liquid mixture.
A gas generation line including a gas generation pipe for taking out the gas generated from the bubbles as a gas to be produced from the riser pipe, and a gas generation line.
A pump that sucks up the liquid in order to discharge the liquid from which the gas has been separated by the gas-liquid separator from the riser pipe.
A liquid discharge line including a liquid discharge pipe that takes out the liquid sucked up by the pump from the riser pipe, and
The conduit serving as the path of the liquid leading from the gas-liquid separator to the pump is the pipe in the flow mode evaluation device, and the bubbles remaining in the liquid and the liquid are provided in the pipe as a gas-liquid mixed phase flow. The image acquisition unit that acquires an image of the gas-liquid mixed phase flow when the gas-liquid mixed phase flow passes through the observation position by imaging the gas-liquid mixed phase flow passing through the observed position, and the flow mode. The machine learning unit of the evaluation device and the evaluation unit of the flow mode evaluation device are provided so that the liquid level of the liquid is formed in the gas-liquid separation device and the decomposition of the gas hydrate can be controlled. In addition, the first control that controls the rotation of the pump according to the pressure at the tip portion, and the rotation speed of the pump that is controlled by the first control based on the evaluation result of the flow mode by the evaluation unit. A gas production system comprising a control device for performing a second control for further controlling the gas. It is a flow mode evaluation method that evaluates the flow mode in a gas-liquid mixed phase flow of liquid and gas flowing in a pipe.
A step of acquiring an image of the gas-liquid multiphase flow when the gas-liquid multiphase flow passes through the observation position by imaging a gas-liquid multiphase flow passing through a predetermined observation position in the pipe.
A plurality of mixed-phase flow images obtained by changing the generation conditions of the gas-liquid mixed-phase flow of the liquid and the gas are used as learning input data, and the classification results of the flow modes of the plurality of mixed-phase flow images are used as learning output data. A step of preparing an evaluation module in which the relationship between the multiphase flow image and the classification result of the flow mode is machine-learned, and
The feature is that the acquired image of the gas-liquid multiphase flow is used as an evaluation image, and the evaluation module is used to evaluate the flow mode of the gas-liquid multiphase flow when passing through the observation position. Flow style evaluation method. In the step of acquiring an image of the gas-liquid multiphase flow, the gas-liquid multiphase flow passing through the observation position is sequentially imaged in time series at a predetermined frame rate, and a plurality of the gas-liquid multiphase flow at the observation position are sequentially imaged. The flow mode evaluation method according to claim 10, further comprising a process of creating one still composite image of the gas-liquid multiphase flow when passing through the observation position by connecting the images in chronological order. Each of the mixed-phase flow images used as the learning input data sequentially images the gas-liquid multiphase flow passing through a region having the same size as the observation position in time series at the frame rate, and the gas in the region. The flow mode evaluation method according to claim 11, which is one static composite image of the gas-liquid multiphase flow created by connecting a plurality of images of the liquid-mixed multiphase flow in chronological order.

Images