JP7664875B2 - Flow behavior prediction system, flow behavior prediction method, prediction model generation method, and trained prediction model - Google Patents

Description

An embodiment of the present invention relates to flow behavior prediction technology.

It is necessary to protect the safety of workers in indoor workplaces where harmful gases or dust are generated. In addition, droplets and other fine particles that are the source of infection for infectious diseases are greatly affected by the surrounding air currents, so ventilation is necessary to reduce the risk of infection indoors. Numerical simulation is therefore considered useful for predicting and verifying the effectiveness of measures that take the air current environment into account. Conventionally, there is known a technique for evaluating ventilation through numerical simulation using general-purpose thermal fluid analysis software. However, airflow changes significantly depending on changes in the conditions within the space, such as the positions of doors, open windows, and air supply and exhaust vents, the flow rate of airflow passing through these openings, and the placement of obstacles in the room. Numerical simulation requires a long time for calculations, making it difficult to obtain results in real time.

Conventional fluid simulations, for example, require the generation of computational grids and the numerical convergence of the Navier-Stokes equations. Therefore, evaluating the behavior of fluids that change from moment to moment generally requires a significant amount of computation time. For this reason, it is not realistic to perform a simulation and grasp the state of the fluid each time the state in space changes, as this involves cumbersome work and takes a long time to calculate.

In recent years, various technologies using machine learning have emerged. The calculation time using neural networks is much shorter than that using simulations. For example, a technology is known in which the fluid state of invisible materials inside manufacturing equipment used for liquid phase growth of semiconductors is calculated using a machine learning predictive model, and an image showing the fluid state of this material is projected in real time onto the exterior wall surface of the manufacturing equipment.

JP 2018-169818 A

"No. 65 "Airflow Simulation"", [Retrieved December 16, 2021], Internet <URL: https://www.jfe-tec.co.jp/jfetec-news/65/2p.html> "Special feature: Virus spread as seen from "correct ventilation" simulations", [Retrieved December 16, 2021], Internet <URL: http://www.env-simulation.com/jp/merumaga/kanki_tokushuu>

The problem that this invention aims to solve is to provide a flow behavior prediction technology that can predict the behavior of a fluid in a space using a machine learning prediction model.

A flow behavior prediction system according to an embodiment of the present invention includes a data acquisition unit that acquires first state data indicating at least a shape of a space and a behavior of a fluid in the space that changes depending on the shape of the space , and a computer that inputs at least one of the first state data or data simulating the first state data into a prediction model whose parameters have been machine-learned using teacher data, and outputs at least one of second state data indicating a behavior of the fluid in the future relative to the first state or data simulating the first state data, and outputs the data of the second state. The data acquisition unit acquires an entrance and exit of the fluid in the space and a position of an object present in the space, Measurements are taken to obtain data indicating the current state of the shape of the space, the data regarding the shape of the space including data on the shapes of the objects that exist within the space and affect the behavior of the fluid, and the predictive model predicts the future behavior of the fluid taking into account the influence that the objects present within the space have on the behavior of the fluid, and when machine learning is performed on the predictive model, multiple patterns are set for the entrances and exits of the fluid in the space and the basic shapes and positions of the objects present in the space, and for each pattern, data including the physical quantities of the fluid is obtained by numerical simulation or statistical data, and machine learning is performed using the data including the physical quantities of the fluid as the training data .

An embodiment of the present invention provides a flow behavior prediction technology that can predict the behavior of a fluid in a space using a machine learning prediction model.

FIG. 1 is a configuration diagram showing a flow behavior prediction system according to a first embodiment. FIG. 1 is a block diagram showing a flow behavior prediction system. FIG. 1 is a block diagram showing a prediction model. FIG. 11 is a screen diagram showing a display mode in which the behavior of a fluid is visualized. FIG. 11 is a screen diagram showing a display mode in which the physical quantities of a fluid are visualized using isosurfaces. FIG. 13 is a screen diagram showing a display mode in which a specific area of a space is emphasized. 1 is a flowchart illustrating a method for generating a predictive model. 1 is a flowchart showing a flow behavior prediction method. 10 is a flowchart showing a mode of switching the display contents on the screen. 11 is a flowchart showing a warning output process. 10 is a flowchart showing a flow behavior prediction method according to a second embodiment. FIG. 13 is a configuration diagram showing a flow behavior prediction system according to a third embodiment. FIG. 1 is a block diagram showing a flow behavior prediction system. 4 is a flowchart showing a control process of the ventilation device.

First Embodiment
Hereinafter, embodiments of a flow behavior prediction system, a flow behavior prediction method, a prediction model generation method, and a trained prediction model will be described in detail with reference to the drawings. First, a first embodiment will be described with reference to FIGS. 1 to 10.

Reference numeral 1 in FIG. 1 denotes a flow behavior prediction system of the first embodiment. This flow behavior prediction system 1 predicts and visualizes the behavior of a fluid inside a specified space 2 by machine learning using a neural network or the like. An example of the space 2 is the inside (room) of a room provided in a specified building. This space 2 may also be a room provided in a vehicle, aircraft, ship, or the like. Examples of the room include a factory, a workshop, a hospital examination room, a waiting room, an underground mall, a train passenger compartment, and the like. The room may also be a containment vessel of a nuclear power plant, or the like.

Inside the space 2, there is a fluid such as air. Inside the space 2, there are also certain objects 3 (obstacles) that affect the behavior of the fluid. Examples of objects 3 include certain furniture such as chairs, desks, tables, and shelves. Note that the object 3 may also be a person in the room. The shape of the room is made up of, for example, a floor surface 4, walls 5, and a ceiling 6. Furthermore, the shape of the room includes the shapes of certain openings 7 through which air enters and exits, such as windows, doors, intake vents, and exhaust vents. The shape of the room and the shapes of the objects 3 present in the room make up the shape of the space 2. The behavior of the fluid changes according to the shape of the space 2.

The flow behavior prediction system 1 comprises a system main body 10 and multiple fluid data acquisition units 11. The system main body 10 is provided in a position that is easily visible to a user using the flow behavior prediction system 1. For example, the system main body 10 is provided on a wall surface 5 of a room. The multiple fluid data acquisition units 11 are provided on the floor surface 4, wall surface 5, ceiling 6, near the object 3, etc. of the room.

The fluid data acquisition unit 11 acquires data on the physical quantities of the fluid at any position in the space 2. In this way, it is possible to predict the future behavior of the fluid from the physical quantities of the fluid at an actual position in the space 2.

The fluid data acquisition unit 11 is composed of a predetermined fluid sensor, etc. For example, the fluid data acquisition unit 11 measures at least one physical quantity of the speed, direction, temperature, pressure, components, etc. of the fluid present in the space 2. Each fluid data acquisition unit 11 is capable of communicating with the system main body 10, and data indicating the measured physical quantity is transmitted to the system main body 10. For example, time-series data indicating the temporal change in the physical quantity of the fluid at each point in the space 2 is acquired by each fluid data acquisition unit 11 and transmitted to the system main body 10.

These time series data include information on the time and calendar (date and day of the week, etc.) when the fluid physical quantities were acquired, as well as information indicating the position of the fluid data acquisition unit 11 that acquired each physical quantity. In the following description, the data acquired by the fluid data acquisition unit 11 is referred to as fluid data (Figure 3).

Fluid behavior refers to at least one of the speed and direction of the fluid. Fluid data includes the physical quantities (e.g., scalars, vectors) of the speed and direction of the fluid, as well as other physical quantities (e.g., temperature, pressure, content of components). In this way, future physical quantities of the fluid can be predicted taking into account not only the speed and direction of the fluid, but also other physical quantities of the fluid.

Next, the configuration of the flow behavior prediction system 1 will be described with reference to the block diagram shown in FIG. 2. Note that the flow behavior prediction system 1 may include components other than those shown in FIG. 2, and some of the components shown in FIG. 2 may be omitted.

The system body 10 of the flow behavior prediction system 1 includes a spatial data acquisition unit 12, a display unit 13, a memory unit 14, a warning output unit 15, a communication unit 16, and a control unit 17.

The control unit 17 is a computer that provides overall control of the flow behavior prediction system 1. This control unit 17 includes a calculation unit 18 and a visualization information generation unit 19. These are realized by the CPU executing a program stored in the memory or HDD.

The flow behavior prediction system 1 is configured as a computer having hardware resources such as a CPU, ROM, RAM, and HDD, and in which software-based information processing is realized using the hardware resources as the CPU executes various programs. Furthermore, the flow behavior prediction method and the prediction model generation method are realized by having the computer execute various programs.

Each component of the flow behavior prediction system 1 does not necessarily have to be provided on one computer. For example, one flow behavior prediction system 1 may be realized by multiple computers connected to each other via a network. For example, the calculation unit 18 and the visualization information generation unit 19 may each be mounted on a separate computer. Also, the display unit 13 may be mounted on another computer connected via a network. Note that the flow behavior prediction method and the prediction model generation method may each be executed on separate computers.

The spatial data acquisition unit 12 acquires data on the shape of the space 2. This data on the shape of the space 2 includes data on the shape of the objects 3 that exist inside the space 2 and affect the behavior of the fluid. In this way, it is possible to predict the future behavior of the fluid by taking into account the influence that the objects 3 present inside the space 2 have on the behavior of the fluid. The spatial data acquisition unit 12 is provided, for example, on the upper front side of the system main body 10 (Figure 1). The spatial data acquisition unit 12 mounted on the system main body 10 senses the positions of the entrances and exits of the fluid in the space 2 (for example, the openings 7), the objects 3 present in the space 2, and the like, and acquires data indicating their current states.

The spatial data acquisition unit 12 is composed of a specified camera or shape measurement sensor. For example, the camera takes an image. This image contains information indicating the shape of the space 2 including the object 3. By taking an image with this camera, data on the shape of the space 2 including the surrounding objects 3 is obtained from the system main body 10.

The shape measurement sensor measures the three-dimensional shape (solid shape) of the space 2 including the object 3. For example, a depth sensor is used as the shape measurement sensor. Note that an infrared sensor or a laser sensor such as LiDAR may also be used as the shape measurement sensor.

For example, the shape measurement sensor can measure the distance from the system body 10 to the surrounding object 3 by projecting a laser onto the object 3 and receiving the reflected light with a light receiving element. The shape measurement sensor also uses a ToF (Time of Flight) method to convert the delay time between the projected pulse and the received pulse into distance, and measures the distance from the system body 10 to the surrounding object 3. Using this shape measurement sensor, it is possible to generate three-dimensional point cloud data indicating the shape of the space 2 including the surrounding object 3 from the system body 10.

The shape of the space 2 includes information on the positions (coordinates) of the floor 4, walls 5, and ceiling 6 that make up the space 2. The shape of the object 3 includes information on the position (coordinates) where the object 3 is placed in the space 2. The shape data of the object 3 also includes information on the time when the object 3 is present and calendar information (date and day of the week, etc.). In the following explanation, the data acquired by the spatial data acquisition unit 12 includes not only shape information, but also information indicating position and time. These are referred to as spatial data (Figure 3).

The display unit 13 displays predetermined information. The display unit 13 is, for example, a display that outputs the results of predicting the behavior of the fluid. The display unit 13 is provided, for example, on the front side of the system main body 10 (Figure 1). The display unit 13 may be separate from the system main body 10, or may be integrated with it.

In the first embodiment, a display is exemplified as the display unit 13 that displays images, but other configurations are also possible. For example, images may be displayed using a head-mounted display or a projector. Furthermore, a printer that prints information on paper media may be used instead of a display. In other words, the display unit 13 may include a head-mounted display, a projector, or a printer.

The memory unit 14 stores various information necessary for predicting the behavior of a fluid. For example, the memory unit 14 stores spatial data and fluid data. Furthermore, the memory unit 14 stores a prediction model 20 (Figure 3).

The warning output unit 15 outputs a predetermined warning according to the state of the fluid. The warning may be output as sound or in a visually recognizable manner. Furthermore, the warning may be sent to another computer (not shown) connected via a predetermined network.

The warning output unit 15 in the first embodiment is composed of a speaker or buzzer that issues a warning. The warning output unit 15 may also be a lamp that flashes when a warning is issued. Additionally or alternatively, the display unit 13 may output a warning using an image or the like.

The communication unit 16 communicates with each of the fluid data acquisition units 11. The communication unit 16 and the fluid data acquisition unit 11 may be connected by wireless communication or by wired communication. The communication unit 16 may also communicate with other computers (not shown) connected via a specified network.

The calculation unit 18 includes a prediction model 20 (Figure 3) that has undergone machine learning. The calculation unit 18 performs calculations to predict the future state of the fluid using this prediction model 20.

The calculation unit 18 includes artificial intelligence (AI) that performs machine learning. The calculation unit 18 also includes a deep learning unit that extracts a specific pattern from multiple patterns based on deep learning.

In the first embodiment, prediction using a computer uses a prediction technique based on artificial intelligence learning. For example, a learning model generated by machine learning using a neural network, a learning model generated by other machine learning, a deep learning algorithm, a mathematical algorithm such as regression analysis, etc. Forms of machine learning include clustering, deep learning, etc.

The flow behavior prediction system 1 includes a computer having artificial intelligence that performs machine learning. For example, this system may be composed of a single computer having a neural network, or may be composed of multiple computers having neural networks.

Here, a neural network is a mathematical model that represents the characteristics of brain functions through computer simulation. For example, this model shows that artificial neurons (nodes) that form a network through synaptic connections change the strength of synaptic connections through learning, and acquire problem-solving abilities. Furthermore, neural networks acquire problem-solving abilities through deep learning.

As shown in FIG. 3, the prediction model 20 includes an input layer 21, one or more intermediate layers 22, and an output layer 23. The input layer 21 receives first state data indicating at least the shape of the space 2 and the behavior of the fluid inside the space 2. The output layer 23 outputs second state data indicating at least the behavior of the fluid in the future relative to the first state.

The data of the first state includes spatial data acquired by the spatial data acquisition unit 12 and fluid data of the current state acquired by the fluid data acquisition unit 11. The spatial data includes information such as the shape of the room, the shape of the furniture, and the shape of a person. The fluid data includes information such as the speed, direction, temperature, pressure, and components of the fluid.

The data in the second state includes fluid data in a state further in the future than the first state. In the first embodiment, the prediction is made under the assumption that the spatial data will not change in the future, so the data in the second state does not include spatial data. However, the data in the second state may also include future spatial data. For example, this future spatial data is identical to the data in the first state except for the time and calendar information. Here, the time and calendar information is information that is determined to change in the future.

The second state that is in the future from the first state may be one second, one minute, ten minutes, one hour, or one day after the first state. The time difference between the first state and the second state can be arbitrarily set by the user in advance.

The intermediate layer 22 has parameters machine-learned using training data in which at least one of data in a first state or data simulating the first state is input, and at least one of data in a second state or data simulating the second state is output.

This prediction model 20 causes the computer to function by inputting data of the first state into the input layer 21, performing calculations in the intermediate layer 22, and outputting data of the second state from the output layer 23.

For example, the neural network that constitutes the prediction model 20 is provided with an intermediate layer 22 having multiple layers. Each layer of this intermediate layer 22 is composed of multiple units. Furthermore, by having the multi-layered neural network learn in advance using training data, it is possible to automatically extract features in the pattern of changes in the state of the circuit or system. Note that the number of intermediate layers, number of units, learning rate, number of times of learning, and activation function of the multi-layered neural network can be set on the user interface.

In addition, deep reinforcement learning may be used in the neural network, in which a reward function is set for each information item to be learned, and the information item with the highest value is extracted based on the reward function.

For example, a convolutional neural network (CNN), which has a proven track record in image recognition, is used. In this CNN, the intermediate layer 22 is composed of a convolutional layer and a pooling layer. The convolutional layer obtains a feature map by applying filtering to nearby nodes in the previous layer. The pooling layer further reduces the feature map output from the convolutional layer to create a new feature map. At this time, by obtaining the maximum value of the pixels contained in the region of interest in the feature map, it is possible to absorb slight deviations in the position of the feature amount.

The convolutional layer extracts local features from an image, and the pooling layer consolidates local features. These processes reduce the size of the input image while maintaining its features. In other words, a convolutional neural network (CNN) can significantly compress (abstract) the amount of information contained in an image. The input image can then be recognized and classified using the abstracted image image stored in the neural network.

Note that there are various machine learning techniques, such as autoencoders, long short-term memory (LSTM), signed distance functions (SDF), generative adversarial networks (GANs), and recurrent neural networks (RNNs). These techniques may be applied to machine learning.

For example, CNN and LSTM are used for machine learning. For example, SDF is used to input spatial data to the prediction model 20. The prediction model 20 has, for example, both a CNN that compresses spatial data and an LSTM that performs time series prediction.

As the teacher data, data actually acquired by the spatial data acquisition unit 12 and the fluid data acquisition unit 11 can be used. Alternatively, data simulating these may be used as the teacher data. Note that data at a specified time is set as data of the first state, and data at a time in the future is set as data of the second state.

For example, when performing machine learning of the prediction model 20, first, multiple patterns of the basic shapes and positions of the entrances and exits of the fluid in the space 2 (e.g., the openings 7), the objects 3 present in the space 2, etc. are set. Then, for each pattern, data including the physical quantities of the fluid is obtained by numerical simulation or statistical data, etc. Machine learning is performed using these as training data.

In other words, the parameters of the prediction model 20 are machine-learned using training data in which data obtained by simulating multiple states of the space 2 having an arbitrary base shape is used as input and output. In this way, a large amount of training data can be obtained in advance by simulations based on the base shape.

Next, when the behavior of the fluid is predicted, the data of the first state and the second state actually acquired by the spatial data acquisition unit 12 and the fluid data acquisition unit 11 are input to the prediction model 20. Time series data of physical quantities indicating the future behavior of the fluid at each point in the space 2 is predicted. The prediction results are displayed on the display unit 13. For example, first, the behavior of the fluid is displayed on the display unit 13 (Figure 1) in accordance with the actual time. Next, the behavior of the fluid in the future beyond the actual time is displayed on the display unit 13. In this way, it is possible to display the prediction results in real time in accordance with the actual state of the fluid.

Calculating a numerical simulation of fluid behavior using basic shapes and machine learning using the results of that calculation take time. However, by completing a prediction model 20 using machine learning in advance, it is possible to instantly predict time series data for fluid behavior in any space 2, and to grasp time series data for future physical quantities of the fluid. As a result, an image of visualized information showing the state of the fluid can be displayed on the display unit 13 in real time.

The visualization information generating unit 19 (FIG. 2) generates visualization information that visualizes the state of the fluid based on the second state data output from the prediction model 20. This visualization information is displayed on the display unit 13. In this way, the user can see the visualized state of the fluid and understand the ventilation status of the space 2.

Next, using Figures 4 to 6, examples of the image form of the visualized information displayed on the screen 30 of the display unit 13 will be shown.

For example, as shown in Figure 4, an image visualizing the behavior of a fluid displays a plan view of a specified space 2 as a room. An object 3 existing in this space 2 is also displayed. The fluid moving through the space 2 is displayed as a flow of arrows. The user can grasp the behavior of the fluid from the present to a specified future state.

The visualization information generating unit 19 (FIG. 2) also calculates the physical quantities of the fluid at each position in the space 2, and generates visualization information that indicates these physical quantities using at least one of contours, vectors, and isosurfaces. For example, the visualization information generating unit 19 converts the physical quantities, such as the fluid speed, direction, temperature, pressure, components, and age of air, predicted by the prediction model 20, into isosurfaces.

As shown in FIG. 5, in an image in which the physical quantities of a fluid are visualized using isosurfaces, regions are displayed that are color-coded according to the physical quantities. For example, the age of air in the fluid is displayed using different colors. Temperature distribution may also be displayed using different colors. Furthermore, contour bars 31 and the like are also displayed. In this way, the user can grasp the state of the fluid using isosurfaces. The user can grasp the age of air in real time or in the future.

The physical quantities of the fluid can be displayed not only as isosurfaces, but also as contours or vectors. Furthermore, these can be displayed in combination, or switched between in sequence. In this way, the user can grasp the physical quantities of the fluid from various perspectives.

Furthermore, the visualization information generating unit 19 (FIG. 2) generates visualization information that highlights and displays an area 32 in the space 2 where the value of a specific component contained in the fluid exceeds a threshold value among the physical quantities of the fluid at each position in the space 2. For example, components such as hydrogen or carbon dioxide are set in advance as targets to be highlighted and displayed.

As shown in FIG. 6, if there is an area 32 where the value of a component such as hydrogen or carbon dioxide exceeds a threshold, this area 32 is highlighted and displayed in a manner different from the others. For example, the highlighted area 32 is heated up (colored), surrounded by a line, or indicated by an arrow. In this way, the user can grasp the area 32 in the space 2 where the value of a particular component exceeds the threshold.

Next, the method for generating a predictive model will be explained using the flowchart in Figure 7. The aforementioned drawings will be referenced as appropriate.

First, in step S1, the user prepares training data. For example, the user sets multiple patterns of basic shapes and positions of objects 3 and the like present in space 2. For example, data showing the behavior of a fluid in a flow path having one simple structure is created. Then, multiple data of various patterns in which the simple structure and the shape of the flow path are changed are created. Note that data actually acquired by the spatial data acquisition unit 12 and the fluid data acquisition unit 11 may be used as training data.

In the next step S2, the user selects a machine learning method. For example, CNN and LSTM are selected as machine learning methods.

In the next step S3, the user selects the data necessary for machine learning. Data that is not suitable for machine learning is removed from the training data.

In the next step S4, the user performs data wrangling on the training data. For example, complex data is organized and integrated so that the training data can be easily accessed and analyzed.

In the next step S5, the user splits the training data into training data and validation data.

In the next step S6, the user sets hyperparameters. For example, the behavior of the machine learning algorithm is set.

In the next step S7, a computer that performs machine learning, for example, the control unit 17 (FIG. 2), performs machine learning of parameters using the acquired training data. Then, the control unit 17 generates a machine-learned prediction model 20 (FIG. 3) that uses the data in the first state as input and the data in the second state as output. Note that machine learning may be performed by a computer other than the control unit 17.

In the next step S8, the control unit 17 inputs the first state data for verification to the prediction model 20 and outputs the second state data that is the prediction result.

In the next step S9, the control unit 17 compares the second state data output from the prediction model 20 with the verification data.

In the next step S10, the control unit 17 determines whether the prediction result of the prediction model 20 has reached the required accuracy. If the required accuracy has been reached (YES in step S10), the process proceeds to step S11. On the other hand, if the required accuracy has not been reached (NO in step S10), the user is notified of this fact and the process returns to step S2.

In step S11, the control unit 17 stores the completed optimal prediction model 20 in the memory unit 14 (Figure 2).

Then, the prediction model generation method is terminated. Note that the above steps are at least a part of the prediction model generation method, and other steps may be included in the prediction model generation method.

Next, the flow behavior prediction method will be explained using the flowchart in Figure 8. The aforementioned drawings will be referenced as appropriate.

First, in step S21, the spatial data acquisition unit 12 (Figure 1) acquires the current spatial data in space 2.

In the next step S22, the fluid data acquisition unit 11 (Figure 1) acquires the current fluid data in space 2.

In the next step S23, the control unit 17 (Figure 2) inputs the acquired spatial data and fluid data as first state data into the input layer 21 of the machine-learned prediction model 20 (Figure 3).

In the next step S24, the prediction model 20 (calculation unit 18) performs calculations in the intermediate layer 22 where the parameters have been machine-learned.

In the next step S25, the prediction model 20 (calculation unit 18) outputs the second state data from the output layer 23. Note that training data may be generated based on this output second state data, and machine learning of the prediction model 20 may be performed.

In the next step S26, the visualization information generator 19 (FIG. 2) generates visualization information that visualizes the state of the fluid based on the second state data output from the prediction model 20, i.e., the future fluid data.

In the next step S27, the display unit 13 (Figure 1) displays an image of the generated visualization information on the screen 30 (Figures 4 to 6).

Then, the flow behavior prediction method is terminated. The processing of this flow behavior prediction method is constantly executed and repeated when the flow behavior prediction system 1 is in operation. Note that the above steps are at least a part of the flow behavior prediction method, and other steps may be included in the flow behavior prediction method.

Next, the manner in which the display content of the screen 30 of the display unit 13 is switched will be explained using the flowchart in FIG. 9. The aforementioned drawings will be referred to as appropriate.

The control unit 17 (Figure 2) generates visualization information that switches between displaying a first type of physical quantity and a second type of physical quantity that is different from the first type of physical quantity at regular intervals, which is a physical quantity of the fluid at an arbitrary position in the space 2. In this way, the user can grasp the state of the various physical quantities of the fluid.

For example, in step S31, the control unit 17 causes the screen 30 of the display unit 13 (FIG. 1) to display an image of visualization information showing the velocity and direction as physical quantities of the fluid (e.g., a first type of physical quantity). Here, the display is continued until a preset arbitrary time has elapsed.

In the next step S32, the control unit 17 causes the screen 30 of the display unit 13 to display an image of visualized information showing the concentration of carbon dioxide, which is a component as a physical quantity (e.g., a second type physical quantity) of the fluid. Here, the display is continued until a given time has elapsed.

In the next step S33, the control unit 17 causes the screen 30 of the display unit 13 to display an image of visualized information showing the concentration of hydrogen, which is a component of the fluid as a physical quantity. Here, the display is continued until a desired time has elapsed.

In the next step S34, the control unit 17 causes the screen 30 of the display unit 13 to display an image of visualized information showing the temperature as a physical quantity of the fluid. Here, the display is continued until a given time has elapsed.

In the next step S35, the control unit 17 causes the screen 30 of the display unit 13 to display an image of visualized information showing the pressure as a physical quantity of the fluid. Here, the display continues until a given time has elapsed.

Then, the process of switching the display content is terminated. This process of switching the display content is constantly executed and repeated when the flow behavior prediction system 1 is in operation. Note that the above steps are at least a part of the process of switching the display content, and other steps may be included in the process of switching the display content.

Next, the warning output process will be explained using the flowchart in Figure 10. The aforementioned drawings will be referenced as appropriate.

In the flow behavior prediction system 1, specific components such as hydrogen, carbon dioxide, and toxic gases are preset as targets for warnings. In addition, thresholds at which warnings are issued are preset.

First, in step S41, the control unit 17 (Fig. 2) determines whether or not a specific component that is the subject of a warning, among various components contained in the fluid present in the space 2, exceeds a threshold value. This determination is made based on the data of the second state output from the prediction model 20 (Fig. 3). Note that this determination may also be made based on the data of the first state.

If the specific component does not exceed the threshold (NO in step S41), the warning output process ends. On the other hand, if the specific component exceeds the threshold (YES in step S41), the process proceeds to step S42.

In the next step S42, the warning output unit 15 outputs a warning. For example, the warning output unit 15 outputs an alarm sound. In this way, a warning is given to the user when the value of a specific component inside the space 2 exceeds the threshold value.

The display unit 13 may output a warning using an image or the like. Additionally or alternatively, the display unit 13 may highlight and display an area 32 (FIG. 6) in which the value of a specific component exceeds the threshold in a manner different from the others. This allows the user to grasp the area 32 in which a specific component is likely to exceed the threshold in real time or in the future.

Then, the warning output process is terminated. This warning output process is executed and repeated at all times while the flow behavior prediction system 1 is in operation. Note that the above steps are at least a part of the warning output process, and other steps may be included in the warning output process.

In the first embodiment, no calculation time is required to perform a numerical simulation of the fluid, and the real-time or future behavior of the fluid is displayed on the display unit 13. The user can grasp the state of the fluid inside the specified space 2.

Second Embodiment
Next, a second embodiment will be described with reference to Fig. 11. Note that the same components as those shown in the above-mentioned embodiment are denoted by the same reference numerals and the duplicated description will be omitted.

The control unit 17 (Figure 2) inputs the second state data output from the prediction model 20 (Figure 3) to the prediction model 20 as first state data, and further outputs the second state data indicating the future behavior of the fluid.

The flow behavior prediction method of the second embodiment differs from the flow behavior prediction method of the first embodiment only in step S25A, and the other steps are the same as those of the first embodiment.

In step S25A, which follows step S25, the control unit 17 (Fig. 2) determines whether the second state data output from the prediction model 20 (Fig. 3) is data for the future time to be predicted. If it is data for the future time to be predicted (YES in step S25A), the process proceeds to step S26. On the other hand, if it is not data for the future time to be predicted (NO in step S25A), the process returns to step S23. Then, in step S23, the second state data is input to the prediction model 20 as first state data. Note that spatial data may be included in this second state data.

For example, the user sets the future time to be predicted as 10 minutes in the future. Then, the user sets the prediction model 20 to output data of the second state one minute into the future. In the second embodiment, the process from step S23 to step S25A is repeated about 10 times to obtain data of the second state at the future time to be predicted.

Continuing to predict future fluid behavior using machine learning will result in errors. For example, predicting the future one minute from now will result in less error than predicting the future 10 minutes from now. Therefore, in the second embodiment, errors can be reduced by repeatedly making predictions at certain intervals.

Third Embodiment
Next, a third embodiment will be described with reference to Figures 12 to 14. Note that the same components as those shown in the above-described embodiment are denoted by the same reference numerals, and duplicated description will be omitted.

As shown in Figs. 12 and 13, the control unit 17 of the third embodiment controls a specified device to lower a value of a specific component contained in the fluid among the physical quantities of the fluid at each position in the space 2 when the value of the specific component exceeds a threshold value. The flow behavior prediction system 1A of the third embodiment is connected to a ventilation device 40 as a specified device instead of the warning output unit 15 (Fig. 2) of the first embodiment. The ventilation device 40 is provided, for example, on the ceiling 6, and is controlled by the control unit 17.

Next, the control process of the ventilation device 40 will be explained using the flowchart in FIG. 14. The aforementioned drawings will be referenced as appropriate.

Specific components such as hydrogen, carbon dioxide, and toxic gases are preset as targets for monitoring in the flow behavior prediction system 1. In addition, a threshold value for activating the ventilation device 40 is preset.

First, in step S51, the control unit 17 (FIG. 13) determines whether a specific component being monitored, among various components contained in the fluid present in the space 2, exceeds a threshold value. This determination is made based on the data of the second state output from the prediction model 20 (FIG. 3). Note that this determination may also be made based on the data of the first state.

If the specific component exceeds the threshold (YES in step S51), proceed to step S52. On the other hand, if the specific component does not exceed the threshold (NO in step S51), proceed to step S53.

In step S52, the control unit 17 operates the ventilation device 40 (FIG. 13). If the ventilation device 40 is constantly operating, the rotation speed of the fan of the ventilation device 40 may be increased. The control unit 17 may also notify the user to encourage ventilation. Then, the control process of the ventilation device 40 is terminated.

In step S53, the control unit 17 stops the ventilation device 40 (FIG. 13). Then, the control process of the ventilation device 40 is terminated.

This control process of the ventilation device 40 is constantly executed and repeated when the flow behavior prediction system 1 is in operation. Note that the above steps are at least a part of the control process of the ventilation device 40, and other steps may be included in the control process of the ventilation device 40.

Although the ventilator 40 is shown as an example of the predetermined device, other modes may be used. For example , it may be a hydrogen removal device that is activated when the hydrogen concentration exceeds a threshold value, or a cooling device that is activated when the temperature of the fluid exceeds a threshold value.

According to the third embodiment, when the value of a particular component inside space 2 is about to exceed a threshold value in real time or in the future, or when this is predicted, the value of the particular component can be automatically lowered.

The flow behavior prediction system, flow behavior prediction method, prediction model generation method, and trained prediction model have been described based on the first to third embodiments, but the configuration applied in any one of the embodiments may be applied to the other embodiments, and the configurations applied in each embodiment may be combined.

In the above-mentioned embodiment, the judgment of an arbitrary value (value of a specific component) using a reference value (threshold value) may be a judgment of "whether the arbitrary value is equal to or greater than the reference value", a judgment of "whether the arbitrary value exceeds the reference value", a judgment of "whether the arbitrary value is equal to or less than the reference value", or a judgment of "whether the arbitrary value is less than the reference value". The reference value may not be fixed, but may be variable. Therefore, a value within a predetermined range may be used instead of the reference value, and a judgment of whether the arbitrary value falls within the predetermined range may be made. Furthermore, the error occurring in the device may be analyzed in advance, and a predetermined range including the error range centered on the reference value may be used for the judgment.

In the flowchart of the above embodiment, the steps are executed in series, but the order of steps is not necessarily fixed, and some steps may be executed in reverse. Some steps may be executed in parallel with other steps.

The system of the above-mentioned embodiment includes a control device that is a highly integrated processor such as an FPGA (Field Programmable Gate Array), a GPU (Graphics Processing Unit), a CPU (Central Processing Unit) and a dedicated chip, a storage device such as a ROM (Read Only Memory) and a RAM (Random Access Memory), an external storage device such as an HDD (Hard Disk Drive) and an SSD (Solid State Drive), a display device such as a display, an input device such as a mouse and a keyboard, and a communication interface. This system can be realized with a hardware configuration that uses a normal computer.

The program executed by the system of the above-mentioned embodiment is provided in advance in a ROM or the like. Additionally or alternatively, the program may be provided in the form of an installable or executable file stored on a non-transitory computer-readable storage medium such as a CD-ROM, CD-R, memory card, DVD, or flexible disk (FD).

The programs executed by this system may also be stored on a computer connected to a network such as the Internet and provided by downloading them via the network. This system may also be configured by combining separate modules that independently perform the functions of the components and connect them together via a network or dedicated lines.

In the above embodiment, a gas such as air is exemplified as a fluid, but other forms are also possible. For example, the prediction model 20 may be used to predict the behavior of a liquid such as water.

In the above embodiment, the interior of a room is exemplified as the space 2 to be predicted, but other configurations are also possible. For example, the space 2 to be predicted may be the interior of a specific manufacturing device, and the future state of the fluid therein may be predicted by the prediction model 20.

According to at least one of the embodiments described above, by providing a computer that inputs data of at least the first state acquired by the data acquisition unit into the prediction model and outputs data of the second state, it is possible to predict the behavior of a fluid in a space using a prediction model based on machine learning.

Although several embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, substitutions, modifications, and combinations can be made without departing from the spirit of the invention. These embodiments or modifications thereof are within the scope of the invention and its equivalents as set forth in the claims, as well as the scope and spirit of the invention.

1 (1A)...flow behavior prediction system, 2...space, 3...object, 4...floor, 5...wall, 6...ceiling, 7...opening, 10...system body, 11...fluid data acquisition unit, 12...spatial data acquisition unit, 13...display unit, 14...storage unit, 15...warning output unit, 16...communication unit, 17...control unit, 18...calculation unit, 19...visualization information generation unit, 20...prediction model, 21...input layer, 22...middle layer, 23...output layer, 30...screen, 31...contour bar, 32...area, 40...ventilation device.

Claims (14)

a data acquisition unit that acquires first state data indicating at least a shape of a space and a behavior of a fluid in the space that changes depending on the shape of the space;
A prediction model in which parameters are machine-learned using teacher data in which at least one of data of the first state or data simulating the first state is input and at least one of data of a second state or data simulating the second state, which indicates the behavior of the fluid in the future from at least the first state, is output.
inputting the data of the first state acquired by at least the data acquisition unit, and outputting the data of the second state;
A computer,
Equipped with
the data acquisition unit measures an inlet/outlet of the fluid in the space and a position of an object present in the space to acquire data indicating a current shape state of the space;
The data on the shape of the space includes data on the shape of the object that exists within the space and affects the behavior of the fluid;
the prediction model predicts future behavior of the fluid taking into consideration the influence of the object present in the space on the behavior of the fluid,
When performing machine learning of the prediction model, a plurality of patterns of the entrance and exit of the fluid in the space and the basic shape and position of the object present in the space are set, and for each pattern, data including the physical quantity of the fluid is obtained by numerical simulation or statistical data, and machine learning is performed using the data including the physical quantity of the fluid as the teacher data.
Flow behavior prediction system. The prediction model uses training data in which data obtained by simulating a plurality of states of the space having an arbitrary basic shape is input and output, and the parameters are machine-learned using the training data.
The flow behavior prediction system according to claim 1 . The data acquisition unit acquires data on a physical quantity of the fluid at an arbitrary position in the space.
The flow behavior prediction system according to claim 1 or 2 . The behavior of the fluid includes at least one physical quantity of a velocity and a direction of the fluid;
The data of the first state and the second state further includes physical quantities other than the velocity and direction of the fluid.
The flow behavior prediction system according to any one of claims 1 to 3 . the computer inputs the second state data output from the prediction model into the prediction model as the first state data, and further causes the computer to output the second state data indicating a future behavior of the fluid;
The flow behavior prediction system according to any one of claims 1 to 4 . a display unit that displays visualization information that visualizes the state of the fluid generated by the computer based on the data of the second state output from the prediction model.
The flow behavior prediction system according to any one of claims 1 to 5 . the computer calculates physical quantities of the fluid at each position in the space based on the second state data output from the prediction model, and generates the visualization information showing these physical quantities as at least one of a contour, a vector, and an isosurface.
The flow behavior prediction system according to claim 6 . the computer generates the visualization information in which a first physical quantity, which is a physical quantity of the fluid at an arbitrary position in the space, and a second physical quantity, which is different from the first physical quantity, are displayed while being switched at regular intervals.
The flow behavior prediction system according to claim 6 or 7 . the computer generates the visualization information in such a way that an area of the space where a value of a specific component contained in the fluid exceeds a threshold value, among the physical quantities of the fluid at each position in the space, is highlighted and displayed.
The flow behavior prediction system according to any one of claims 6 to 8 . the computer outputs a warning when a value of a specific component contained in the fluid, among the physical quantities of the fluid at each position in the space, exceeds a threshold value.
The flow behavior prediction system according to any one of claims 1 to 9 . When a value of a specific component contained in the fluid among the physical quantities of the fluid at each position in the space exceeds a threshold value, the computer controls a device for lowering the value.
The flow behavior prediction system according to any one of claims 1 to 10 . a data acquisition unit acquires first state data indicating at least a shape of a space and a behavior of a fluid in the space that changes depending on the shape of the space ;
A prediction model in which parameters are machine-learned using teacher data in which at least one of data of the first state or data simulating the first state is input and at least one of data of a second state or data simulating the second state, which indicates the behavior of the fluid in the future from at least the first state, is output.
inputting the data of the first state acquired by at least the data acquisition unit, and outputting the data of the second state;
1. A computer-implemented method of processing comprising the steps of :
the data acquisition unit measures an inlet/outlet of the fluid in the space and a position of an object present in the space to acquire data indicating a current shape state of the space;
The data on the shape of the space includes data on the shape of the object that exists within the space and affects the behavior of the fluid;
the prediction model predicts future behavior of the fluid taking into consideration the influence of the object present in the space on the behavior of the fluid,
When performing machine learning of the prediction model, a plurality of patterns of the entrance and exit of the fluid in the space and the basic shape and position of the object present in the space are set, and for each pattern, data including the physical quantity of the fluid is obtained by numerical simulation or statistical data, and machine learning is performed using the data including the physical quantity of the fluid as the teacher data.
Methods for predicting flow behavior. A training data set is obtained by inputting at least one of data of a first state, which indicates at least a shape of a space and a behavior of a fluid in the space that changes according to the shape of the space , or data simulating the first state, and outputting at least one of data of a second state, which indicates a behavior of the fluid in the future from the first state, or data simulating the second state;
Performing machine learning of parameters using the training data,
generating a prediction model using at least the first state data acquired by a data acquisition unit as an input and the second state data as an output;
1. A computer-implemented method of processing comprising the steps of :
the data acquisition unit measures an inlet/outlet of the fluid in the space and a position of an object present in the space to acquire data indicating a current shape state of the space;
The data on the shape of the space includes data on the shape of the object that exists within the space and affects the behavior of the fluid;
the prediction model predicts future behavior of the fluid taking into consideration the influence of the object present in the space on the behavior of the fluid,
When performing machine learning of the prediction model, a plurality of patterns of the entrance and exit of the fluid in the space and the basic shape and position of the object present in the space are set, and for each pattern, data including the physical quantity of the fluid is obtained by numerical simulation or statistical data, and machine learning is performed using the data including the physical quantity of the fluid as the teacher data.
Methods for generating predictive models. an input layer to which first state data indicating at least a shape of a space and a behavior of a fluid in the space that changes according to the shape of the space is input;
an output layer that outputs data of a second state that indicates a behavior of the fluid at least in the future relative to the first state;
an intermediate layer in which parameters are machine-learned using teacher data that receives as input at least one of the first state data or data simulating the first state data and outputs at least one of the second state data or data simulating the second state data;
Equipped with
inputting the first state data acquired by at least a data acquisition unit into the input layer, performing an operation in the intermediate layer, and outputting the second state data from the output layer;
A predictive model that causes a computer to function in such a way that
the data acquisition unit measures an inlet/outlet of the fluid in the space and a position of an object present in the space to acquire data indicating a current shape state of the space;
The data on the shape of the space includes data on the shape of the object that exists within the space and affects the behavior of the fluid;
the prediction model predicts future behavior of the fluid taking into consideration the influence of the object present in the space on the behavior of the fluid,
When performing machine learning of the prediction model, a plurality of patterns of the entrance and exit of the fluid in the space and the basic shape and position of the object present in the space are set, and for each pattern, data including the physical quantity of the fluid is obtained by numerical simulation or statistical data, and machine learning is performed using the data including the physical quantity of the fluid as the teacher data.
A trained predictive model.

Images