Environment estimation device, display system, air conditioning system, environment estimation method, program, environment estimation system, and trained model manufacturing method

Abstract

To simply estimate stagnation of air in a target space.SOLUTION: A control unit (14) of an environment estimation device (10) inputs parameters based on a propagation time in a second space (R2) into a trained model (22) that has been trained using, as training data (41), parameters based on the propagation time from when a transmitter (31) transmits a sound wave to a first space (R1) until a receiver (32) receives it and information related to stagnation of air in the first space (R1) corresponding to the parameters; and outputs information related to stagnation of air in the second space (R2) estimated by the trained model (22).SELECTED DRAWING: Figure 8

Description

[Technical Field]

[0001] The present disclosure relates to an environment estimation device, a display system, an air conditioning system, an environment estimation method, a program, an environment estimation system, and a method for manufacturing a trained model. [Background technology]

[0002] Patent Document 1 discloses an air conditioning system. The air conditioning system includes an air conditioner that conditions a target space, a ventilation device that introduces outside air into the target space, a detection means for detecting the environmental state of the target space, and a fluid analysis means for calculating thermal environment distribution information in the target space and an air age distribution, which is a distribution of air ages that is an index of air freshness. The air conditioning system of Patent Document 1 determines control patterns to be set for the air conditioner and the ventilation device based on the analysis results of the fluid analysis means. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] International Publication No. 2023 / 062681 Summary of the Invention [Problem to be solved by the invention]

[0004] In Patent Document 1, the fluid analysis means uses CFD (Computational Fluid Dynamics) to calculate the air age distribution as information related to the air retention in the target space. However, CFD has the following disadvantages. Complex boundary conditions require detailed data on the room shape, including intake and exhaust vents, heat entering through the walls, and internal heat generation. Since the target space is first modeled in 3D and then analyzed, it is difficult to reflect the actual conditions of the target space even if detailed room data is available. - It requires specialized knowledge and advanced calculation skills, and requires a large amount of man-hours.

[0005] The purpose of the present disclosure is to easily estimate the stagnation of air in a target space. [Means for solving the problem]

[0006] The first aspect is directed to an environment estimation device including a control unit (14). The control unit (14) inputs parameters based on the propagation time in a second space (R2) to a trained model (22, 122, 222, 422) that has been trained using, as training data (41), parameters based on the propagation time from when a sound wave emitted by a transmitter (31) to a first space (R1) is received by a receiver (32) and information on the air stagnation in the first space (R1) corresponding to the parameters, and outputs information on the air stagnation in the second space (R2) estimated by the trained model (22, 122, 222, 422).

[0007] In a first aspect, the trained model (22, 122, 222, 422) is trained using parameters based on the propagation time of sound waves emitted to a first space (R1) and information related to the air stagnation in the first space (R1) corresponding to the parameters. By using the trained model (22, 122, 222, 422), information related to the air stagnation in a second space (R2), which is a target space, can be obtained from parameters based on the propagation time of sound waves emitted to the second space (R2). Because the parameters based on the propagation time can be obtained without requiring advanced computing power, the environment estimation device can easily estimate the air stagnation in the target space without performing complex calculations such as CFD.

[0008] In the second aspect, in the first aspect, the control unit (14) inputs information regarding the airflow for each of the multiple sections into which the second space (R2) is divided, which information is calculated based on the propagation time, into the trained model (22, 122, 222, 422).

[0009] In the second aspect, the estimation accuracy of the environment estimation device can be improved by using information about the airflow for each section.

[0010] In a third aspect, in the first or second aspect, the control unit (14) inputs information regarding the temperature of each of the multiple sections into which the second space (R2) is divided, calculated based on the propagation time, into the trained model (22, 122, 222, 422).

[0011] In the third aspect, a temperature difference is likely to occur between the area near the walls and the center of the target space. Even when the target space has a window, a temperature difference is likely to occur between the area near the window and the surrounding area. By using information about the temperature of each section, it is possible to obtain information about the room shape of the target space to a certain extent. This can improve the estimation accuracy of the environment estimation device.

[0012] In a fourth aspect, in any one of the first to third aspects, the control unit (14) inputs to the trained model a first parameter that is the sum of the propagation time of the outbound path and the propagation time of the return path when a sound wave travels back and forth along the same propagation path, and a second parameter that is the difference between the propagation time of the outbound path and the propagation time of the return path.

[0013] In the fourth aspect, since information about temperature can be obtained from the first parameter and information about airflow can be obtained from the second parameter, the estimation accuracy of the environment estimation device can be improved.

[0014] In a fifth aspect, in any one of the first to fourth aspects, the control unit (14) outputs information about the stagnation of air for each of a plurality of sections obtained by dividing the second space (R2).

[0015] In the fifth aspect, it is possible to identify positions in the target space where there is a high probability of stagnation.

[0016] In a sixth aspect, in the fifth aspect, the trained model (22, 122, 222, 422) is a model machine-learned by multi-output learning that outputs the air stagnation for each of the multiple compartments into which the first space (R1) is divided, and the control unit (14) outputs the air stagnation for each of the compartments of the second space (R2).

[0017] In the sixth aspect, by using multi-output learning, estimation can be made taking into account the conditions of not only a specific section but also the surrounding sections. When estimating the air stagnation for each section, the estimation accuracy of the environment estimation device (10) can be improved.

[0018] In a seventh aspect, in any one of the first to sixth aspects, the information relating to the stagnation of air includes an age of air, a remaining life of air, or a lifespan of air.

[0019] In the seventh aspect, the stagnation of air in a target space can be qualitatively evaluated.

[0020] The eighth aspect is directed to a display system (DS), which includes any one of the first to seventh environment estimation devices (10, 310) and a display unit (50) that displays information related to the stagnation of air in the second space (R2).

[0021] In the eighth aspect, the stagnation of air in the target space can be visually grasped.

[0022] A ninth aspect is directed to an air conditioning system (AS). The air conditioning system (AS) includes any one of the first to seventh environment estimation devices (10, 410) and an air conditioner (AC) that conditions the air in the second space (R2), and the control unit (14) controls the air conditioner based on information related to the stagnation of air in the second space (R2).

[0023] In the ninth aspect, information regarding the stagnation of air in the target space can be easily obtained without performing complex calculations such as CFD, and therefore air conditioning can be performed efficiently with as little load as possible.

[0024] A tenth aspect relates to an environment estimation method, which includes the steps of: inputting, into any one of the first to seventh environment estimation devices (10, 310), a parameter based on a propagation time from when a sound wave is transmitted from a transmitter (31) to the second space (R2) until it is received by a receiver (32); and displaying information on the stagnation of air in the second space (R2) output by the environment estimation device (10, 310).

[0025] An eleventh aspect relates to another environment estimation method, which includes the steps of: inputting a parameter based on the propagation time in a second space (R2) into a trained model (22, 122, 222, 422) that has been trained using, as training data (41), a parameter based on the propagation time from when a sound wave emitted by a transmitter (31) to a first space (R1) is received by a receiver (32) and information on the air stagnation in the first space (R1) corresponding to the parameter; and outputting information on the air stagnation in the second space (R2) estimated by the trained model (22, 122, 222, 422).

[0026] A twelfth aspect is directed to a program for causing a computer to execute the environment estimation method of the eleventh aspect.

[0027] A thirteenth aspect relates to an environment estimation system (ES). The environment estimation system (ES) includes a measurement unit (30) that measures a propagation time from when a sound wave is emitted by a transmitter (31) to when it is received by a receiver (32), and a control unit (14), wherein the control unit (14) inputs parameters based on the propagation time measured by the measurement unit (30) in a second space (R2) to a trained model (22, 122, 222, 422) that has been trained using, as training data (41), parameters based on the propagation time measured by the measurement unit (30) in a first space (R1) and information related to the air stagnation in the first space (R1) that corresponds to the parameters, and outputs information related to the air stagnation in the second space (R2) estimated by the trained model (22, 122, 222, 422).

[0028] A fourteenth aspect is the thirteenth aspect, further comprising a server device (120) connected to the measurement unit (30) via a network (N), and the server device (120) includes the control unit (14).

[0029] A fifteenth aspect relates to a method for manufacturing a trained model. The manufacturing method includes the steps of: acquiring a propagation time from when a sound wave is transmitted from a transmitter (31) to a first space until it is received by a receiver (32); acquiring information about the air stagnation in the first space (R1); and performing machine learning using, as training data (41), parameters calculated based on the propagation time and information about the air stagnation in the first space (R1) corresponding to the parameters, to manufacture a trained model (22, 122, 222, 422) that inputs parameters based on the propagation time in a second space (R2) and outputs information about the air stagnation in the second space (R2). [Brief explanation of the drawings]

[0030] [Figure 1] FIG. 1 is a block diagram of an environment estimation system including an environment estimation device according to the first embodiment. [Figure 2] FIG. 2 is a layout diagram of the measurement unit in the first space. [Figure 3] FIG. 3 is a flowchart for calculating the specific parameters. [Figure 4] FIG. 4 is a diagram for explaining the speed of air currents in the x and y directions within a section. [Figure 5] FIG. 5 is a conceptual diagram illustrating a method for producing a trained model. [Figure 6] FIG. 6 is a schematic diagram showing the airflow distribution. [Figure 7] FIG. 7 is a schematic diagram showing the temperature distribution. [Figure 8] Figure 8 is a conceptual diagram showing the use of a trained model. [Figure 9] FIG. 9 is a graph showing the estimation accuracy of a trained model trained using a decision tree model. [Figure 10] FIG. 10 is a graph showing the estimation accuracy of trained models trained by ensemble learning. [Figure 11] FIG. 11 is a block diagram of a display system having an environment estimation device. [Figure 12] FIG. 12 is a flowchart of the environment estimation. [Figure 13] FIG. 13 is an example of the age of air distribution displayed on the display unit. [Figure 14] FIG. 14 is a block diagram showing a first modification of the display system. [Figure 15] FIG. 15 is a block diagram showing a second modification of the display system. [Figure 16] FIG. 16 is a block diagram showing a third modification of the display system. [Figure 17] FIG. 17 is a block diagram of an air conditioning system having an environment estimation device according to the second embodiment. [Figure 18] FIG. 18 is a piping diagram of an air conditioner. [Figure 19] FIG. 19 is a schematic perspective view of an air conditioning apparatus. [Figure 20] FIG. 20 is a block diagram showing a modified example of the air conditioning system. DETAILED DESCRIPTION OF THE INVENTION

[0031] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings. Note that the present disclosure is not limited to the embodiments shown below, and various modifications are possible within the scope of the technical concept of the present disclosure. Since the drawings are intended to conceptually explain the present disclosure, dimensions, ratios, or numbers may be exaggerated or simplified as necessary to facilitate understanding.

[0032] First Embodiment (1) Overall configuration of the environmental estimation system As shown in FIG. 1, an environment estimation system (ES) having an environment estimation device (10) includes a measurement unit (30) and a server device (20). The measurement unit (30) emits sound waves into a target space and measures the time of flight (ToF) of the sound waves. The target space is an indoor space. The environment estimation device (10) outputs information related to the air stagnation in the target space estimated based on the time of flight. The server device (20) is connected to the environment estimation device (10) via a network (N). The server device (20) is used to estimate the air stagnation in the target space. The air stagnation represents stagnation of air. In the first embodiment, the information related to the air stagnation is the air age in the target space. The air age indicates the time that air that has entered the target space remains in the target space. A shorter air age indicates fresher air, while a higher air age indicates stagnation of air. The age of air allows us to understand the state of stagnation in the target space.

[0033] In the first embodiment, information about the stagnation of air in a target space is output using a trained model (22) described later. The trained model (22) is a model that has been machine-learned based on the propagation time of sound waves measured in an indoor space other than the target space and the age of air in the other indoor space. In the following description, the indoor space that is the subject of the training data (41) of the trained model (22) is referred to as a first space (R1), and the indoor space that is the subject of estimation by the environment estimation device (10) is referred to as a second space (R2).

[0034] (1-1) Measuring part The measurement unit 30 includes a transmitter 31, a receiver 32, and a temperature sensor 33. A plurality of measurement units 30 are arranged in one second space R2.

[0035] FIG. 2 shows an example of the arrangement of the measuring units (30) in the second space (R1). The second space (R1) is an actual indoor space, and includes a window (W) and an air conditioner (AC). In this example, four measuring units (30) are arranged in the second space (R2). Each measuring unit (30) is arranged at a corner of the second space (R2). When the second space (R2) is large, the measuring units (30) are also arranged in areas other than the corners.

[0036] Each measurement unit (30) has, for example, a plurality of freestanding poles that respectively support a transmitter (31), a receiver (32), and a temperature sensor (33). Each pole extends vertically. Each pole is provided with an adjustment mechanism for adjusting the height of the transmitter (31), etc. It is preferable that the heights of the transmitters (31), etc. of each measurement unit (30) are set to the same height position by the adjustment mechanism. Each pole is movable. An operator transports each pole to the second space (R2) and installs it in the second space (R2).

[0037] The transmitter 31 emits sound waves into the second space R2. The transmitter 31 emits sound waves within an angular range that allows the sound waves to be received by the receivers 32 of all other measurement units 30. For example, the transmitter 31 generates sound waves over a range of approximately 90° in a planar view. The transmitter 31 may be a directional transmitter or an omnidirectional transmitter.

[0038] The receiver 32 receives the sound waves transmitted from the transmitter 31. The receiver 32 may be a directional receiver or an omnidirectional receiver.

[0039] The sound waves received by the receiver 32 include direct waves and reflected waves. Direct waves are sound waves that are emitted from the transmitter 31 and reach the receiver 32 without colliding with the walls of the second space R2. Reflected waves are sound waves that are emitted from the transmitter 31 and then reflect off the walls of the second space R2 before reaching the receiver 32.

[0040] The temperature sensor (33) measures the temperature around the measuring part (30) in the second space (R2). The temperature sensor (33) is, for example, an infrared sensor.

[0041] (1-2)Environmental estimation device The environment estimation device (10) is a terminal such as a smartphone or a PC (Personal Computer). In the first embodiment, the environment estimation device (10) is a mobile terminal such as a smartphone or a tablet PC. In the first embodiment, the worker carries the environment estimation device (10) and goes to the second space (R2), and operates the environment estimation device (10) to check the age of air in the second space (R2). The environment estimation device (10) may be a stationary terminal such as a desktop PC.

[0042] The environment estimation device (10) includes a first communication unit (11), an operation unit (12), a storage unit (13), and a control unit (14).

[0043] The first communication unit 11 communicates with the measurement unit 30 and the server device 20 via wired or wireless communication. The first communication unit 11 includes a communication module (communication device) such as a LAN board, and communicates wirelessly with the server device 20 via a network N such as the Internet.

[0044] The operation unit 12 externally operates the environment estimation device 10. The operation unit 12 is composed of input devices such as a touch panel, a keyboard, and a mouse.

[0045] The storage unit (13) is composed of a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk drive, or a combination thereof. The storage unit (13) stores sound wave path information. The path information includes multiple propagation paths and the propagation path lengths corresponding to these propagation paths.

[0046] The storage unit (13) stores various first programs (131) executed by the control unit (14).

[0047] The control unit 14 includes a processor such as a CPU (Central Processing Unit) and an MPU (Micro Processing Unit), an electric circuit, and an electronic circuit. The control unit 14 reads and executes a first program 131 to control each component of the environment estimation device 10.

[0048] The control unit (14) calculates various specific parameters for estimating the stagnation of air in the second space (R2) from the propagation time of the sound wave measured by the measurement unit (30). In the first embodiment, the control unit (14) calculates, as specific parameters, the airflow distribution in the second space (R2), the temperature distribution in the second space (R2), a first parameter which is the sum of the propagation time on the outbound path and the propagation time on the return path when the sound wave travels back and forth along the same propagation path, and a second parameter which is the difference between the propagation time on the outbound path and the propagation time on the return path. Note that the airflow distribution refers to the airflow in each of the multiple sections into which the second space (R2) is divided, and the temperature distribution refers to the temperature in each section. The airflow is not only a scalar such as wind speed, but also a vector including the direction of flow.

[0049] The control unit (14) inputs the calculated specific parameter and the temperature detected by the temperature sensor (33) to the server device (20) via the first communication unit (11).

[0050] (1-3) Server device The server device 20 includes a second communication unit 21 and a trained model 22. The trained model 22 is stored as a program in a storage unit of the server device 20.

[0051] The second communication unit 21 communicates with the environment estimation device 10 via wired or wireless communication. The second communication unit 21 includes a communication module (communication device) such as a LAN board, and communicates wirelessly with the environment estimation device 10 via the network N.

[0052] The trained model (22) is a model for estimating the air age distribution in the target space from the specific parameters and the detected temperature. The specific parameters and the detected temperature are input to the trained model (22) from the control unit (14) via the network (N). The trained model (22) returns the estimated air age distribution to the control unit (14) via the network (N). The air age distribution means the air age for each of the multiple sections into which the second space (R2) is divided.

[0053] (2) Calculation of specific parameters The method of calculating the specific parameter by the control unit (14) will be described with reference to FIGS.

[0054] As shown in FIG. 3, in step ST101, if the initial setting of the control unit (14) is not completed, the initial setting of steps ST102 to ST104 is executed.

[0055] In step ST102, the control unit 14 acquires three-dimensional coordinates relating to the position of each measurement unit 30. The storage unit 13 stores the three-dimensional coordinates measured in step ST102 as coordinate data.

[0056] In step ST103, the control unit (14) divides the second space (R2) into a plurality of sections (A n ) (n=1, 2, ... n). In FIG. 2, the second space (R2) is divided into 12 sections (A1 to A 12 In addition, in step ST103, the control unit (14) divides the sound wave propagation path (P m ) (m=1, 2, ... m). The control unit (14) determines the n ) over one or more propagation paths (P m ) through multiple propagation paths (P m The memory unit (13) determines the divided sections (A n ) and the three-dimensional coordinates of the determined propagation paths (P m ) are stored as coordinate data. 12 ) are divided into Section 1 (A1) to Section 12 (A 12 ) are sometimes explained as being divided into 12 sections (A1 to A 12 When there is no need to distinguish between the compartments (A n )

[0057] In FIG. 2, each measurement section (30) is made up of a first section (A1), a fourth section (A4), a ninth section (A9), and a twelfth section (A 12) are placed in the second space (R2). In addition, in the second space (R2), the window (W) is placed in the 11th section (A 11 ) and Section 12 (A 12 ), and the air conditioning unit (AC) is located in the ninth section (A9). 10 ) faces the wall.

[0058] In Figure 2, there are 10 propagation paths (P1 to P 10 ) These propagation paths (P1 to P 10 ), the first propagation path (P1), the third propagation path (P3), the fifth propagation path (P5), the seventh propagation path (P7), the eighth propagation path (P8), and the tenth propagation path (P 10 ) are the propagation paths of the direct wave. The second propagation path (P2), fourth propagation path (P4), sixth propagation path (P6), and ninth propagation path (P9) are the propagation paths of the indirect wave reflected by the wall surface.

[0059] In step ST104, the control unit (14) determines the propagation path (P m ) in the section (A n ) propagation path length (D m,n ) is calculated. m,n ) is the propagation path (P m ) the whole area (A n ) is the length of the path divided into each of the sections. In FIG. 2, the propagation path length of the first propagation path (P1) is illustrated. The first propagation path (P1) passes through the first section (A1), the fifth section (A5), and the ninth section (A9). In this case, the propagation path length corresponding to the first propagation path (P1) is D 1,1 , D 1,5 , and D 1,9 The total length of the first propagation path (P1) is D 1,1 +D 1,5 +D 1,9 becomes.

[0060] In step ST104, the control unit (14) calculates the propagation path length (D m,nIn addition, the control unit (14) calculates a plurality of propagation paths (P m ) to each receiver (32). In step ST104, the storage unit (13) calculates the incident angle θ of the sound wave incident on each receiver (32) from the plurality of propagation paths (P m ) and their propagation paths (P m ) and the corresponding incident angles and propagation paths (P m ) corresponding to the propagation path length (D m,n ) and store it as initial data.

[0061] In step ST105, a first process is executed. The first process is executed by m The first process is a process for measuring the propagation time of a sound wave from the transmitter (31) to the receiver (32). Here, the propagation time is the time from when the transmitter (31) transmits the sound wave until the sound wave is received by the receiver (32). The details of the first process will be described later.

[0062] In step ST106, the control unit (14) calculates the propagation time for each propagation path (P m ) to calculate the first parameter.

[0063] In step ST107, the control unit (14) calculates the temperature distribution in the second space (R2), i.e., the temperature distribution in each section (A n ) temperature is calculated.

[0064] In step ST108, the control unit (14) calculates the propagation time for each propagation path (P m ) to calculate the second parameter.

[0065] In step ST109, the control unit (14) calculates the airflow distribution in the second space (R2), i.e., the airflow distribution in each section (A n ) airflow.

[0066] Next, the relationship between the first parameter and the temperature distribution, and the relationship between the second parameter and the airflow distribution will be described.

[0067] The propagation velocity of sound waves (C) can be expressed by the following equation (A).

[0068]

number

[0069] Here, C0 is the speed of sound (331.5 [m / sec]) when the temperature is 0°C, t is the temperature, and α is a coefficient that is a predetermined constant.

[0070] Propagation time T of each measurement sound wave + is expressed by the following relational expression using the propagation path length D and wind speed V. As shown in Figure 4, V x cosθ represents the velocity of the airflow in the x direction, and V y sinθ represents the airflow velocity in the y direction.

[0071]

number

[0072] Return propagation time T of each measurement sound wave - is expressed by the following relational expression using the propagation path length D and wind speed V.

[0073]

number

[0074] A formula for calculating the temperature can be obtained based on the first parameter, which is the sum of the propagation time on the outward path and the propagation time on the return path. Specifically, the following relational expression is established based on formulas (A) to (C).

[0075]

number

[0076] Based on formula (D), each section (A1 to A 12) is created to calculate the temperature. Specifically, the propagation path of the sound wave (P1 to P 10 ) for the six detection sound waves passing through the 10 ) and each section (A1~A 12 ) Propagation path length D m,n (m=1,2,…,10,n=1,2,…,12) and the speed of sound C n For (n=1,2,…,12), the following determinant (5) holds.

[0077]

number

[0078] The simultaneous equations obtained from the determinant of this equation (E) are solved for each section (A1 to A 12 ) temperature t n (n=1, 2, ..., 12). The control unit (14) can calculate, for example, the first section (A1), the fourth section (A4), the ninth section (A9), and the twelfth section (A 12 ) is the detected temperature, and for the remaining sections, the temperature (t n ) may be adopted.

[0079] A formula for calculating the airflow can be obtained based on the second parameter, which is the difference between the propagation time on the outbound path and the propagation time on the return path. Specifically, the following relational expression is established based on formulas (A) to (C).

[0080]

number

[0081] Based on formula (F), each section (A1 to A 12 ) is created. Specifically, the determinant for the airflow along the propagation path (P1 to P 10 ) for the six detection sound waves passing through the 10 ) and each section (A1~A 12 ) Propagation path length D m,n(m=1,2,…,10,n=1,2,…,12) and the speed of sound C xn , C yn (n=1,2,…,12) The following determinant (G) holds:

[0082]

number

[0083] The temperature of each region (t n ) from equation (E) and calculate the sound speed C, and then each section (A1 to A 12 ) wind speed V n (n=1,2,...,12) can be calculated.

[0084] (3) How to manufacture trained models The trained model (22) is a model that has undergone machine learning using, as training data (41), a data set of parameters based on the propagation time from when a sound wave is transmitted by a transmitter (31) to when it is received by a receiver (32), a detected temperature detected by a temperature sensor (33), and an age-of-air distribution in the first space (R1) corresponding to the parameters and the detected temperature. The parameters and the detected temperature correspond to explanatory variables, and the age-of-air corresponds to a target variable.

[0085] As shown in FIG. 5, the parameters are calculated by the calculation device (40) based on the propagation time measured by the measurement unit (30). The parameters are the same as the specific parameters described above, namely, the airflow distribution in the first space (R1), the temperature distribution in the first space (R1), the first parameter, and the second parameter. The calculation device (40) may be the environment estimation device (10). The calculation device (40) calculates the specific parameters according to the flowchart of FIG. 3, similar to the control unit (14). In the first embodiment, the air age distribution is a binarized distribution indicating a region where the air age is equal to or greater than a predetermined value and a region where the air age is less than the predetermined value. The air age distribution is estimated separately, for example, by CFD. Note that the measurement unit (30) used to calculate the training data (41) and the measurement unit (30) used to estimate the air age distribution in the second space (R2) do not need to be exactly the same as long as they have the same configuration.

[0086] The training data (41) is sent to a learning device (43). The learning device (43) performs machine learning using the training data (41) to produce a trained model (22). The learning device (43) divides the first space (R1) into a plurality of sections (A n The machine learning is performed by multi-output learning using the age of air for each time period (times) as a response variable. The produced trained model (22) is stored in the server device (20). The learning device (43) may be the server device (20).

[0087] Fig. 6 shows an example of the airflow distribution calculated by the calculation device (40). In Fig. 6, it is assumed that the first space (R1) has the same configuration as in Fig. 2. In Fig. 6, for convenience, the direction of the airflow is indicated by the direction of the arrow, and the magnitude of the wind speed is indicated by the size of the arrow. In practice, the wind speed is calculated numerically.

[0088] As shown in FIG. 6, the fifth section (A5), the ninth section (A9), and the eleventh section (A 11 ), and Section 12 (A 12 The wind speed in the fifth section (A5) and the ninth section (A9) is due to the influence of the air conditioning unit (AC), and the wind speed in the eleventh section (A 11) and Section 12 (A 12 ) is the effect of introducing outside air through the window (W).

[0089] It can be seen that the second section (A2), the third section (A3), the fourth section (A4), the seventh section (A7), and the eighth section (A8) have lower wind speeds than the other sections. A lower wind speed means that air tends to stagnate. In other words, the airflow distribution can be used to estimate which sections are likely to have a high air age.

[0090] FIG. 7 shows an example of the temperature distribution calculated by the calculation device (40). In FIG. 7, it is assumed that the first space (R1) has the same configuration as in FIG. 2. In FIG. 7, for convenience, each section (A n ) is shown in four levels. Level 1 is the lowest temperature, followed by levels 2, 3, and 4. In reality, the temperature can be calculated numerically.

[0091] As shown in Figure 7, the 11th section (A 11 ) and Section 12 (A 12 ) has a low temperature. This is because the temperature in the 11th compartment (A 11 ) and Section 12 (A 12 The first compartment (A1), the second compartment (A2), the fifth compartment (A5), and the ninth compartment (A9) have relatively low temperatures because they include walls.

[0092] As mentioned above, the temperature distribution reflects the positions of windows and walls in the first space (R1). By using the temperature distribution, it is possible to estimate the room shape, including the position of furniture, to some extent, even if no room shape data is available.

[0093] Furthermore, the fourth compartment (A4) has a higher temperature than the other compartments despite not having a heat source. In locations where air tends to stagnate, the temperature difference with the surrounding compartments tends to be large. In other words, it can be assumed that compartments such as the fourth compartment (A4), which have a higher temperature than the surrounding compartments, have a higher age of air.

[0094] The server device (20) performs machine learning using the airflow distribution of the entire first space (R1) shown in FIG. 6 and the temperature distribution of the entire first space (R1) shown in FIG. 7 as explanatory variables. For example, the server device (20) does not associate only the airflow and temperature of the sixth section (A6) with the age of air of the sixth section (A6), but n ) and learn how the airflow and temperature in the sixth section (A6) correspond to the age of air in the sixth section (A6).

[0095] The server device (20) performs machine learning to output a binarized air age distribution indicating whether the air age is greater than or equal to a predetermined value or less than a predetermined value. The machine learning method used by the server device (20) may be a neural network, a support vector machine, a decision tree, gradient boosting, bagging, a random forest, linear regression, ridge regression, or Lasso regression. The server device (20) may perform machine learning by combining the above-mentioned machine learning methods.

[0096] (4) Use of pre-trained models 8, when the environment estimation system (ES) estimates the age-of-air distribution in the second space (R2), the measurement unit (30) inputs the propagation times and detected temperatures measured in the second space (R2) to the control unit (14). The control unit (14) calculates the airflow distribution in the second space (R2), the temperature distribution in the second space (R2), the first parameter, and the second parameter from the propagation times.

[0097] The control unit (14) inputs explanatory variables including the airflow distribution, the temperature distribution, the first parameter, the second parameter, and the detected temperature to the trained model (22). The trained model (22) outputs an age-of-air distribution estimated from the explanatory variables as a response variable. The output age-of-air distribution is transmitted to the control unit (14). Thereafter, the control unit (14) outputs the estimated age-of-air distribution.

[0098] (4-1) Estimation accuracy of trained model The estimation accuracy of the trained model (22) varies depending on the explanatory variables included in the training data (41). Figures 9 and 10 show the results of verifying the change in estimation accuracy due to differences in explanatory variables. Figure 9 shows the estimation accuracy of a trained model created using only decision trees. Figure 10 shows the estimation accuracy of a trained model created by ensemble learning that combines multiple models.

[0099] 9 and 10, Model 1, Model 2, Model 3, Model 4, and Model 5 have different explanatory variables included in the training data (41). Specifically, Model 1: Airflow distribution, temperature distribution, first parameter, second parameter, and detected temperature Model 2: Airflow Distribution Model 3: First and second parameters Model 4: Airflow and temperature distribution Model 5: Airflow distribution, temperature distribution, first parameter, and second parameter Model 1 corresponds to the trained model (22) in the first embodiment.

[0100] As shown in Figure 9, when any of Models 1 to 5 is used, the estimation accuracy is 60% or more. Specifically, Model 2 has the lowest estimation accuracy of the five models, at about 60%. Models 3 to 5 have similar estimation accuracy of about 80%. Model 1 has the highest estimation accuracy of the five models, at about 83%.

[0101] Models 1, 4, and 5 use temperature distribution as an explanatory variable in addition to airflow distribution. As mentioned above, temperature distribution is affected by the room shape, so when temperature distribution is used, the estimation results reflect the room shape to some extent. Furthermore, since both airflow distribution and temperature distribution provide information that can be used to estimate compartments where air is likely to stagnate, using both airflow distribution and temperature distribution improves the accuracy of estimating the age of air.

[0102] Model 3 uses only the first parameter and the second parameter. As mentioned above, the first parameter is a parameter related to the temperature distribution, and the second parameter is a parameter related to the airflow distribution. Learning using both the first parameter and the second parameter as explanatory variables can be said to be learning while taking into account both the airflow distribution and the temperature distribution.

[0103] Therefore, although a certain degree of accuracy in estimating the age of air can be achieved by simply using parameters related to airflow distribution as explanatory variables, adding parameters related to temperature distribution as explanatory variables can provide even higher accuracy in estimating the age of air. Furthermore, as in Model 1, the estimation accuracy can be improved by using not only the airflow distribution and temperature distribution but also the first parameter, second parameter, and detected temperature, which are the original data used to calculate them, as explanatory variables.

[0104] As shown in Figure 10, when a trained model is created using ensemble learning that combines multiple models, the estimation accuracy of Model 1 increases to about 85% compared to when a trained model is created using only decision trees. The estimation accuracy of Models 3 to 5 also increases to about 85%. The estimation accuracy of Model 2 also increases slightly.

[0105] Comparing Figure 9 and Figure 10, we can see that estimation accuracy differs depending on the learning method. By creating a trained model through ensemble learning that combines multiple models, the estimation accuracy of the trained model can be improved compared to when a single model is used.

[0106] (5) Display system FIG. 11 shows the configuration of a display system (DS) equipped with an environment estimation device (10).

[0107] The display system (DS) includes an environment estimation device (10), a server device (20), and a display unit 50. The environment estimation device (10) and the server device (20) are the same as the environment estimation device (10) and the server device (20) in the above-described environment estimation system, and therefore detailed description thereof will be omitted.

[0108] The display unit 50 displays the age-of-air distribution in the second space R2 output by the environment estimation device 10. The display unit 50 includes, for example, a liquid crystal panel and displays an image. The display unit 50 may function as a touch panel.

[0109] Next, the operation of the display system (DS) to display the age-of-air distribution output by the environment estimation device (10) will be described.

[0110] 12, in step ST201, the control unit 14 acquires the propagation time and the detected temperature. The control unit 14 acquires the propagation time and the detected temperature in the same manner as in steps ST101 to ST105 described above.

[0111] In step ST202, the control unit (14) calculates specific parameters: the airflow distribution in the second space (R2), the temperature distribution in the second space (R2), the first parameter, and the second parameter.

[0112] In step ST203, the control unit (14) inputs the specific parameters calculated in step ST202 and the detected temperature into the trained model (22).

[0113] In step ST204, the control unit (14) acquires the age-of-air distribution of the second space (R2) output by the trained model.

[0114] In step ST205, the control unit (14) causes the display unit (50) to display the acquired age-of-air distribution.

[0115] FIG. 13 is an example of the age of air distribution displayed on the display unit (50). In the second space (R2) shown in FIG. 13, the ninth section (B9) and the tenth section (B 10 ) has a window (W) and the 12th section (B 12 ) are equipped with air conditioning units (AC). The first section (B1) to the fifth section (B5), the eighth section (B8), and the eleventh section (B 11 ), and Section 12 (A 12 ) faces the wall.

[0116] The trained model (22) of the first embodiment outputs a binarized air age distribution. The black hatched areas are areas where the air age is estimated to be higher than a predetermined value. In this example, the first section (B1), the second section (B3), the fifth section (B5), and the eleventh section (B 11 ) is estimated to have an age of air higher than a predetermined value. By displaying the age of air distribution on the display unit (50), the age of air distribution can be visually grasped and can be used as an index for performing air conditioning such as ventilation.

[0117] (6) Effects of the First Embodiment In the first embodiment, the control unit (14) inputs parameters based on the propagation time in the second space (R2) into a trained model (22) that has been trained using training data (41) including parameters based on the propagation time from when a sound wave emitted from a transmitter (31) to the first space (R1) is received by a receiver (32) and an air age distribution in the first space (R1) corresponding to the parameters, and outputs the air age distribution in the second space (R2) estimated by the trained model (22). By using the trained model (22), the air age distribution in the second space (R2) can be obtained from parameters based on the propagation time of the sound wave emitted to the second space (R2). The parameters based on the propagation time can be obtained without requiring advanced computing power. The environment estimation device (10) can easily estimate the air age distribution in the second space (R2) without performing complex calculations such as CFD. Furthermore, since the air age distribution can be estimated with a smaller load than CFD, it is possible to obtain the air age distribution in real time.

[0118] In the first embodiment, the control unit (14) inputs information about the airflow for each of the multiple sections into which the second space (R2) is divided, calculated based on the propagation time, into the trained model (22). Air tends to stagnate in sections with a lower wind speed compared to the surrounding sections. Furthermore, when the airflow forms a vortex, it can be estimated that air is stagnating even if the wind speed is relatively high. Therefore, by using the information about the airflow for each section, the estimation accuracy of the environment estimation device (10) can be improved.

[0119] In the first embodiment, the control unit (14) inputs information about the temperature of each of the compartments into which the second space (R2) is divided, calculated based on the propagation time, to the trained model (22). The temperature of each compartment is affected by objects contained in the compartment, such as windows, furniture, and walls. By using the temperature of each compartment, it is possible to estimate the room shape, including the position of furniture, to some extent, even without room shape data. In other words, the information about the temperature of each compartment can serve as information to replace room shape data. Furthermore, since the temperature difference between the compartment and the surrounding compartments is likely to be large in locations where air is likely to stagnate, the air age distribution can also be estimated from the information about the temperature of each compartment. Therefore, by using the information about the temperature of each compartment, it is possible to improve the estimation accuracy of the environment estimation device (10).

[0120] In the first embodiment, the control unit (14) inputs a first parameter, which is the sum of the propagation time on the outbound path and the propagation time on the return path when a sound wave travels back and forth along the same propagation path, and a second parameter, which is the difference between the propagation time on the outbound path and the propagation time on the return path, into the trained model (22). The first parameter is a parameter related to temperature, and the second parameter is a parameter related to airflow. Using the first parameter means taking temperature into consideration, and using the second parameter means taking airflow into consideration. The first parameter and the second parameter are information closer to the raw value of the propagation time than wind speed or temperature, and therefore can be said to be parameters that more accurately reflect the actual air condition. Therefore, using the first parameter and the second parameter can improve the estimation accuracy of the environment estimation device (10).

[0121] Furthermore, by using the first parameter and the second parameter in addition to the airflow distribution and the temperature distribution, the estimation accuracy of the environment estimation device (10) can be improved.

[0122] In the first embodiment, the control unit (14) outputs the age of air distribution for each of the sections obtained by dividing the second space (R2) into a plurality of sections, thereby making it possible to identify positions in the second space (R2) with high stagnation.

[0123] In the first embodiment, the trained model (22) is a model trained by machine learning through multi-output learning that outputs the age of air for each of the sections obtained by dividing the first space (R1), and the control unit (14) outputs the age of air for each section of the second space (R2). Multi-output learning enables estimation that takes into account the conditions of not only a specific section but also the surrounding sections. When estimating the age of air for each section, the estimation accuracy of the environment estimation device (10) can be improved.

[0124] In the first embodiment, the age of air distribution in the second space (R2) is displayed on the display unit (50). The age of air distribution output by the environment estimation device (10) can be visually grasped by the display unit (50).

[0125] (7) Modification of the first embodiment 14, in the display system (DS), a network may be interposed between the environment estimation device (10), the measurement unit (30), and the display unit (50). The environment estimation device (10) may be realized by cloud computing.

[0126] As shown in FIG. 15, in the display system (DS), the control unit (114) of the environment estimation device may be integrated with the server device (120). In this case, the server device (120) functions as the environment estimation device. A program (115) for the control unit (114) to execute processing is stored in a memory unit (113) of the server device (120). A trained model (122) is stored in the memory unit (113).

[0127] As shown in FIG. 16, in the display system (DS), the trained model (222) may be stored in a storage unit (213) of the environment estimation device (210).

[0128] In the display system (DS), the display unit (50) may be an integrated device with the environment estimation device (10).

[0129] Second Embodiment The embodiments of the present disclosure will be described in detail with reference to the drawings. In the following description, parts common to the first embodiment will be denoted by the same reference numerals, and detailed description thereof will be omitted.

[0130] (8) Overall structure of the air conditioning system As shown in Fig. 17, an air conditioning system (AS) having an environment estimation device (10) includes a measurement unit (30), a server device (20), and an air conditioner (AC). The environment estimation device (10) is connected to the measurement unit (30), the server device (20), and the air conditioner (AC) via a network (N) such as the Internet. The configurations of the measurement unit (30), the environment estimation device (10), and the server device (20) are the same as those in the first embodiment, and therefore description thereof will be omitted.

[0131] (8-1) Air conditioning equipment FIG. 18 is a schematic piping diagram of the air conditioner (AC), and FIG. 19 is a schematic perspective view of the air conditioner (AC).

[0132] The air conditioner (AC) adjusts the temperature of the air in the second space (R2). The air conditioner (AC) performs cooling and heating operations. In cooling operation, the air conditioner (AC) cools the air in the second space (R2). In heating operation, the air conditioner (AC) heats the air in the second space (R2).

[0133] The air conditioner (AC) includes a refrigerant circuit (301). The refrigerant circuit (301) is filled with refrigerant. The refrigerant circuit (301) performs a refrigeration cycle by circulating the refrigerant.

[0134] The air conditioner (AC) includes an outdoor unit (320), an indoor unit (330), a first connecting pipe (302), a second connecting pipe (303), and a remote controller (RC). The air conditioner (AC) is a pair type having one outdoor unit (320) and one indoor unit (230). The first connecting pipe (302) is a gas connecting pipe, and the second connecting pipe (303) is a liquid connecting pipe.

[0135] The outdoor unit (320) is installed outdoors. The outdoor unit (320) includes an outdoor casing (320a) and outdoor elements housed in the outdoor casing (320a). The outdoor elements include a compressor (321), an outdoor heat exchanger (322), an expansion valve (323), a four-way selector valve (324), and an outdoor fan (325).

[0136] The compressor (321) is a rotary compressor such as a swing piston type, a rotary type, or a scroll type. The outdoor heat exchanger (322) is a fin-and-tube type. The four-way selector valve (324) is switchable between a first state (a state indicated by a solid line in FIG. 1 ) and a second state (a state indicated by a dashed line in FIG. 1 ). In the first state, the four-way selector valve (324) communicates the discharge port of the compressor (321) with the gas end of the outdoor heat exchanger (322) and also communicates the suction port of the compressor (321) with the first connecting pipe (302). In the second state, the four-way selector valve (324) communicates the discharge port of the compressor (321) with the first connecting pipe (302) and also communicates the suction port of the compressor (321) with the gas end of the outdoor heat exchanger (322). The outdoor fan (325) is a propeller fan.

[0137] The outdoor unit (320) is controlled by an outdoor control unit (OC).

[0138] The indoor unit (330) includes an indoor casing (330a) and indoor elements housed in the indoor casing (330a). The indoor elements include an indoor heat exchanger (331), an indoor fan (332), and a flap (333). The indoor heat exchanger (331) is a fin-and-tube type. The indoor fan (332) is a cross-flow fan. The flap (333) adjusts the direction of airflow into the second space (R2).

[0139] The indoor unit (330) is controlled by an indoor control unit (IC).

[0140] The remote controller (RC) has a display unit (341) and an operation unit (342). The display unit (341) is an example of a notification unit of the present disclosure. The display unit (341) is a display such as an LCD screen. The operation unit (342) is composed of buttons operated by a user. The display unit (341) and the operation unit (342) may be combined into a touch panel.

[0141] The air age distribution estimated by the trained model (22) is input to the environment estimation device (10) via the second communication unit (21). The control unit (14) inputs the estimated air age distribution to the outdoor control unit (OC), the indoor control unit (IC), or the remote controller (RC).

[0142] The air conditioner (AC) performs air conditioning of the second space (R2) based on the input air age distribution. For example, the air volume from the indoor fan (332) and the direction of the flap (333) are adjusted so that air is blown to an area with a high air age.

[0143] (9) Effects of the Second Embodiment In the second embodiment, the air conditioning of the second space (R2) is performed by the air conditioner (AC) based on the air age distribution of the second space (R2) output from the control unit (14). When performing air conditioning based on the air age distribution, it is necessary to estimate the air age distribution as quickly as possible. If complex calculations such as CFD are performed, it takes a long time to calculate the air age distribution, and there is a risk that the air age distribution will have changed by the time air conditioning is actually performed. As in the second embodiment, if the trained model (22) is used to calculate the air age distribution from parameters based on the propagation time of sound waves, the air age distribution of the second space (R2) can be easily obtained with as little load as possible. The air conditioning system (AS) can perform air conditioning efficiently.

[0144] (10) Modification of the second embodiment As shown in Fig. 20, in the air conditioning system (AS), the environment estimation device (410) may be an integrated device with the air conditioning device (AC). In this case, the trained model (422) is stored in the storage unit (413).

[0145] The environment estimation device (410) may be mounted on any of the outdoor unit (320), the indoor unit (330), and the remote controller (RC), or may be configured as a combination of these. When the environment estimation device (410) is mounted on the outdoor unit (320), the control unit (414) is configured as an outdoor control unit (OC). Similarly, when the environment estimation device (410) is mounted on the indoor unit (330), the control unit (414) is configured as an indoor control unit (IC).

[0146] (11) Other embodiments The control unit (14) may calculate the airflow distribution and the temperature distribution using the room shape data.

[0147] The trained model (22, 122, 222, 422) may be a model to which the CO2 concentration in the second space (R2) is further added as an explanatory variable. The CO2 concentration in the second space (R2) may be measured, for example, by installing a CO2 concentration sensor in the measurement unit (30).

[0148] The trained model (22, 122, 222, 422) used was the aforementioned Model 1. The trained models (22, 122, 222, 422) may also use Models 2 to 5. Model 2, which estimates only from the airflow distribution, has lower estimation accuracy than the other models, but is more useful than using CFD in that it only requires measuring the propagation time of sound waves in the second space (R2), it can obtain estimation results that reflect the actual shape of the second space (R2), and it can obtain information on air retention in real time.

[0149] The trained model (22, 122, 222, 422) may output information about air retention, such as air lifetime and remaining air life. The air lifetime is the time from when outside air is introduced into the second space (R2) until it is exhausted. The remaining air life is the value obtained by subtracting the air age from the air lifetime.

[0150] The trained model (22, 122, 222, 422) may be a model trained by machine learning to output the age of air as a numerical value. In this case, the control unit (14) may control the display unit (50) to display the age of air distribution in the second space (R2) as a contour diagram. In addition, the trained model (22, 122, 222, 422) may be a model trained to output the difference from the standard deviation of the calculated age of air, skewness, kurtosis, etc.

[0151] The learning method for producing the trained model (22, 122, 222, 422) does not have to be ensemble learning. As shown in Figure 9, even if a simple decision tree is used as the learning method, a certain degree of estimation accuracy can be obtained.

[0152] If the trained model (22, 122, 222, 422) is not produced by multi-output learning, the trained model (22) may be a model that estimates the age of air in a specific compartment from the airflow and temperature in the specific compartment. In this case, the control unit (14) may perform processing so that, after receiving the age of air in each compartment estimated by the trained model (22, 122, 222, 422), the display unit (50) displays the air age distribution in the entire second space (R2).

[0153] The first space (R1) and the second space (R2) may be the same space. That is, the trained model (22, 122, 222, 422) may be a model produced by performing real-time learning using parameters based on the propagation time in the second space (R2) and information indicating the stagnation of air in the second space (R2). When performing real-time learning, the calculation device (40) shown in FIG. 3 is configured as a control unit (14).

[0154] The measuring unit 30 does not have to be portable, and may be embedded in a wall or ceiling of an indoor space.

[0155] Although the embodiments and modifications have been described above, it will be understood that various modifications in form and details are possible without departing from the spirit and scope of the claims. Furthermore, the above embodiments, modifications, and other embodiments may be combined or substituted as appropriate as long as the functionality of the subject matter of the present disclosure is not impaired.

[0156] The terms "first," "second," "third," etc. mentioned above are used to distinguish the terms to which these terms are attached, and do not limit the number or order of the terms. [Industrial Applicability]

[0157] As described above, the present disclosure is useful for an environment estimation device, a display system, an air conditioning system, an environment estimation method, a program, an environment estimation system, and a method for manufacturing a trained model. [Explanation of symbols]

[0158] 10 Environment estimation device 14 Control Unit 20 Server device 22 Pre-trained models 30 Measuring part 31 Transmitter 32 Receiver 41 Teacher data 50 Display section 120 Server device 122 trained models 222 trained models 310 Environment estimation device 410 Environment estimation device 422 trained models AC air conditioning unit AS Air Conditioning System ES Environmental Estimation System DS Display System

Claims

1. An environment estimation device including a control unit (14), The control unit (14) inputting a parameter based on the propagation time in the second space (R2) into a trained model (22,122,222,422) that has been trained using, as training data (41), a parameter based on the propagation time from when a sound wave is transmitted from a transmitter (31) to a first space until it is received by a receiver (32) and information on the stagnation of air in the first space (R1) corresponding to the parameter; An environment estimation device that outputs information regarding the air stagnation in the second space (R2) estimated by the trained model (22,122,222,422).

2. The environment estimation device according to claim 1 , The control unit (14) is an environment estimation device that inputs information about airflow for each of multiple sections into which the second space (R2) is divided, calculated based on the propagation time, into the trained model (22, 122, 222, 422).

3. The environment estimation device according to claim 2, The control unit (14) is an environmental estimation device that inputs information regarding the temperature of each of the multiple sections into which the second space (R2) is divided, calculated based on the propagation time, into the trained model (22, 122, 222, 422).

4. The environment estimation device according to claim 1 , The control unit (14) is an environmental estimation device that inputs a first parameter, which is the sum of the propagation time of the outbound path and the propagation time of the return path when a sound wave travels back and forth along the same propagation path, and a second parameter, which is the difference between the propagation time of the outbound path and the propagation time of the return path, into the trained model (22,122,222,422).

5. The environment estimation device according to any one of claims 1 to 4, The control unit (14) is an environment estimation device that outputs information about the stagnation of air for each of a plurality of sections obtained by dividing the second space (R2).

6. The environment estimation device according to claim 5, The trained model (22, 122, 222, 422) is a model machine-learned by multi-output learning, which outputs the air retention for each of the multiple sections obtained by dividing the first space (R1), The control unit (14) is an environment estimation device that outputs the air stagnation for each of the compartments of the second space (R2).

7. The environment estimation device according to any one of claims 1 to 4, The information about the air stagnation includes air age, air remaining life, or air lifespan.

8. An environment estimation device (10, 310) according to any one of claims 1 to 4; and a display unit (50) that displays information about the stagnation of air in the second space (R2).

9. An environment estimation device (10, 410) according to any one of claims 1 to 4; an air conditioning device (AC) that conditions the air in the second space (R2); The control unit (14) controls the air conditioner based on information relating to the stagnation of air in the second space (R2).

10. a step of inputting a parameter based on a propagation time from a sound wave transmitted from a transmitter to the second space until the sound wave is received by a receiver, into the environment estimation device according to any one of claims 1 to 4; and displaying information relating to the stagnation of air in the second space (R2) output by the environment estimation device (10, 310).

11. inputting a parameter based on the propagation time in the second space (R2) into a trained model (22,122,222,422) trained using, as training data (41), a parameter based on the propagation time from when a sound wave transmitted from a transmitter (31) to a first space (R1) is received by a receiver (32) and information on the stagnation of air in the first space (R1) corresponding to the parameter; An environmental estimation method comprising a step of outputting information regarding the air stagnation in the second space (R2) estimated by the trained model (22,122,222,422).

12. A program for causing a computer to execute the environment estimation method according to claim 11.

13. a measuring unit (30) that measures the propagation time from when a sound wave is transmitted by a transmitter (31) until it is received by a receiver (32); a control unit (14), The control unit (14) inputting the parameters based on the propagation time measured by the measurement unit (30) in the second space (R2) into a trained model (22,122,222,422) that has been trained using, as training data (41), parameters based on the propagation time measured by the measurement unit (30) in the first space (R1) and information related to the air stagnation in the first space (R1) corresponding to the parameters; An environmental estimation system that outputs information regarding the air stagnation in the second space (R2) estimated by the trained model (22,122,222,422).

14. The environment estimation system according to claim 13, a server device (120) connected to the measurement unit (30) via a network (N), The server device (120) is an environment estimation system including the control unit (14).

15. acquiring a propagation time from when a sound wave is transmitted from a transmitter (31) to a first space until when the sound wave is received by a receiver (32); Obtaining information about the stagnation of air in the first space (R1); a step of performing machine learning using parameters calculated based on the propagation time and information regarding the air stagnation in the first space (R1) corresponding to the parameters as training data (41), inputting parameters based on the propagation time in the second space (R2), and producing a trained model (22,122,222,422) that outputs information regarding the air stagnation in the second space (R2).

Images