Ventilator, learning device and inference device

Abstract

To provide a ventilator that can present a proper timing of maintenance of an air filter for collecting dust contained in a ventilation airflow before clogging of the air filter is detected.SOLUTION: A ventilator comprises a blower for forming a ventilation airflow, an air filter for collecting dust contained in the ventilation airflow, and an inference device 114 which infers a proper maintenance timing for the air filter by using a learned model 113a as a model for inferring a proper maintenance timing for the air filter.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present disclosure relates to a ventilation device that supplies outdoor air and exhausts indoor air, a learning device, and an inference device. [Background technology]

[0002] Ventilating devices generally have an air filter installed at the air intake port, and the air filter collects dust contained in the ventilation flow to prevent dust from being taken into the room and the inside of the ventilation device from becoming dirty.

[0003] In a ventilation device equipped with an air filter, if the air filter becomes dirty due to dust collection, the pressure loss in the air filter increases and the ability to take in outside air decreases. For this reason, ventilation devices equipped with an air filter need to periodically clean or replace the air filter.

[0004] Patent Document 1 discloses a ventilation device that detects clogging of an air filter based on the current or rotation speed of an air supply fan. The ventilation device disclosed in Patent Document 1 notifies the user of the clogging of the air filter, enabling the user to clean the air filter at an appropriate time. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] JP 2020-153542 A Summary of the Invention [Problem to be solved by the invention]

[0006] However, the ventilation device disclosed in Patent Document 1 cannot detect the clogging of the air filter until the air filter actually becomes clogged and the ventilation capacity drops below a threshold. Therefore, the ventilation device disclosed in Patent Document 1 prepares to clean the air filter or arranges for a replacement air filter after detecting the clogging of the air filter. Therefore, the ventilation device disclosed in Patent Document 1 has a problem in that it must stop operation not only during the air filter replacement work but also while arranging for a replacement air filter, and must operate with a reduced ventilation capacity using a clogged air filter.

[0007] The present disclosure has been made in consideration of the above, and has an object to provide a ventilation device that can indicate the time for air filter maintenance before clogging of the air filter is detected. [Means for solving the problem]

[0008] In order to solve the above-mentioned problems and achieve the object, the present disclosure provides a ventilation device including a blower that forms a ventilation flow and an air filter that collects dust contained in the ventilation flow. The ventilation device also includes an inference device that infers the maintenance timing of the air filter by using a trained model that is a model for inferring the maintenance timing of the air filter. Effect of the Invention

[0009] According to the present disclosure, it is possible to obtain an effect of providing a ventilation device that can indicate the time for air filter maintenance before clogging of the air filter is detected. [Brief description of the drawings]

[0010] [Figure 1] FIG. 1 is a diagram showing a configuration of a heat exchange ventilation device according to a first embodiment. [Diagram 2] FIG. 1 is a diagram showing a configuration of a control device for a heat exchange ventilation device according to a first embodiment. [Diagram 3]FIG. 1 is a diagram showing the configuration of a learning device for a heat exchange ventilation device according to a first embodiment. [Figure 4] FIG. 1 is a diagram showing an example of the configuration of a neural network included in a model generating unit of a heat exchange ventilator according to a first embodiment. [Diagram 5] A flowchart of a learning process of a learning device for a heat exchange ventilation device according to the first embodiment. [Figure 6] FIG. 1 is a diagram showing a configuration of an inference device for a heat exchange type ventilation device according to a first embodiment. [Figure 7] A flowchart of an inference process of the inference device for the heat exchange ventilation device according to the first embodiment. [Figure 8] FIG. 13 is a diagram showing the configuration of a learning device for a heat exchange ventilation device according to a second embodiment. [Figure 9] FIG. 13 is a diagram showing a configuration of an inference device for a heat exchange ventilation device according to a second embodiment. [Figure 10] FIG. 13 is a diagram showing the configuration of a learning device for a heat exchange ventilation device according to a third embodiment. [Figure 11] FIG. 13 is a diagram showing a configuration of an inference device for a heat exchange ventilation device according to a third embodiment. [Figure 12] FIG. 13 is a diagram showing a configuration of an inference device for a heat exchange ventilation device according to a fourth embodiment. [Figure 13] FIG. 13 is a diagram showing the configuration of a heat exchange ventilation system according to a fifth embodiment. [Figure 14] FIG. 1 shows a hardware configuration of a learning device according to a first to fifth embodiments. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A ventilation device, a learning device, and an inference device according to embodiments will be described in detail below with reference to the drawings.

[0012] Embodiment 1 FIG. 1 is a diagram showing the configuration of a heat exchange ventilator according to the first embodiment. The heat exchange ventilator 50 according to the first embodiment includes a housing 1, a heat exchanger 2, an intake air blower 3, an exhaust air blower 4, and an air filter 15. The heat exchanger 2, the intake air blower 3, the exhaust air blower 4, and the air filter 15 are housed inside the housing 1. The housing 1 is formed with an outdoor air inlet 7, an indoor air inlet 8, an indoor air outlet 9, and an outdoor air outlet 10. The intake air blower 3 forms an intake air flow G1 that is sucked into the housing 1 from the outdoor air inlet 7 and blown out of the housing 1 from the indoor air outlet 9. The exhaust air blower 4 forms an exhaust air flow G2 that is sucked into the housing 1 from the indoor air inlet 8 and blown out of the housing 1 from the outdoor air outlet 10.

[0013] The heat exchange ventilator 50 ventilates while exchanging heat between the outside air OA drawn into the housing 1 by the supply air blower 3 through the outdoor air inlet 7 and the indoor air RA drawn into the housing 1 by the exhaust air blower 4 through the indoor air inlet 8 by the heat exchanger 2. The outside air OA after heat exchange in the heat exchanger 2 is supplied to the room as supply air SA through the indoor air outlet 9. The indoor air RA after heat exchange in the heat exchanger 2 is discharged to the outdoors as exhaust air EA through the outdoor air outlet 10.

[0014] The air filter 15 collects dust contained in the ventilation flow. The air filter 15 includes an intake air filter 5 installed upstream of the intake air blower 3 in the intake air duct through which the intake air flow G1 passes, and an exhaust air filter 6 installed upstream of the exhaust air blower 4 in the exhaust air duct through which the exhaust air flow G2 passes. The intake air filter 5 collects dust contained in the outside air OA. The intake air temperature and humidity sensor 13 measures the temperature and humidity of the outside air OA. The exhaust air filter 6 collects dust contained in the room air RA. The exhaust air temperature and humidity sensor 14 measures the temperature and humidity of the room air RA.

[0015] The heat exchange ventilator 50 includes a control device 11 and a remote controller 12. The remote controller 12 is connected to the control device 11. The control device 11 controls the supply air blower 3 and the exhaust air blower 4 based on a request from the remote controller 12. The remote controller 12 and the control device 11 may be connected wirelessly or wired. When the remote controller 12 and the control device 11 are connected wired, they may be directly connected by a dedicated line or may be connected via a network.

[0016] 2 is a diagram showing the configuration of a control device of a heat exchange ventilation device according to embodiment 1. The control device 11 includes a ventilation control unit 111 that controls the intake air blower 3 and the exhaust air blower 4, a learning device 112 that learns a learned model 113a that is a model for inferring the maintenance time of the air filter 15, a learned model storage unit 113 that stores the learned model 113a, an inference device 114 that infers the maintenance time Ts of the intake air filter 5, and a user notification unit 115 that notifies a user of the maintenance time Ts of the intake air filter 5. In embodiment 1, the model for inferring the maintenance time of the air filter 15 is a model for inferring the maintenance time Ts of the intake air filter 5.

[0017] 3 is a diagram showing the configuration of a learning device for the heat exchanging ventilator according to Embodiment 1. The learning device 112 includes a learning data acquiring unit 1121 and a model generating unit 1122.

[0018] The learning data acquisition unit 1121 acquires the total amount Qs of air passing through the intake air filter 5 in a unit time, the amount of change Ps of pressure loss of the intake air filter 5 in the unit time, and the average humidity Rs of air passing through the intake air filter 5 in the unit time as learning data. The unit time here is a control unit for the learning device 112 to perform learning, and can be set to any time. The total amount Qs of air passing through the intake air filter 5 in a unit time, the amount of change Ps of pressure loss of the intake air filter 5 in the unit time, and the average humidity Rs of air passing through the intake air filter 5 in the unit time acquired as learning data are not instantaneous values ​​measured in one unit time, but are average values ​​of measurement results obtained by performing measurements multiple times per unit time. In other words, the total amount Qs of air passing through the intake air filter 5 in a unit time, the amount of change Ps of pressure loss of the intake air filter 5 in the unit time, and the average humidity Rs of air passing through the intake air filter 5 in the unit time are continuously measured multiple times, and the average values ​​of the measurement results up to the present time are used as learning data.

[0019] By storing in advance as table data the relationship between the current value and rotation speed of air supply fan 3 and the air volume for each air volume notch, which is the command value for air supply fan 3, it is possible to determine the current air volume from the current command value and the current value and rotation speed of air supply fan 3. The total amount Qs of air that has passed through air supply filter 5 per unit time can be estimated by accumulating the air volumes passing through air supply filter 5, but the method for estimating the total amount Qs of air that has passed through air supply filter 5 per unit time is not limited to this.

[0020] The amount of change in pressure loss Ps of the intake air filter 5 per unit time can be estimated, for example, by installing differential pressure sensors before and after the intake air filter 5, but the method of estimating the amount of change in pressure loss Ps of the intake air filter 5 per unit time is not limited to this.

[0021] Dust in the air becomes less likely to fly up when it adsorbs water vapor. In other words, the higher the humidity, the more the dust is prevented from scattering, and the slower the rate at which dust accumulates on the intake air filter 5. Therefore, the learning device 112 uses humidity as one piece of learning data for predicting clogging of the intake air filter 5. The average humidity Rs of the air that has passed through the intake air filter 5 per unit time is calculated based on the measurement results of the intake air temperature and humidity sensor 13.

[0022] The model generating unit 1122 learns a model for inferring the maintenance timing Ts of the intake air filter 5 based on the learning data created based on a combination of the total amount Qs of air passing through the intake air filter 5 per unit time, the amount of change in pressure loss Ps of the intake air filter 5 per unit time, and the average humidity Rs of the air passing through the intake air filter 5 per unit time output from the learning data acquiring unit 1121. That is, a trained model 113a is generated that infers the maintenance timing Ts of the intake air filter 5 from the total amount Qs of air passing through the intake air filter 5 of the heat exchange ventilation device 50 per unit time, the amount of change in pressure loss Ps of the intake air filter 5 per unit time, and the average humidity Rs of the air passing through the intake air filter 5 per unit time. Here, the learning data is data in which the total amount Qs of air passing through the intake air filter 5 per unit time, the amount of change in pressure loss Ps of the intake air filter 5 per unit time, and the average humidity Rs of the air passing through the intake air filter 5 per unit time are associated with each other.

[0023] The learning algorithm used by the model generation unit 1122 may be a known algorithm such as supervised learning, unsupervised learning, reinforcement learning, etc. As an example, a case where a neural network is applied to the model generation unit 1122 will be described.

[0024] The model generation unit 1122 learns a model for inferring the maintenance time Ts of the intake air filter 5 by so-called supervised learning, for example, according to a neural network model. Here, supervised learning refers to a method in which a pair of input and label data, which is a result, is provided to the learning device 112, whereby the learning device 112 learns the features of the learning data and infers a result from the input.

[0025] 4 is a diagram showing a configuration example of a neural network included in the model generation unit of the heat exchange ventilator according to embodiment 1. The neural network 20 is composed of an input layer 21 consisting of a plurality of neurons X1, X2, and X3, an intermediate layer 22 which is a hidden layer consisting of a plurality of neurons Y1 and Y2, and an output layer 23 consisting of a plurality of neurons Z1, Z2, and Z3. Here, a neural network 20 having one intermediate layer 22 is taken as an example, but the intermediate layer 22 may have two or more layers.

[0026] For example, in the case of a three-layer neural network 20 including one intermediate layer 22 as shown in Fig. 4, when multiple inputs are input to the input layer 21, the values ​​are multiplied by weights w11, w12, w13, w14, w15, and w16 before being input to the intermediate layer 22, and the results are further multiplied by weights w21, w22, w23, w24, w25, and w26 before being output from the output layer 23. The output results from the output layer 23 vary depending on the values ​​of the weights w11, w12, w13, w14, w15, and w16 and w21, w22, w23, w24, w25, and w26.

[0027] The neural network 20 learns a model for inferring the maintenance time Ts of the intake air filter 5 by so-called supervised learning according to learning data created based on a combination of the total amount Qs of air passing through the intake air filter 5 per unit time, the amount of pressure loss change Ps of the intake air filter 5 per unit time, and the average humidity Rs of the air passing through the intake air filter 5 per unit time acquired by the learning data acquisition unit 1121.

[0028] In other words, the neural network 20 learns by inputting the total amount Qs of air passing through the intake air filter 5 per unit time and the average humidity Rs of the air passing through the intake air filter 5 per unit time into the input layer 21, and adjusting the weights w11, w12, w13, w14, w15, w16 and w21, w22, w23, w24, w25, w26 so that the result output from the output layer 23 approaches the maintenance time Ts for the intake air filter 5.

[0029] The model generation unit 1122 executes the above-mentioned learning to generate and output the trained model 113a.

[0030] The trained model 113a output from the model generation unit 1122 is stored in the trained model storage unit 113.

[0031] Next, a description will be given of the learning process performed by the learning device 112. Fig. 5 is a flowchart showing the learning process performed by the learning device of the heat exchanging ventilator according to the first embodiment.

[0032] In step S11, the learning data acquisition unit 1121 acquires the total amount Qs of air passing through the intake air filter 5 per unit time, the amount of change in pressure loss Ps of the intake air filter 5 per unit time, and the average humidity Rs of the air passing through the intake air filter 5 per unit time. Note that the total amount Qs of air passing through the intake air filter 5 per unit time, the amount of change in pressure loss Ps of the intake air filter 5 per unit time, and the average humidity Rs of the air passing through the intake air filter 5 per unit time are acquired simultaneously, but it is sufficient if the total amount Qs of air passing through the intake air filter 5 per unit time, the amount of change in pressure loss Ps of the intake air filter 5 per unit time, and the average humidity Rs of the air passing through the intake air filter 5 per unit time are input in an associated manner, and the data of the total amount Qs of air passing through the intake air filter 5 per unit time, the amount of change in pressure loss Ps of the intake air filter 5 per unit time, and the average humidity Rs of the air passing through the intake air filter 5 per unit time may be acquired at different times.

[0033] In step S12, the model generation unit 1122 performs a learning process to generate a learned model 113a by so-called supervised learning in accordance with learning data created based on a combination of the total amount Qs of air passing through the intake air filter 5 per unit time, the amount of change Ps in pressure loss of the intake air filter 5 per unit time, and the average humidity Rs of the air passing through the intake air filter 5 per unit time.

[0034] In step S13, the model generation unit 1122 stores the generated trained model 113a in the trained model storage unit 113.

[0035] 6 is a diagram showing the configuration of an inference device for the heat exchange ventilation device according to Embodiment 1. The inference device 114 includes an inference data acquisition unit 1141 and an inference unit 1142.

[0036] The inference data acquisition unit 1141 acquires the amount of change Ps in pressure loss of the supply air filter 5 per unit time.

[0037] The inference unit 1142 infers the maintenance time Ts of the intake air filter 5 obtained by using the trained model 113a. That is, by inputting the pressure loss change amount Ps of the intake air filter 5 per unit time acquired by the inference data acquisition unit 1141 into the trained model 113a, the maintenance time Ts of the intake air filter 5 inferred from the pressure loss change amount Ps of the intake air filter 5 per unit time can be output to the user notification unit 115.

[0038] In addition, in embodiment 1, it has been described that the maintenance time Ts of the intake air filter 5 is output to the user notification unit 115 using the trained model 113a trained by the model generation unit 1122 of the heat exchanger ventilation device 50. However, it is also possible to obtain the trained model 113a from an external source, such as another heat exchanger type ventilation device, and output the maintenance time Ts of the intake air filter 5 to the user notification unit 115 based on this trained model 113a.

[0039] Next, a description will be given of a process for obtaining the maintenance timing Ts of the supply air filter 5 using the inference device 114. Fig. 7 is a flowchart relating to the inference process of the inference device of the heat exchange ventilator according to the first embodiment.

[0040] In step S21, the inference data acquisition unit 1141 acquires data on the amount of change Ps in pressure loss of the supply air filter 5 per unit time.

[0041] In step S22, the inference unit 1142 inputs the amount of change in pressure loss Ps of the intake air filter 5 in unit time into the learned model 113a stored in the learned model storage unit 113, and obtains the maintenance timing Ts of the intake air filter 5. Note that the amount of change in pressure loss Ps of the intake air filter 5 in unit time input to the inference unit 1142 is not an instantaneous value measured in one unit time, but an average value of measurement results obtained by performing multiple measurements per unit time.

[0042] In step S23, the inference unit 1142 outputs to the user notification unit 115 the maintenance timing Ts of the intake air filter 5 obtained by the trained model 113a.

[0043] In step S24, the user notification unit 115 notifies the user of the maintenance time Ts by displaying the predicted maintenance time on the remote controller 12 using the maintenance time Ts for the intake air filter 5 input from the inference device 114. This allows the user of the heat exchanging ventilator 50 to perform maintenance of the intake air filter 5 in accordance with the predicted maintenance time for the intake air filter 5.

[0044] In the first embodiment, a case has been described in which supervised learning is applied to the learning algorithm used by the model generation unit 1122, but the present invention is not limited to this. As for the learning algorithm, reinforcement learning, unsupervised learning, semi-supervised learning, or the like can also be applied in addition to supervised learning.

[0045] The model generating unit 1122 may learn the maintenance time Ts of the intake air filter 5 according to the learning data created for a plurality of heat exchange type ventilation devices. The model generating unit 1122 may acquire learning data from a plurality of heat exchange type ventilation devices used in the same area, or may learn the maintenance time Ts of the intake air filter 5 by using learning data collected from a plurality of heat exchange type ventilation devices operating independently in different areas. It is also possible to add or remove a heat exchange type ventilation device from which learning data is collected as a target midway. Furthermore, the learning device 112 that has learned the maintenance time Ts of the intake air filter 5 for a certain heat exchange type ventilation device may be applied to another heat exchange type ventilation device, and the maintenance time Ts of the intake air filter 5 for the other heat exchange type ventilation device may be re-learned and updated.

[0046] In addition, the learning algorithm used in the model generation unit 1122 can be deep learning, which learns to extract the features themselves, or machine learning can be performed according to other well-known methods, such as genetic programming, functional logic programming, support vector machines, function approximation, etc.

[0047] Since the heat exchange ventilator 50 according to the first embodiment displays the predicted maintenance time on the remote controller 12 using the maintenance time Ts of the intake air filter 5, the user of the heat exchange ventilator 50 can arrange for a replacement intake air filter 5 before the intake air filter 5 becomes clogged. Therefore, the heat exchange ventilator 50 according to the first embodiment can reduce the downtime for maintenance of the intake air filter 5. In addition, it is not necessary to operate the heat exchange ventilator 50 in a state where the ventilation capacity is reduced by using a clogged intake air filter 5 while arranging for a replacement intake air filter 5.

[0048] Embodiment 2 FIG. 8 is a diagram showing the configuration of a learning device for a heat exchange ventilation device according to the second embodiment. In the learning device 112 according to the second embodiment, the model for inferring the maintenance timing of the air filter 15 is a model for inferring the maintenance timing Te of the exhaust filter 6. In the learning device 112 according to the second embodiment, the learning data acquisition unit 1121 acquires the total amount Qe of air passing through the exhaust filter 6 in a unit time, the pressure loss change amount Pe of the exhaust filter 6 in a unit time, and the average humidity Re of the air passing through the exhaust filter 6 in a unit time as learning data. Note that the total amount Qe of air passing through the exhaust filter 6 in a unit time, the pressure loss change amount Pe of the exhaust filter 6 in a unit time, and the average humidity Re of the air passing through the exhaust filter 6 in a unit time acquired as learning data are not instantaneous values ​​measured in one unit time, but are average values ​​of measurement results obtained by performing measurements multiple times per unit time. That is, the total amount Qe of air passing through the exhaust filter 6 per unit time, the amount of change in pressure loss Pe of the exhaust filter 6 per unit time, and the average humidity Re of the air passing through the exhaust filter 6 per unit time are continuously measured multiple times, and the average values ​​of the measurement results up to the present time are used as learning data.

[0049] The model generation unit 1122 learns a model for inferring the maintenance timing Te of the exhaust filter 6 based on learning data created based on the total amount Qe of air passing through the exhaust filter 6 in a unit time, the pressure loss change amount Pe of the exhaust filter 6 in a unit time, and the average humidity Re of the air passing through the exhaust filter 6 in a unit time. That is, a trained model 113a is generated that infers the maintenance timing Te of the exhaust filter 6 from the total amount Qe of air passing through the exhaust filter 6 of the heat exchange ventilation device 50 in a unit time, the pressure loss change amount Pe of the exhaust filter 6 in a unit time, and the average humidity Re of the air passing through the exhaust filter 6 in a unit time. Here, the learning data is data in which the total amount Qe of air passing through the exhaust filter 6 in a unit time, the pressure loss change amount Pe of the exhaust filter 6 in a unit time, and the average humidity Re of the air passing through the exhaust filter 6 in a unit time are associated with each other.

[0050] 9 is a diagram showing a configuration of an inference device for a heat exchange ventilator according to embodiment 2. In an inference device 114 according to embodiment 2, an inference data acquisition unit 1141 acquires a pressure loss change amount Pe of the exhaust filter 6 per unit time.

[0051] The inference unit 1142 infers the maintenance timing Te of the exhaust filter 6 obtained by using the trained model 113a. That is, by inputting the pressure loss change amount Pe of the exhaust filter 6 in a unit time acquired by the inference data acquisition unit 1141 into the trained model 113a, it is possible to output the maintenance timing Te of the exhaust filter 6 inferred from the pressure loss change amount Pe of the exhaust filter 6 in a unit time. Note that the pressure loss change amount Pe of the exhaust filter 6 in a unit time input to the inference unit 1142 is not an instantaneous value measured in one unit time, but an average value of the measurement results obtained by performing multiple measurements per unit time.

[0052] The user notification unit 115 uses the maintenance timing Te of the exhaust filter 6 input from the inference device 114 to display the predicted maintenance timing on the remote controller 12.

[0053] The heat exchange ventilator 50 according to the second embodiment displays the predicted maintenance time on the remote controller 12 using the maintenance time Te of the exhaust filter 6, so that the user of the heat exchange ventilator 50 can arrange for a replacement exhaust filter 6 before the exhaust filter 6 becomes clogged. Therefore, the heat exchange ventilator 50 according to the second embodiment can reduce the downtime for maintenance of the exhaust filter 6. In addition, it is not necessary to operate the heat exchange ventilator 50 in a state where the ventilation capacity is reduced by using a clogged exhaust filter 6 while arranging for a replacement exhaust filter 6.

[0054] Embodiment 3 10 is a diagram showing the configuration of a learning device for a heat exchange ventilator according to embodiment 3. In learning device 112 according to embodiment 3, learning data acquisition unit 1121 acquires, as learning data, a total amount Qs of air passing through intake air filter 5 per unit time, a pressure loss change amount Ps of intake air filter 5 per unit time, an average humidity Rs of air passing through intake air filter 5 per unit time, a total amount Qe of air passing through exhaust filter 6 per unit time, a pressure loss change amount Pe of exhaust filter 6 per unit time, and an average humidity Re of air passing through exhaust filter 6 per unit time. The total amount of air Qs passing through the intake filter 5 in a unit time, the amount of change in pressure loss Ps of the intake filter 5 in a unit time, the average humidity Rs of the air passing through the intake filter 5 in a unit time, the total amount of air Qe passing through the exhaust filter 6 in a unit time, the amount of change in pressure loss Pe of the exhaust filter 6 in a unit time, and the average humidity Re of the air passing through the exhaust filter 6 in a unit time, which are acquired as learning data, are not instantaneous values ​​measured in one unit time, but are average values ​​of measurement results obtained by performing measurements multiple times per unit time. That is, the total amount of air Qs passing through the intake filter 5 in a unit time, the amount of change in pressure loss Ps of the intake filter 5 in a unit time, the average humidity Rs of the air passing through the intake filter 5 in a unit time, the total amount of air Qe passing through the exhaust filter 6 in a unit time, the amount of change in pressure loss Pe of the exhaust filter 6 in a unit time, and the average humidity Re of the air passing through the exhaust filter 6 in a unit time are continuously measured multiple times, and the average values ​​of the measurement results up to the present time are used as learning data.

[0055] The model generation unit 1122 learns a model for inferring the maintenance timing Ts of the intake air filter 5 based on learning data created based on the total amount Qs of air passing through the intake air filter 5 per unit time, the amount of change in pressure loss Ps of the intake air filter 5 per unit time, and the average humidity Rs of the air passing through the intake air filter 5 per unit time. The model generation unit 1122 also learns a model for inferring the maintenance timing Te of the exhaust filter 6 based on learning data created based on the total amount Qe of air passing through the exhaust filter 6 per unit time, the amount of change in pressure loss Pe of the exhaust filter 6 per unit time, and the average humidity Re of the air passing through the exhaust filter 6 per unit time.

[0056] 11 is a diagram showing a configuration of an inference device for a heat exchange ventilator according to embodiment 3. In an inference device 114 according to embodiment 3, an inference data acquisition unit 1141 acquires a pressure loss change amount Ps of the supply air filter 5 per unit time and a pressure loss change amount Pe of the exhaust filter 6 per unit time.

[0057] The inference unit 1142 infers the maintenance time Ts of the intake air filter 5 and the maintenance time Te of the exhaust filter 6 obtained by using the trained model 113a. That is, by inputting the pressure loss change amount Ps of the intake air filter 5 in a unit time and the pressure loss change amount Pe of the exhaust filter 6 in a unit time acquired by the inference data acquisition unit 1141 into the trained model 113a, it is possible to output the maintenance time Ts of the intake air filter 5 inferred from the pressure loss change amount Ps of the intake air filter 5 in a unit time and the maintenance time Te of the exhaust filter 6 inferred from the pressure loss change amount Pe of the exhaust filter 6 in a unit time. Note that the pressure loss change amount Ps of the intake air filter 5 in a unit time and the pressure loss change amount Pe of the exhaust filter 6 in a unit time input to the inference unit 1142 are not instantaneous values ​​measured in one unit time, but are average values ​​of measurement results obtained by performing measurements multiple times per unit time.

[0058] The user notification unit 115 uses the maintenance timing Ts of the intake air filter 5 and the maintenance timing Te of the exhaust air filter 6 input from the inference unit 1142 to display the predicted maintenance timing on the remote controller 12.

[0059] The heat exchange ventilator 50 according to the third embodiment displays the predicted maintenance time on the remote controller 12 using the maintenance time Ts of the intake air filter 5 and the maintenance time Te of the exhaust filter 6, so that the user of the heat exchange ventilator 50 can prepare replacement intake air filter 5 and exhaust filter 6 before the intake air filter 5 and exhaust filter 6 become clogged. Therefore, the heat exchange ventilator 50 according to the third embodiment can reduce the downtime for maintenance of the intake air filter 5 and exhaust filter 6. In addition, it is not necessary to operate the heat exchange ventilator 50 in a state where the ventilation capacity is reduced by using the clogged intake air filter 5 and exhaust filter 6 while preparing the replacement intake air filter 5 and exhaust filter 6.

[0060] Embodiment 4 FIG. 12 is a diagram showing the configuration of an inference device for a heat exchange ventilation device according to the fourth embodiment. In the inference device 114 according to the fourth embodiment, the inference data acquisition unit 1141 acquires the total amount Qs of air passing through the intake air filter 5 in a unit time. The inference unit 1142 infers the maintenance time Ts of the intake air filter 5 obtained by using the trained model 113a. That is, by inputting the total amount Qs of air passing through the intake air filter 5 in a unit time acquired by the inference data acquisition unit 1141 into the trained model 113a, the maintenance time Ts of the intake air filter 5 inferred from the pressure loss change amount Ps of the intake air filter 5 in the unit time can be output. Note that the total amount Qs of air passing through the intake air filter 5 in a unit time input to the inference unit 1142 is not an instantaneous value measured in one unit time, but is an average value of the measurement results obtained by performing multiple measurements per unit time.

[0061] The user notification unit 115 uses the maintenance timing Ts for the intake air filter 5 input from the inference device 114 to display the predicted maintenance timing on the remote controller 12.

[0062] Like the heat exchanger ventilator 50 according to the first embodiment, the heat exchanger ventilator 50 according to the fourth embodiment can reduce the downtime for maintenance of the intake air filter 5. In addition, it is not necessary to operate the heat exchanger ventilator 50 in a state where the ventilation capacity is reduced due to a clogged intake air filter 5 while a replacement intake air filter 5 is being prepared.

[0063] In the above first to fourth embodiments, the heat exchanger ventilator 50 including the learning device 112 and the inference device 114 has been given as an example, but it is also possible to implement the heat exchanger ventilator 50 in a configuration that omits the learning device 112. In this case, the inference device 114 may infer the maintenance timing of the intake air filter 5 or the exhaust filter 6 using a trained model trained by a learning device configured as a device separate from the heat exchanger ventilator 50.

[0064] Embodiment 5. 13 is a diagram showing the configuration of a heat exchange ventilation system according to embodiment 5. The heat exchange ventilation system 100 includes a heat exchange ventilation device 50, a learning device 116, a trained model storage device 117, and an inference device 118. The heat exchange ventilation device 50 differs from the heat exchange ventilation device 50 according to embodiment 1 in that the control device 11 does not include the learning device 112, the trained model storage unit 113, and the inference device 114.

[0065] The learning device 116 is an independent device formed from the learning device 112 included in the heat exchange ventilator 50 according to the first embodiment, and acquires the total amount Qs of air passing through the intake air filter 5 in a unit time, the amount of change in pressure loss Ps of the intake air filter 5 in a unit time, and the average humidity Rs of the air passing through the intake air filter 5 in a unit time as learning data to generate a trained model 113a. The total amount Qs of air passing through the intake air filter 5 in a unit time, the amount of change in pressure loss Ps of the intake air filter 5 in a unit time, and the average humidity Rs of the air passing through the intake air filter 5 in a unit time acquired as learning data are not instantaneous values ​​measured in one unit time, but are average values ​​of measurement results obtained by performing measurements multiple times per unit time. That is, the total amount Qs of air passing through the intake air filter 5 in a unit time, the amount of change in pressure loss Ps of the intake air filter 5 in a unit time, and the average humidity Rs of the air passing through the intake air filter 5 in a unit time are continuously measured multiple times, and the average values ​​of the measurement results up to the present time are used as learning data. The trained model 113a is stored in a trained model storage device 117. The inference device 118 is an independent device configured from the inference device 114 included in the heat exchange ventilation device 50 according to the first embodiment, and acquires the pressure loss change amount Ps of the intake air filter 5 per unit time, and infers the maintenance timing Ts of the intake air filter 5 obtained by using the trained model 113a. Note that the pressure loss change amount Ps of the intake air filter 5 per unit time input to the inference unit 1142 is not an instantaneous value measured in one unit time, but an average value of the measurement results obtained by performing multiple measurements per unit time.

[0066] The user notification unit 115 uses the maintenance timing Ts of the intake air filter 5 to display the predicted maintenance timing on the remote controller 12. This makes it possible to minimize the equipment downtime due to the maintenance of the intake air filter 5 according to the predicted maintenance timing of the intake air filter 5.

[0067] The heat exchange ventilation system 100 according to the fifth embodiment, like the heat exchange ventilation device 50 according to the first embodiment, displays the predicted maintenance time on the remote controller 12 using the maintenance time Ts of the intake air filter 5, so that the user of the heat exchange ventilation device 50 can arrange for a replacement intake air filter 5 before the intake air filter 5 becomes clogged. Therefore, the heat exchange ventilation device 50 according to the fifth embodiment can reduce the downtime for maintenance of the intake air filter 5. In addition, it is not necessary to operate the heat exchange ventilation device 50 in a state where the ventilation capacity is reduced by using a clogged intake air filter 5 while arranging for a replacement intake air filter 5.

[0068] Here, the configuration in which the learning device 116 and the inference device 118 are directly connected to the heat exchange ventilation device 50 has been exemplified, but the learning device 116 and the inference device 118 may be connected to the heat exchange ventilation device 50 via a network. Also, the learning device 116 and the inference device 118 may not be independent devices, but may be provided on a server on a network, and the server on which the learning device 116 and the inference device 118 are provided may be connected to the heat exchange ventilation device 50. Also, the learning device 116 and the inference device 118 may be configured as independent devices, such as the learning device 112 and the inference device 114 of the heat exchange ventilation device 50 according to the second, third, or fourth embodiment.

[0069] In each of the above embodiments, the learning data input to the learning device 112, 116 was the total amount of air passing through at least one of the intake air filter 5 and the exhaust air filter 6 per unit time, the amount of change in pressure loss per unit time, and the average humidity of the air passing through per unit time, but the learning data may be the total amount of air passing through at least one of the intake air filter 5 and the exhaust air filter 6 per unit time and the amount of change in pressure loss per unit time. The learning data may also be the total amount of air passing through at least one of the intake air filter 5 and the exhaust air filter 6 per unit time and the average humidity of the air passing through per unit time. The learning data may also be the amount of change in pressure loss per unit time and the average humidity of the air passing through per unit time, of at least one of the intake air filter 5 and the exhaust air filter 6. In any case, the average value of the measurement results up to the present time, which are continuously measured multiple times, is used as the learning data.

[0070] Furthermore, in each of the above embodiments, the data for inference input to the inference devices 114, 118 was the total amount of air passing through at least one of the intake air filter 5 and the exhaust air filter 6 per unit time or the amount of change in pressure loss per unit time, but the data for inference may be the total amount of air passing through per unit time and the amount of change in pressure loss per unit time. Furthermore, the data for inference may include the average humidity of the air passing through per unit time in addition to at least one of the total amount of air passing through per unit time and the amount of change in pressure loss per unit time. In either case, the average value of the measurement results obtained by continuously measuring multiple times up to the present time is used as the data for inference.

[0071] The following describes the hardware configuration of the learning devices 112, 116 and the inference devices 114, 118 according to embodiments 1 to 5. Fig. 14 is a diagram showing the hardware configuration of the learning devices according to embodiments 1 to 5. The learning devices 112, 116 are realized by a computer system including a processor 91 that executes various processes, a memory 92 that is a main memory, and a storage device 93 that stores information.

[0072] The processor 91 may be a calculation means such as a calculation device, a microprocessor, a microcomputer, a CPU (Central Processing Unit), or a DSP (Digital Signal Processor). The memory 92 may be a non-volatile or volatile semiconductor memory such as a RAM (Random Access Memory), a ROM (Read Only Memory), a flash memory, an EPROM (Erasable Programmable Read Only Memory), or an EEPROM (Electrically Erasable Programmable Read Only Memory). The storage device 93 stores a program for performing a process of learning a model for inferring the maintenance time of the air filter 15.

[0073] In the above computer system, the processor 91 reads out into the memory 92 a program corresponding to the processing of each component stored in the storage device 93 and executes the program, thereby realizing the function of the model generation unit 1122. The memory 92 is also used as a temporary memory for each processing executed by the processor 91. The program executed by the processor 91 may be provided in a state stored in a storage medium, or may be provided via a network.

[0074] Like the learning devices 112 and 116, the inference devices 114 and 118 are realized by a computer system including a processor 91 that executes various processes, a memory 92 that is a main memory, and a storage device 93 that stores information.

[0075] In the above first to fifth embodiments, the heat exchange ventilator 50 including the heat exchanger 2 has been taken as an example, but a ventilator not including the heat exchanger 2 can also be implemented in the same manner.

[0076] The configurations shown in the above embodiments are merely examples of the contents, and may be combined with other known technologies. Parts of the configurations may be omitted or modified without departing from the spirit of the invention.

[0077] Various aspects of the present disclosure are summarized below as appendices.

[0078] (Appendix 1) A blower for forming a ventilation flow; an air filter that collects dust contained in the ventilation flow; A ventilation device comprising: an inference device that infers the maintenance timing of the air filter using a trained model that is a model for inferring the maintenance timing of the air filter. (Appendix 2) The ventilation device according to claim 1, further comprising a learning device that learns the trained model. (Appendix 3) The learning device includes: a learning data acquisition unit that acquires learning data including a total amount of air passing through the air filter per unit time and a change in pressure loss of the air filter per unit time; and a model generation unit that generates the trained model using the learning data. (Appendix 4) The learning device includes: a learning data acquisition unit that acquires learning data including a total amount of air that has passed through the air filter per unit time and an average humidity of the air that has passed through the air filter per unit time; and a model generation unit that generates the trained model using the learning data. (Appendix 5) The learning device includes: a learning data acquisition unit that acquires learning data including a pressure loss change amount of the air filter per unit time and an average humidity of air passing through the air filter per unit time; and a model generation unit that generates the trained model using the learning data. (Appendix 6) The learning device includes: a learning data acquisition unit that acquires learning data including a total amount of air passing through the air filter per unit time, a pressure loss change amount of the air filter per unit time, and an average humidity of the air passing through the air filter per unit time; and a model generation unit that generates the trained model using the learning data. (Appendix 7) The ventilation device according to any one of Appendix 3, Appendix 4, and Appendix 6, wherein the trained model is a model for inferring the timing of maintenance of the air filter from the total amount of air passing through the air filter per unit time. (Appendix 8) The ventilation device according to any one of Appendix 3, Appendix 5, and Appendix 6, characterized in that the trained model is a model for inferring the timing of maintenance of the air filter from the amount of change in pressure loss of the air filter per unit time. (Appendix 9) The ventilation device described in Appendix 3 or Appendix 6, characterized in that the trained model is a model for inferring the timing of maintenance of the air filter from the total amount of air passing through the air filter per unit time and the amount of change in pressure loss of the air filter per unit time. (Appendix 10) The ventilation device according to claim 4 or 6, wherein the trained model is a model for inferring the timing of maintenance of the air filter from the total amount of air passing through the air filter per unit time and the average humidity of the air passing through the air filter per unit time. (Appendix 11) The ventilation device according to claim 5 or 6, wherein the trained model is a model for inferring the timing of maintenance of the air filter from the amount of change in pressure loss of the air filter per unit time and the average humidity of the air passing through the air filter per unit time. (Appendix 12) The ventilation device described in Appendix 6, wherein the trained model is a model for inferring the time for maintenance of the air filter from the total amount of air passing through the air filter per unit time, the change in pressure loss of the air filter per unit time, and the average humidity of the air passing through the air filter per unit time. (Appendix 13) The blower includes an intake air blower that forms an intake air flow and an exhaust air blower that forms an exhaust air flow, The air filter includes an intake air filter that collects dust contained in an intake air flow and an exhaust air filter that collects dust contained in an exhaust air flow, The learning device learns the learned model for at least one of the intake air filter and the exhaust filter, The ventilation device described in any one of Appendix 2 to Appendix 12, characterized in that the inference device infers the maintenance timing for the intake air filter and the exhaust filter, for which the learning device has learned the trained model. (Appendix 14) the trained model is a model for inferring a maintenance timing of the air filter from a change in pressure loss of the air filter per unit time, The inference device includes an inference data acquisition unit that acquires a pressure loss change amount of the air filter per unit time; The ventilation device described in any one of Appendix 1 to Appendix 6, characterized in that it is equipped with an inference unit that uses the learned model to infer the time for maintenance of the air filter from the amount of change in pressure loss of the air filter per unit time. (Appendix 15) The trained model is a model for inferring a maintenance time for the air filter from a total amount of air passing through the air filter per unit time, The inference device includes an inference data acquisition unit that acquires a total amount of air passing through the air filter per unit time; The ventilation device according to any one of claims 1 to 6, further comprising an inference unit that uses the learned model to infer the time for maintenance of the air filter from the total amount of air that passes through the air filter per unit time. [Explanation of symbols]

[0079] 1 housing, 2 heat exchanger, 3 intake air blower, 4 exhaust air blower, 5 intake air filter, 6 exhaust air filter, 7 outdoor air inlet, 8 indoor air inlet, 9 indoor air outlet, 10 outdoor air outlet, 11 control device, 12 remote controller, 13 intake air temperature and humidity sensor, 14 exhaust air temperature and humidity sensor, 15 air filter, 20 neural network, 21 input layer, 22 intermediate layer, 23 output layer, 50 heat exchange ventilation device, 91 processor, 92 memory, 93 storage device, 100 heat exchange ventilation system, 111 ventilation control unit, 112, 116 learning device, 113 learned model storage unit, 113a learned model, 114, 118 inference device, 115 user notification unit, 117 learned model storage unit, 1121 learning data acquisition unit, 1122 Model generation unit, 1141 inference data acquisition unit, 1142 inference unit.

Claims

1. A blower for forming a ventilation flow; an air filter that collects dust contained in the ventilation flow; A ventilation device comprising: an inference device that infers the maintenance timing of the air filter using a trained model that is a model for inferring the maintenance timing of the air filter.

2. The ventilation apparatus according to claim 1 , further comprising a learning device that learns the learned model.

3. The learning device includes: a learning data acquisition unit that acquires learning data including a total amount of air passing through the air filter per unit time and a change in pressure loss of the air filter per unit time; The ventilation device according to claim 2, further comprising a model generation unit that generates the trained model using the learning data.

4. The learning device includes: a learning data acquisition unit that acquires learning data including a total amount of air that has passed through the air filter per unit time and an average humidity of the air that has passed through the air filter per unit time; The ventilation device according to claim 2, further comprising a model generation unit that generates the trained model using the learning data.

5. The learning device includes: a learning data acquisition unit that acquires learning data including a pressure loss change amount of the air filter per unit time and an average humidity of air passing through the air filter per unit time; The ventilation device according to claim 2, further comprising a model generation unit that generates the trained model using the learning data.

6. The learning device includes: a learning data acquisition unit that acquires learning data including a total amount of air passing through the air filter per unit time, a pressure loss change amount of the air filter per unit time, and an average humidity of the air passing through the air filter per unit time; The ventilation device according to claim 2, further comprising a model generation unit that generates the trained model using the learning data.

7. The ventilation device according to claim 3, wherein the learned model is a model for inferring the time for maintenance of the air filter from the total amount of air that passes through the air filter per unit time.

8. The ventilation device according to claim 3, wherein the trained model is a model for inferring the timing of maintenance of the air filter from the amount of change in pressure loss of the air filter per unit time.

9. The ventilation device described in claim 3, characterized in that the learned model is a model for inferring the time for maintenance of the air filter from the total amount of air passing through the air filter per unit time and the change in pressure loss of the air filter per unit time.

10. The ventilation device described in claim 4, characterized in that the learned model is a model for inferring the time for maintenance of the air filter from the total amount of air passing through the air filter per unit time and the average humidity of the air passing through the air filter per unit time.

11. The ventilation device described in claim 5, characterized in that the learned model is a model for inferring the time for maintenance of the air filter from the amount of change in pressure loss of the air filter per unit time and the average humidity of the air passing through the air filter per unit time.

12. The ventilation device described in claim 6, characterized in that the learned model is a model for inferring the time for maintenance of the air filter from the total amount of air passing through the air filter per unit time, the change in pressure loss of the air filter per unit time, and the average humidity of the air passing through the air filter per unit time.

13. The blower includes an intake air blower that forms an intake air flow and an exhaust air blower that forms an exhaust air flow, The air filter includes an intake air filter that collects dust contained in an intake air flow and an exhaust air filter that collects dust contained in an exhaust air flow, The learning device learns the learned model for at least one of the intake air filter and the exhaust filter, The ventilation apparatus according to any one of claims 2 to 12, characterized in that the inference device infers the maintenance timing for the intake air filter and the exhaust filter, for which the learning device has learned the trained model.

14. the trained model is a model for inferring a maintenance timing of the air filter from a change in pressure loss of the air filter per unit time, The inference device includes an inference data acquisition unit that acquires a pressure loss change amount of the air filter per unit time; The ventilation device according to any one of claims 1 to 6, characterized in that it comprises an inference unit that uses the learned model to infer the time for maintenance of the air filter from the amount of change in pressure loss of the air filter per unit time.

15. The trained model is a model for inferring a maintenance time for the air filter from a total amount of air passing through the air filter per unit time, The inference device includes an inference data acquisition unit that acquires a total amount of air passing through the air filter per unit time; The ventilation device according to any one of claims 1 to 6, characterized in that it comprises an inference unit that uses the learned model to infer the time for maintenance of the air filter from the total amount of air that passes through the air filter per unit time.

16. a learning data acquisition unit that acquires learning data including a total amount of air passing through an air filter of the ventilation device per unit time, a pressure loss change amount of the air filter per unit time, and an average humidity of the air passing through the air filter per unit time; and a model generation unit that uses the learning data to generate a trained model for inferring the time for maintenance of the air filter from the amount of change in pressure loss of the air filter per unit time or the total amount of air passing through the air filter per unit time.

17. An inference device comprising: an inference data acquisition unit that acquires a change in pressure loss of an air filter of a ventilation device per unit time; and an inference unit that infers the time for maintenance of the air filter from the total amount of air passing through the air filter per unit time using a learned model for inferring the time for maintenance of the air filter from the change in pressure loss of the air filter per unit time.

Images