Inference device, inference method, and learning device

Abstract

An inference device (10) infers sound-related data associated with the sound of a compressor (6) that compresses a refrigerant. The inference device (10) comprises: a data acquisition unit (111) that acquires individual data that indicates variations among compressors (6) and is correlated with the sound-related data associated with the sound of the compressor compressing the refrigerant; and an inference unit (113) that uses a trained model (20) to infer sound-related data on the basis of individual data acquired by the data acquisition unit (111).

Description

[Technical Field]

[0001] The present disclosure relates to an inference device, an inference method, and a learning device that infer sound-related data related to the sound of a compressor that compresses a refrigerant. [Background technology]

[0002] Conventionally, compressors that compress a refrigerant are known. For example, International Publication No. 2020 / 208777 (Patent Document 1) discloses a compressor that suppresses noise by preventing an increase in the inner diameter out-of-roundness of a stator. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] International Publication No. 2020 / 208777 Summary of the Invention [Problem to be solved by the invention]

[0004] According to the compressor disclosed in International Publication No. 2020 / 208777 (Patent Document 1), noise can be suppressed by preventing an increase in the inner diameter roundness of the stator used in the electric motor. However, because the inner diameter roundness of the stator is merely adjusted for each component, preventing an increase in the inner diameter roundness of the stator does not necessarily result in suppression of noise in the actually manufactured compressor.

[0005] To check the noise of a compressor, it is possible to actually inspect the compressor. However, the pressure conditions in the inspection equipment must be stabilized each time the compressor noise is inspected, and inspecting every compressor during the manufacturing process would require a significant amount of time. For this reason, compressor noise inspections are typically performed by sampling inspections. However, because the multiple components that make up a compressor have individual variations, the noise of a compressor made up of multiple components varies from compressor to compressor. Furthermore, when multiple components are assembled, the multiple components may interact with each other, affecting the compressor noise. For this reason, it is difficult to guarantee that the noise of all manufactured compressors is within the acceptable range through sampling inspections.

[0006] The present disclosure has been made to solve the above-mentioned problem, and aims to provide a technology for easily checking data related to the sound of a compressor. [Means for solving the problem]

[0007] An inference device according to the present disclosure is an inference device that infers sound-related data related to the sound of a compressor that compresses a refrigerant. The inference device includes a data acquisition unit that acquires individual data indicating individual variations of the compressor that are correlated with the sound-related data, and an inference unit that infers the sound-related data based on the individual data acquired by the data acquisition unit, using a trained model for inferring the sound-related data based on the individual data. The compressor includes a compression mechanism for compressing a refrigerant and an electric motor for supplying power to the compression mechanism for compressing the refrigerant. The compression mechanism includes a cylinder, a rolling piston that rotates along the inner circumferential surface of the cylinder based on power from the electric motor, a vane that divides a compression chamber formed by the inner circumferential surface of the cylinder and the outer circumferential surface of the rolling piston into an intake side and a compression side, a vane groove in the cylinder into which the vane is inserted, an upper frame that supports the cylinder from above, a lower frame that supports the cylinder from below, and a shaft inserted into the upper and lower frames to rotate the rolling piston. The shaft includes an upper shaft portion inserted into the upper frame and a lower shaft portion inserted into the lower frame. The individual data includes at least one of the coaxiality between the upper shaft portion and the lower shaft portion and the dimension of the gap between the vane and the vane groove. In the inference device according to the present disclosure, the electric motor may also include a stator, a winding wound around the stator, and a rotor disposed inside the stator. The individual data may include at least one of the amount of magnetic flux of the rotor interlinking with the winding, the roundness of the inner diameter of the stator, and the amount of eccentricity of the rotor.

[0008] An inference method according to the present disclosure is an inference method for inferring sound-related data related to sound of a compressor that compresses a refrigerant by a computer. The inference method includes the steps of acquiring individual data indicating individual variations of the compressor correlated with the sound-related data, and inferring the sound-related data based on the individual data acquired in the acquiring step, using a trained model for inferring the sound-related data based on the individual data. The compressor includes a compression mechanism for compressing a refrigerant and an electric motor for supplying power to the compression mechanism for compressing the refrigerant. The compression mechanism includes a cylinder, a rolling piston that rotates along the inner circumferential surface of the cylinder based on power from the electric motor, a vane that divides a compression chamber formed by the inner circumferential surface of the cylinder and the outer circumferential surface of the rolling piston into a suction side and a compression side, a vane groove in the cylinder into which the vane is inserted, an upper frame that supports the cylinder from above, a lower frame that supports the cylinder from below, and a shaft inserted into the upper and lower frames to rotate the rolling piston. The shaft includes an upper shaft portion inserted into the upper frame and a lower shaft portion inserted into the lower frame. The individual data includes at least one of the coaxiality between the upper shaft portion and the lower shaft portion and the dimension of the gap between the vane and the vane groove. In the inference method according to the present disclosure, the electric motor may also include a stator, a winding wound around the stator, and a rotor disposed inside the stator. The individual data may include at least one of the amount of magnetic flux of the rotor interlinking with the winding, the roundness of the inner diameter of the stator, and the amount of eccentricity of the rotor.

[0009] A learning device according to the present disclosure is a learning device for performing supervised learning, and includes a data acquisition unit that acquires learning data including individual data indicating individual variations of compressors that are correlated with associated data related to the sound of the compressor that compresses a refrigerant and sound-related data, and a model generation unit that uses the learning data to generate a trained model for inferring the sound-related data based on the individual data. The compressor includes a compression mechanism for compressing a refrigerant and an electric motor for supplying power to the compression mechanism for compressing the refrigerant. The compression mechanism includes a cylinder, a rolling piston that rotates along the inner circumferential surface of the cylinder based on power from the electric motor, a vane that divides a compression chamber formed by the inner circumferential surface of the cylinder and the outer circumferential surface of the rolling piston into an intake side and a compression side, a vane groove in the cylinder into which the vane is inserted, an upper frame that supports the cylinder from above, a lower frame that supports the cylinder from below, and a shaft inserted into the upper and lower frames to rotate the rolling piston. The shaft includes an upper shaft portion inserted into the upper frame and a lower shaft portion inserted into the lower frame. The individual data includes at least one of the coaxiality between the upper shaft portion and the lower shaft portion and the dimension of the gap between the vane and the vane groove. In the learning device according to the present disclosure, the electric motor may also include a stator, a winding wound around the stator, and a rotor disposed inside the stator. The individual data may include at least one of the amount of magnetic flux of the rotor interlinking with the winding, the roundness of the inner diameter of the stator, and the amount of eccentricity of the rotor.

[0010] A learning device according to the present disclosure is a learning device for performing unsupervised learning, and includes a data acquisition unit that acquires learning data including individual data indicating individual variations of compressors that are correlated with sound-related data related to the sound of the compressor that compresses a refrigerant, and a model generation unit that uses the learning data to generate a trained model for inferring the sound-related data based on the individual data. The compressor includes a compression mechanism for compressing a refrigerant and an electric motor for supplying power to the compression mechanism for compressing the refrigerant. The compression mechanism includes a cylinder, a rolling piston that rotates along the inner circumferential surface of the cylinder based on power from the electric motor, a vane that divides a compression chamber formed by the inner circumferential surface of the cylinder and the outer circumferential surface of the rolling piston into an intake side and a compression side, a vane groove in the cylinder into which the vane is inserted, an upper frame that supports the cylinder from above, a lower frame that supports the cylinder from below, and a shaft inserted into the upper and lower frames to rotate the rolling piston. The shaft includes an upper shaft portion inserted into the upper frame and a lower shaft portion inserted into the lower frame. The individual data includes at least one of the coaxiality between the upper shaft portion and the lower shaft portion and the dimension of the gap between the vane and the vane groove. In the learning device according to the present disclosure, the electric motor may also include a stator, a winding wound around the stator, and a rotor disposed inside the stator. The individual data may include at least one of the amount of magnetic flux of the rotor interlinking with the winding, the roundness of the inner diameter of the stator, and the amount of eccentricity of the rotor. [Effects of the Invention]

[0011] According to the present disclosure, a trained model can be used to infer sound-related data related to the sound of a compressor based on individual variations in the compressor, thereby allowing the user to easily check sound-related data related to the sound of the compressor. [Brief explanation of the drawings]

[0012] [Figure 1] 1 is a diagram illustrating a configuration of a compressor according to an embodiment. [Figure 2] FIG. [Figure 3] FIG. 2 is a cross-sectional view of a compression mechanism portion. [Figure 4] FIG. 1 is a diagram illustrating a configuration of an inference device according to an embodiment. [Figure 5] FIG. 1 is a diagram for explaining an overview of supervised learning. [Figure 6] FIG. 2 is a diagram for explaining input and output of supervised learning in the inference device according to the embodiment. [Figure 7] FIG. 10 is a diagram illustrating a configuration of a learning device in a learning phase. [Figure 8] FIG. 1 is a diagram illustrating a configuration of a neural network. [Figure 9] 10 is a flowchart showing processing executed by a learning device (control unit) in a learning phase. [Figure 10] FIG. 10 is a diagram illustrating the configuration of an inference device in the utilization phase. [Figure 11] 10 is a flowchart showing the processing executed by the inference device (control unit) in the utilization phase. DETAILED DESCRIPTION OF THE INVENTION

[0013] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings. While multiple embodiments will be described below, it is anticipated from the beginning that the configurations described in each embodiment will be appropriately combined. Note that identical or corresponding parts in the drawings will be designated by the same reference numerals, and their description will not be repeated.

[0014] Embodiment [Compressor configuration] A compressor 6 according to an embodiment will be described with reference to Figures 1 to 3. The compressor 6 can be used in an air conditioner that cools or heats an object to be air-conditioned, such as a room, by circulating a refrigerant through a refrigerant circuit. The compressor 6 may also be used in a refrigeration system that cools an object to be cooled, such as a showcase or a unit cooler, by circulating a refrigerant.

[0015] Fig. 1 is a diagram showing the configuration of a compressor 6 according to an embodiment. In Fig. 1, when the compressor 6 is properly installed, the horizontal direction of the compressor 6 is defined as the X-axis direction, the vertical direction of the compressor 6 is defined as the Y-axis direction, and the direction perpendicular to the X-axis and Y-axis is defined as the Z-axis direction. Fig. 1 shows a longitudinal cross section of the compressor 6 taken along the XY plane.

[0016] The compressor 6 is a rotary compressor and includes a housing 60, a compression mechanism 62 for compressing a refrigerant (for example, a refrigerant gas), an electric motor 61 for supplying power to the compression mechanism 62 for compressing the refrigerant, a shaft 613, a glass terminal 67 for supplying power to the electric motor 61, an accumulator 63 for drawing the refrigerant into the compressor 6, and a discharge pipe 66 for discharging the refrigerant compressed by the compressor 6 from inside the housing 60.

[0017] The housing 60 accommodates the electric motor 61, the compression mechanism 62, and the shaft 613. The electric motor 61 is fixed in the housing 60 by press fitting or shrink fitting. Note that the stator 611 of the electric motor 61, which will be described later, may be directly attached to the housing 60 by welding. Inside the housing 60, the compression mechanism 62 is arranged below the electric motor 61. Refrigeration oil is stored at the bottom of the housing 60 to lubricate sliding parts such as a rolling piston 622, which will be described later. The compression mechanism 62 is connected to the electric motor 61 via the shaft 613.

[0018] The accumulator 63 has a suction pipe 64 through which the refrigerant is drawn into the accumulator 63 , and a supply pipe 65 through which the refrigerant is supplied to the compression mechanism 62 .

[0019] Fig. 2 is a diagram showing a cross section of electric motor 61. Fig. 2 shows the cross section of electric motor 61 when electric motor 61 is cut along the XZ plane at line A-A' shown in Fig. 1. As shown in Fig. 2, electric motor 61 includes a stator 611, a winding 615 wound around stator 611, and a rotor 612 arranged inside stator 611. Electric motor 61 is, for example, a PM (Permanent Magnet) motor in which a permanent magnet is provided in rotor 612.

[0020] Stator 611 is formed of an iron core or a coil, and has a circular or nearly circular cross section. A central hole 619 having a circular cross section is formed in the center of stator 611, in which rotor 612 is disposed. Rotor 612 can rotate in a direction along the X-Z plane in central hole 619 formed in stator 611.

[0021] Furthermore, a plurality of slots 614 are formed in the stator 611 along the circumferential direction. A winding 615 is attached to each of the plurality of slots 614. Power is supplied to the winding 615 via glass terminals 67. Note that the winding 615 may be attached to the stator 611 using a well-known winding method such as a distributed winding method or a concentrated winding method, and the method of attaching the winding 615 is not particularly limited.

[0022] The rotor 612 has a circular or nearly circular cross section. The rotor 612 is disposed inside the stator 611 so as to fit into a central hole 619 without contacting the stator 611. A shaft hole 616 having a circular cross section is formed in the center of the rotor 612, for passing the shaft 613 along the Y-axis direction. The rotor 612 also has a plurality of air holes 617 formed surrounding the shaft hole 616. A plurality of permanent magnets 618 are provided outside the plurality of air holes 617. The electric motor 61 is not limited to an interior permanent magnet (IPM) motor in which the permanent magnets 618 are embedded inside the rotor 612, but may also be a surface permanent magnet (SPM) motor in which the permanent magnets 618 are attached to the outer circumferential surface of the rotor 612.

[0023] Fig. 3 is a diagram showing a cross section of the compression mechanism 62. Fig. 3 shows a transverse cross section of the compression mechanism 62 when the compression mechanism 62 is cut along the XZ plane at line B-B' shown in Fig. 1. As shown in Fig. 3, the compression mechanism 62 includes a cylinder 621 and a rolling piston 622 arranged inside the cylinder 621.

[0024] Cylinder 621 has a circular or nearly circular cross section. A compression chamber 630 having a circular cross section and for compressing a refrigerant and accommodating rolling piston 622 is formed in the center of cylinder 621. Rolling piston 622 is rotatable in a direction along the X-Z plane in compression chamber 630 formed in cylinder 621.

[0025] Furthermore, a back pressure chamber 628 and a vane groove 624 are formed in cylinder 621. Vane groove 624 connects compression chamber 630 and back pressure chamber 628. A long vane 625 is provided in vane groove 624. In the example of Fig. 3, vane 625 is slidable along vane groove 624 in the Z-axis direction.

[0026] The rolling piston 622 has a circular or nearly circular cross section. The rolling piston 622 is attached to the outer periphery of an eccentric shaft portion 626 that also has a circular or nearly circular cross section. A shaft hole portion 627 having a circular cross section is formed in the eccentric shaft portion 626 at a position offset from the center of the rolling piston 622 and the eccentric shaft portion 626, for passing the shaft 613 along the Y-axis direction. In other words, the shaft 613 is inserted into the rolling piston 622 and the eccentric shaft portion 626 along the Y-axis direction.

[0027] The tip of the vane 625 ideally contacts a part of the outer circumferential surface of the rolling piston 622, dividing the compression chamber 630 formed by the inner circumferential surface of the cylinder 621 and the outer circumferential surface of the rolling piston 622 into an intake side and a compression side.

[0028] The rolling piston 622 rotates in a direction along the XZ plane in accordance with the rotation of the shaft 613. However, because the shaft 613 is inserted at a position off-center of the rolling piston 622, the rolling piston 622 rotates eccentrically along the inner circumferential surface of the cylinder 621, with the off-center position as its axis. When the rolling piston 622 rotates eccentrically within the cylinder 621, part of the outer circumferential surface of the rolling piston 622 ideally comes into close contact with part of the inner circumferential surface of the cylinder 621.

[0029] As shown in FIG. 1, compression mechanism 62 further includes an upper frame 623A, a lower frame 623B, an upper muffler 624A, and a lower muffler 624B.

[0030] Upper frame 623A and lower frame 623B support cylinder 621 and rolling piston 622 of compression mechanism 62 so as to sandwich them from above and below (in the Y-axis direction). Upper frame 623A supports cylinder 621 and rolling piston 622 by ideally being in close contact with the upper parts of cylinder 621 and rolling piston 622. Lower frame 623B supports cylinder 621 and rolling piston 622 by ideally being in close contact with the lower parts of cylinder 621 and rolling piston 622.

[0031] Furthermore, the upper frame 623A and the lower frame 623B allow the shaft 613 to be inserted along the Y-axis direction, and bearings (not shown) support the shaft 613 rotatably in directions along the X-Z plane. An upper shaft portion 613A constituting a part of the long shaft 613 is inserted into the upper frame 623A, and the shaft 613 is rotatably supported by the upper frame 623A at the upper shaft portion 613A. A lower shaft portion 613B constituting a part of the long shaft 613 is inserted into the lower frame 623B, and the shaft 613 is rotatably supported by the lower frame 623B at the lower shaft portion 613B.

[0032] [Compressor operation] In the compressor 6 configured as described above, when a current flows through the windings 615 of the stator 611 due to the power supplied from the glass terminal 67, a rotating magnetic field is generated in the stator 611. The rotating magnetic field generated in the stator 611 acts on the permanent magnets 618 so as to be attracted to the rotating magnetic field, causing the rotor 612 to rotate. As the rotor 612 rotates, the shaft 613 inserted in the rotor 612 rotates. The rotational force of the shaft 613 is then transmitted to the rolling piston 622, causing the rolling piston 622 to rotate eccentrically along the inner circumferential surface of the cylinder 621.

[0033] The refrigerant drawn into the accumulator 63 is supplied to the compression chamber 630 of the compression mechanism 62 via the supply pipe 65. In the compression chamber 630, the refrigerant is compressed by the rotation of the rolling piston 622. As shown in FIG. 3, the compression chamber 630 includes a suction-side region where the drawn refrigerant is present and a compression-side region where the compressed refrigerant (hereinafter also referred to as "compressed refrigerant") is present. These suction-side and compression-side regions are formed by the outer circumferential surface of the rolling piston 622 contacting the inner circumferential surface of the cylinder 621 and the tips of the vanes. The compressed refrigerant is discharged from the compression-side region and rises inside the housing 60 through the upper muffler 624A. Refrigerating machine oil is mixed into the compressed refrigerant.

[0034] The mixture of compressed refrigerant and refrigerating machine oil is separated into the compressed refrigerant and the refrigerating machine oil when passing through air holes 617 formed in rotor 612. This prevents the refrigerating machine oil from flowing into discharge pipe 66. The compressed refrigerant separated from the refrigerating machine oil is supplied through discharge pipe 66 to the high-pressure side of the refrigerant circuit in which the refrigerant circulates.

[0035] [Correlation between individual compressor variations and compressor noise-related data] The sound-related data related to the sound of the compressor 6 includes the sound pressure level (for example, in decibels) of the sound (noise) generated by the compressor 6 when the compressor 6 is driven. Furthermore, the sound pressure level of the compressor 6 tends to increase as the vibration level, which indicates the degree to which the compressor 6 vibrates when the compressor 6 is driven, increases, and the sound pressure level of the compressor 6 tends to decrease as the vibration level decreases. For this reason, the vibration level of the compressor 6 is closely related to the sound pressure level of the compressor 6. In other words, the sound-related data related to the sound of the compressor 6 includes at least one of the sound pressure level of the compressor 6 and the vibration level of the compressor 6.

[0036] Factors that cause fluctuations in the sound-related data of the compressor 6 include individual variations in the components of the compression mechanism 62. In particular, major factors that affect the sound-related data of the compressor 6 in the compression mechanism 62 include the concentricity of the upper shaft portion 613A and the lower shaft portion 613B, and the dimension of the gap (G1 in FIG. 3) between the vane 625 and the vane groove 624.

[0037] The coaxiality of the upper shaft portion 613A and the lower shaft portion 613B represents the degree of misalignment between the central axes of the upper shaft portion 613A and the lower shaft portion 613B, and when the coaxiality is 0, the central axes of the upper shaft portion 613A and the lower shaft portion 613B completely coincide. In other words, the greater the coaxiality of the upper shaft portion 613A and the lower shaft portion 613B, the greater the misalignment between the rotation centers of the shaft 613 above and below the cylinder 621, which hinders the transmission of rotational energy from the electric motor 61 to the compression mechanism portion 62 and converts the rotational energy into vibrational energy. Therefore, qualitatively, the greater the coaxiality of the upper shaft portion 613A and the lower shaft portion 613B, the greater the sound pressure level of the noise from the compressor 6 and the vibration level of the compressor 6. Thus, variations in the coaxiality between the upper shaft portion 613A and the lower shaft portion 613B tend to deteriorate the noise and vibration characteristics of the compressor 6.

[0038] If the gap between the vane 625 and the vane groove 624 is too large, the vane 625 will easily vibrate in the vane groove 624, and the vane 625 will collide with the vane groove 624, generating vibration energy. That is, if the gap between the vane 625 and the vane groove 624 is too large, qualitatively, the sound pressure level of the noise from the compressor 6 will increase, and the vibration level of the compressor 6 will also increase. On the other hand, if the gap between the vane 625 and the vane groove 624 is too small, vibration energy will be generated due to the frictional force generated between the vane 625 and the vane groove 624. The frictional force generated between the vane 625 and the vane groove 624 causes the rotational speed of the motor 61 to pulsate, making the compressor 6 more likely to vibrate and generate noise. That is, if the gap between the vane 625 and the vane groove 624 is too small, qualitatively, the sound pressure level of the noise from the compressor 6 will increase, and the vibration level of the compressor 6 will also increase. Thus, variations in the gap between the vane 625 and the vane groove 624 tend to deteriorate the noise and vibration characteristics of the compressor 6.

[0039] To prevent deterioration of the noise and vibration characteristics of the compressor 6, each component of the compression mechanism 62 described above is precisely machined and surface-treated before assembly. However, there is inevitably dimensional variation among individual components. This individual variation may result in excessive concentricity between the upper shaft portion 613A and the lower shaft portion 613B, or the gap between the vane 625 and the vane groove 624 exceeding the allowable range. During the manufacture of the compressor 6, the dimensions of each component are inspected to determine whether they are within a predetermined allowable range, and the compressor 6 is manufactured using components whose dimensions are determined to be within the allowable range.

[0040] Compressor 6 is manufactured by assembling the various parts of compression mechanism 62, and it is possible to qualitatively understand that the dimensions of each part affect the noise and vibration characteristics of compressor 6. However, even if the dimensions of each part of compression mechanism 62 are within the allowable range, this does not necessarily mean that the coaxiality of upper shaft portion 613A and lower shaft portion 613B and the dimensions of the gap between vane 625 and vane groove 624 in the assembled compressor 6 are within the allowable range.

[0041] For example, when a component having a dimension close to the lower limit of the tolerance range (e.g., the thickness of the vane 625) is combined with a component having a dimension close to the upper limit of the tolerance range (e.g., the thickness of the vane groove 624), the gap between those components becomes large, and the gap dimension may exceed the tolerance range. Also, when a component having a dimension close to the upper limit of the tolerance range (e.g., the thickness of the vane 625) is combined with a component having a dimension close to the lower limit of the tolerance range (e.g., the thickness of the vane groove 624), the gap between those components becomes small, and the gap dimension may fall below the tolerance range.

[0042] Furthermore, multiple gaps are generated when the components of the compression mechanism 62 are assembled, and these multiple gaps may affect each other. For this reason, even if the gap between two components can be determined to some extent, it is not possible to determine the gaps after the components of the compression mechanism 62 are assembled, and ultimately it is difficult to confirm the sound and vibration levels of the compressor 6 as a whole until after the compressor 6 has been manufactured.

[0043] Another factor that causes fluctuations in the sound-related data of the compressor 6 is individual variations in the components of the motor 61. In particular, the main factors that affect the sound-related data of the compressor 6 in the motor 61 include the amount of magnetic flux of the rotor 612 that interlinks with the windings 615, the roundness of the inner diameter of the stator 611, and the amount of eccentricity of the rotor 612. The amount of magnetic flux of the rotor 612 that interlinks with the windings 615 corresponds to the induced voltage that occurs when the rotating magnetic field generated in the stator 611 interacts with the permanent magnets 618.

[0044] The amount of magnetic flux of the rotor 612 interlinked with the windings 615 varies mainly depending on the saturation magnetic flux density of the permanent magnets 618 inserted in the rotor 612, the dimensions of the permanent magnets 618, the outer diameter of the rotor 612, and the inner diameter of the stator 611. Variations in these factors also tend to cause fluctuations in the input power of the compressor 6 (input power supplied from the glass terminal 67). Generally, to control the electric motor 61, data indicating the amount of magnetic flux of the rotor 612 interlinked with the windings 615 of the stator 611 is input in advance to a control device for the electric motor 61. However, the amount of magnetic flux input in advance is a representative value that does not reflect individual variations in the compressor 6. Therefore, the greater the deviation of the actual amount of magnetic flux from the representative value of the amount of magnetic flux input to the control device, the more likely the input to the compressor 6 will fluctuate, resulting in unstable operation of the electric motor 61. Therefore, the greater the variation in the amount of magnetic flux of the rotor 612 interlinked with the windings 615, the more likely the sound pressure level of the noise of the compressor 6 and the vibration level of the compressor 6 will fluctuate qualitatively. In this way, the variation in the amount of magnetic flux of the rotor 612 interlinked with the windings 615 of the stator 611 tends to deteriorate the noise and vibration characteristics of the compressor 6.

[0045] The inner diameter circularity of the stator 611 indicates whether the circle forming the inner circumferential surface of the stator 611, which has a circular cross section, is close to being a perfect circle. When the inner diameter circularity is 0, the stator 611 is a perfect circle. As shown in FIG. 2 , a gap is generated between the inner circumferential surface of the stator 611 and the outer circumferential surface of the rotating rotor 612. The size of the gap between the inner circumferential surface of the stator 611 and the outer circumferential surface of the rotor 612 increases or decreases depending on the inner diameter circularity of the stator 611. If the size of this gap varies, the magnetic attraction force acting between the stator 611 and the rotor 612 becomes unstable. This qualitatively increases the sound pressure level of the noise from the compressor 6 and the vibration level of the compressor 6. As such, variations in the inner diameter circularity of the stator 611 tend to deteriorate the noise and vibration characteristics of the compressor 6.

[0046] The amount of eccentricity of rotor 612 represents the amount of deviation between the rotation center axis of rotor 612 and the ideal position when the rotation center axis deviates from the ideal position, and when the amount of eccentricity is 0, the rotation center axis is located at the ideal position. As described above, the outer circumferential surface of rolling piston 622 and the inner circumferential surface of cylinder 621 ideally come into close contact with each other, but depending on the amount of eccentricity of rotor 612, a gap (G2 in FIG. 3) may occur between the outer circumferential surface of rolling piston 622 and the inner circumferential surface of cylinder 621. Also, as shown in FIG. 3, the tip of vane 625 and the outer circumferential surface of rotating rolling piston 622 ideally come into close contact with each other, but depending on the amount of eccentricity of rotor 612, a gap (G3 in FIG. 3) may occur between the tip of vane 625 and the outer circumferential surface of rolling piston 622. That is, the size of the gap between the inner circumferential surface of the stator 611 and the outer circumferential surface of the rotor 612 and the size of the gap between the tip of the vane 625 and the outer circumferential surface of the rolling piston 622 increase or decrease depending on the amount of eccentricity of the rotor 612. When the sizes of these gaps fluctuate, the magnetic attraction force acting between the stator 611 and the rotor 612 becomes unstable, and qualitatively, the sound pressure level of the noise of the compressor 6 increases, and the vibration level of the compressor 6 also increases. In this way, variations in the amount of eccentricity of the rotor 612 tend to deteriorate the noise and vibration characteristics of the compressor 6.

[0047] Furthermore, there is an interaction between individual variations in the compression mechanism 62 and individual variations in the electric motor 61. For example, if a large amount of refrigerant leaks from gaps in the compression mechanism 62, less refrigerant is available for compression, resulting in a smaller compression torque and, as a result, a smaller torque generated in the electric motor 61. If the torque of the electric motor 61 is small, this has the effect of reducing the input power to the electric motor 61. Furthermore, the effect on the electric motor 61 varies depending on the location of the gaps generated in the compression mechanism 62.

[0048] As such, the noise and vibration characteristics of the compressor are determined by the complex intertwining of the components in the compression mechanism 62 and the components in the electric motor 61, and therefore it is difficult to accurately confirm the noise and vibration characteristics of the compressor 6 until after the components in the compression mechanism 62 and the components in the electric motor 61 have been combined to manufacture the compressor 6.

[0049] To check the noise and vibration characteristics of the compressor 6, it is sufficient to inspect the compressor 6 after manufacturing. However, the pressure conditions in the inspection equipment must be stabilized each time the compressor 6 is inspected, and inspecting the noise and vibration characteristics of every compressor 6 during the manufacturing process would take a significant amount of time. For this reason, the compressors 6 are typically inspected by random sampling. However, as described above, there are individual variations in the components of the compression mechanism 62 and the electric motor 61. Therefore, when multiple components are combined, the noise and vibration characteristics of the compressor 6 vary from compressor to compressor. In other words, even if the dimensions and performance of each component of the compression mechanism 62 and the electric motor 61 are guaranteed individually, the synergistic effect of combining these components cannot be grasped. Therefore, it is difficult to guarantee the noise and vibration characteristics of every compressor 6 manufactured through random sampling.

[0050] Furthermore, noise and vibration occur in various frequency bands. Therefore, the correlation between noise characteristics and vibration characteristics and individual variations of each component in the compression mechanism 62 and the electric motor 61 also varies depending on the frequency band. Therefore, it is even more difficult to derive the influence of individual variations of each component while taking into account all frequency bands.

[0051] Furthermore, as mentioned above, individual component variations qualitatively affect vibration and noise characteristics, but the tendency of the impact on vibration and noise characteristics can change depending on the combination of individual component variations. For example, the greater the coaxiality between upper shaft portion 613A and lower shaft portion 613B and the greater the eccentricity of rotor 612, the worse the vibration and noise characteristics tend to be. However, depending on the combination of these, the vibration and noise may cancel each other out, actually improving the vibration and noise characteristics. When considering the interaction between such a combination of multiple components, it becomes even more difficult to derive the impact of individual component variations.

[0052] Therefore, the present disclosure provides a technology that uses AI (Artificial Intelligence) to infer sound-related data related to the sound of the compressor 6 based on individual data that indicates individual variations of the compressor 6 regarding the compression mechanism section 62 and the electric motor 61.

[0053] [Inference device] Fig. 4 is a diagram showing the configuration of inference device 10 according to the embodiment. As shown in Fig. 4, inference device 10 includes control unit 11, storage unit 12, and input unit 13 as its main functional components.

[0054] The control unit 11 is a computing entity that executes various processes by executing various programs, and an example thereof is a computer (e.g., a processor). The control unit 11 is configured, for example, with at least one of a CPU (Central Processing Unit), an FPGA (Field-Programmable Gate Array), and a GPU (Graphics Processing Unit). The control unit 11 may also be configured with processing circuitry such as an ASIC (Application Specific Integrated Circuit).

[0055] The storage unit 12 is a memory that provides a storage area for temporarily storing program code or work memory when the control unit 11 executes various programs, and examples thereof include volatile memories such as DRAM (dynamic random access memory) and SRAM (static random access memory), or non-volatile memories such as ROM (read only memory) and flash memory. Furthermore, the storage unit 12 may be a storage device that provides a storage area for storing various data necessary for the control unit 11 to execute various programs, and examples thereof may include storage devices such as SSD (solid state drive) or HDD (hard disk drive).

[0056] The input unit 13 is an interface into which individual data indicating individual variations of the compressor 6 relating to the compression mechanism unit 62 and the electric motor 61 is input.

[0057] The control unit 11 includes a data acquisition unit 111 , a model generation unit 112 , an inference unit 113 , and a presentation unit 114 .

[0058] The data acquisition unit 111 acquires individual data of the compressor 6 via the input unit 13. The model generation unit 112 generates a trained model 20 (described later) for inferring sound-related data based on the individual data, using training data 30 (described later) that is a set of individual data and sound-related data related to the sound of the compressor 6, which is ground truth data corresponding to the individual data. The inference unit 113 infers sound-related data based on the individual data, using the trained model 20. The presentation unit 114 presents a method for adjusting individual variations to the user, based on the sound-related data output by the inference unit 113. For example, the presentation unit 114 displays on the screen to present to the user the optimal combination in the compression mechanism unit 62, the coaxiality of the upper shaft portion 613A and the lower shaft portion 613B in the compression mechanism unit 62, a method for adjusting the dimension of the gap (G1 in Figure 3) between the vane 625 and the vane groove 624, the amount of magnetic flux in the rotor 612 that links with the winding 615 in the electric motor 61, the inner diameter circularity of the stator 611, and a method for adjusting the eccentricity of the rotor 612.

[0059] [Learning Phase] 5 to 9, an application example of the inference device 10 in the learning phase will be described. As described above, the inference device 10 performs supervised learning using learning data 30, which is a set of individual data of the compressor 6 and sound-related data of the compressor 6, which is ground truth data corresponding to the individual data. Supervised learning is a technique for learning features in the learning data 30 using a data set of factors and results (labels), and inferring results from inputs.

[0060] 5 is a diagram for explaining an overview of supervised learning. As shown in Fig. 5, in the learning phase, an inference device 10 executes a learning program 40 to generate (update) a trained model 20 based on training data 30 including an input 1 and an input 2 (correct answer).

[0061] In the utilization phase, the inference device 10 uses the trained model 20 to obtain an output based on the input 1.

[0062] 6 is a diagram for explaining the input and output of supervised learning in inference device 10 according to the embodiment. As shown in Fig. 6, inference device 10 uses individual data of compression mechanism unit 62 and individual data of electric motor 61 as input 1.

[0063] The individual data of the compression mechanism 62 includes at least one of the coaxiality of the upper shaft portion 613A and the lower shaft portion 613B, and the size of the gap between the vane 625 and the vane groove 624 (G1 in FIG. 3).

[0064] The individual data of the electric motor 61 includes at least one of the amount of magnetic flux of the rotor 612 interlinked with the windings 615, the roundness of the inner diameter of the stator 611, and the amount of eccentricity of the rotor 612.

[0065] In the inference device 10, at least one of the sound pressure level and the vibration level, which are sound-related data of the compressor 6, is used as the input 2, which is the correct answer data. That is, at least one of the sound pressure level and the vibration level, which are sound-related data of the compressor 6 assembled using the compression mechanism part 62 and the electric motor 61 having the individual data of the input 1, is used as the input 2, which is the correct answer data. In addition, in the inference device 10, at least one of the sound pressure level and the vibration level, which are sound-related data of the compressor 6, is obtained as an output.

[0066] The individual data of the input 1 described above can be acquired before assembling the compressor 6. Note that all of the individual data shown in FIG. 6 may be used as the input 1, or at least one of the individual data shown in FIG. 6 may be used. For example, only the individual data of the compression mechanism unit 62 may be used as the input 1 data, or only the individual data of the electric motor 61 may be used as the input 1 data. Furthermore, among the individual data of the compression mechanism unit 62, at least one of the coaxiality between the upper shaft portion 613A and the lower shaft portion 613B and the dimension of the gap (G1 in FIG. 3) between the vane 625 and the vane groove 624 may be used as the input 1 data. Among the individual data of the electric motor 61, at least one of the amount of magnetic flux of the rotor 612 interlinked with the winding 615, the inner diameter roundness of the stator 611, and the eccentricity of the rotor 612 may be used as the input 1 data.

[0067] 6, only data that is likely to affect the noise characteristics and vibration characteristics of the compressor 6 may be used as the individual data of the input 1. In this way, the inference device 10 can be trained efficiently.

[0068] 7 is a diagram showing the configuration of the learning device 110 in the learning phase. The learning device 110 is realized by the control unit 11 of the inference device 10. The learning device 110 can exchange data with each of the learning program storage unit 121 and the trained model storage unit 122. The learning program storage unit 121 and the trained model storage unit 122 are realized by the storage unit 12 of the inference device 10.

[0069] 7, the learning device 110 includes a data acquisition unit 111 and a model generation unit 112. The learning device 110 executes a learning program 40 stored in a learning program storage unit 121 to generate a trained model 20 based on training data 30 including an input 1 and an input 2 (correct answer).

[0070] The data acquisition unit 111 acquires learning data 30 including input 1 and input 2 (correct answer). Specifically, the data acquisition unit 111 acquires individual data of the compressor 6 shown in Fig. 6 as input 1. The data acquisition unit 111 acquires sound-related data of the compressor 6 as input 2 (correct answer).

[0071] The model generation unit 112 generates a trained model 20 that infers sound-related data of the compressor 6 based on the individual data of the compressor 6, using training data 30 including input 1 and input 2 (correct answer) acquired by the data acquisition unit 111. The model generation unit 112 stores the generated trained model 20 in the trained model storage unit 122.

[0072] 8 is a diagram showing the configuration of a neural network. The model generation unit 112 generates a trained model 20 by supervised learning, for example, in accordance with a neural network model.

[0073] A neural network consists of an input layer consisting of multiple neurons, an intermediate layer (hidden layer) consisting of multiple neurons, and an output layer consisting of multiple neurons. The intermediate layer may be one layer, or two or more layers.

[0074] Figure 8 shows a three-layer neural network. Figure 8 shows a configuration with three inputs and three outputs. When multiple inputs are input to input layers X1, X2, and X3, the values ​​are multiplied by weights w11 to w16, and the resulting values ​​are input to intermediate layers Y1 and Y2. The results are then further multiplied by weights w21 to w26, and the resulting values ​​are output from output layers Z1, Z2, and Z3. The output results vary depending on the values ​​of the weights w11 to w16 and w21 to w26.

[0075] The neural network performs supervised learning based on learning data 30 including input 1 and input 2 (correct answer) acquired by data acquisition unit 111. That is, the neural network learns by inputting input 1 to the input layer and adjusting the weights so that the result output from the output layer approaches input 2 (correct answer).

[0076] The model generation unit 112 generates the trained model 20 by performing the supervised learning described above.

[0077] Fig. 9 is a flowchart of the processing executed by the learning device 110 (inference device 10) in the learning phase. Note that Fig. 9 shows the processing executed by the inference device 10 corresponding to the learning device 110. Also, in Fig. 9, "S" is used as an abbreviation for "STEP."

[0078] 9, the inference device 10 acquires learning data 30 including input 1 and input 2 (correct answer) by the data acquisition unit 111 (S1). Note that the inference device 10 is not limited to acquiring input 1 and input 2 (correct answer) simultaneously, and may acquire input 1 and input 2 (correct answer) at different times.

[0079] The inference device 10 generates a trained model 20 by performing supervised learning using the model generation unit 112 based on the training data 30 (S2). The inference device 10 stores the generated trained model 20 in the trained model storage unit 122 (S3), and ends this process.

[0080] [Utilization phase] An application example of the inference device 10 in the utilization phase will be described with reference to Figures 10 and 11. Figure 10 is a diagram showing the configuration of the inference device 10 in the utilization phase. The inference device 10 can exchange data with the trained model storage unit 122.

[0081] 10, the inference device 10 includes a data acquisition unit 111 and an inference unit 113. The inference device 10 uses a trained model 20 to obtain an output based on an input 1.

[0082] The data acquisition unit 111 acquires an input 1. Specifically, the data acquisition unit 111 acquires, as the input 1, individual data of the compressor 6 shown in FIG.

[0083] The inference unit 113 uses the trained model 20 to obtain sound-related data of the compressor 6 as an output based on the input 1. Specifically, the inference unit 113 reads out the trained model 20 from the trained model storage unit 122. The inference unit 113 uses the trained model 20 to infer sound-related data of the compressor 6 as an output based on the individual data of the compressor 6, which is the input 1 acquired by the data acquisition unit 111.

[0084] 11 is a flowchart showing the processing executed by the inference device 10 (control unit 11) in the utilization phase. In FIG. 11, "S" is used as an abbreviation for "STEP."

[0085] As shown in FIG. 11, the inference device 10 acquires input 1 by the data acquisition unit 111 (S11). The inference device 10 inputs the acquired input 1 to the trained model 20 (S12). The inference device 10 uses the trained model 20 to infer sound-related data of the compressor 6 as output based on the individual data of the compressor 6, which is input 1 (S13). As a result, the inference device 10 can use the trained model 20 to obtain sound-related data related to the sound of the compressor 6 based on the individual data indicating individual variations of the compressor 6.

[0086] The inference device 10 executes a determination process to determine whether the coefficient of performance of the compressor 6 inferred using the trained model 20 is within the allowable range of the specifications and notify the user of the determination result by displaying it on a screen or the like (S14).

[0087] If the sound pressure level or vibration level included in the inferred sound-related data of the compressor 6 is outside the allowable range of the specifications, the inference device 10 executes a presentation process to present to the user, by displaying on a screen or the like, a method for adjusting the individual variation so that the sound pressure level or vibration level of the compressor 6 falls within the allowable range of the specifications (S15). Thereafter, the inference device 10 ends this process.

[0088] As described above, the inference device 10 according to the embodiment can use the trained model 20 to infer sound-related data related to the sound of the compressor 6 based on the individual data indicating individual variations of the compressor 6 related to the compression mechanism 62 and the electric motor 61. This allows the sound pressure level or vibration level of the compressor 6 to be confirmed based on the individual data that can be obtained before the compressor 6 is assembled, without inspecting the compressor 6, and the sound pressure level or vibration level of the compressor 6 can be easily confirmed. If it is confirmed that the noise characteristics or vibration characteristics are poor, since the compressor 6 has not yet been assembled, the parts can be replaced and the sound-related data of the compressor 6 can be inferred again using the trained model 20 to confirm whether the inferred sound-related data is within the allowable range of the specifications. In this case, it is also possible to present to the user, via a screen display or the like, the dimensions or dimensional range of replacement parts that would bring the sound-related data within the allowable range, present to the user a part replacement method that minimizes the number of replacement parts, and further present to the user the dimensions or dimensional range of the replacement parts. In this case, by prioritizing parts proposals that have dimensions close to the upper or lower limits, or parts that have a large impact on the sound-related data, it is possible to more efficiently present combinations of parts. Furthermore, even if the sound-related data of a compressor 6 is within the tolerance range of the specifications, if it is close to the upper or lower limits of the tolerance range, by subjecting that compressor 6 to inspection, it is possible to prevent compressors 6 with sound-related data that is outside the tolerance range from being released, and to prevent defective products from being released.

[0089] In addition, the inference device 10 determines whether the sound pressure level or vibration level of the compressor 6 inferred using the trained model 20 is within the allowable range of the predetermined specifications and outputs the determination result, so that only compressors 6 whose sound pressure level or vibration level is outside the allowable range of the specifications can be subject to inspection.

[0090] Furthermore, if the inferred sound pressure level or vibration level of the compressor 6 is outside the allowable range of the specifications, the inference device 10 presents a method for adjusting the individual variations so that the sound pressure level or vibration level of the compressor 6 falls within the allowable range of the specifications, thereby improving the manufacturing efficiency of the compressor 6.

[0091] <Modification> The data of each part used in the input 1 may be data obtained by a sampling inspection carried out during the manufacture of the compressor 6. In this way, the more compressors 6 are manufactured and the longer the manufacturing period, the more individual data that can be used for the learning data 30 can be collected.

[0092] The individual data of the input 1 may include data indicating individual variations outside the tolerance range of the compressor 6. Specifically, a combination of components that results in a sound pressure level or vibration level outside the specification range of the compressor 6 may be deliberately produced in advance, and machine learning of the inference device 10 may be performed using learning data including the obtained sound-related data and individual data of each component used. In this way, the inference accuracy of the inference device 10 can be improved by deliberately including individual data indicating individual variations outside the tolerance range in the learning data 30. During mass production of the compressor 6, individual variations of the components tend to converge around the median, so components that deviate from the median are not used. In other words, during mass production of the compressor 6, component dimensions or gaps near the median are more common, and therefore sound pressure levels or vibration levels near the median are also more common during sampling inspection. Therefore, if an attempt is made to infer the sound pressure level or vibration level of the compressor 6 using component dimensions close to the upper or lower limits of the specification, the inference accuracy of the inference device 10 may be reduced. Therefore, as described above, by performing machine learning using individual data that indicates individual variations in the compressor 6 that are outside the tolerance range, it is possible to improve the accuracy of inferring sound-related data for a compressor 6 that uses parts whose dimensions, etc. are close to the upper and lower limits.

[0093] The inference device 10 may be a server device communicatively connected via a network to a control device that controls the compressor 6, or may be a cloud server. The inference device 10 may acquire, as the learning data 30, individual data and sound-related data collected from multiple compressors 6 located in the same area, or individual data and sound-related data collected from multiple compressors 6 located in different areas. In this case, area information may also be included in the learning data 30, allowing machine learning to take differences between areas into consideration. The area may be treated as a different area even if the individual inspection device that inspects the noise and vibration characteristics of the compressor 6 is different. After performing machine learning on a certain compressor 6, machine learning may be performed again on another compressor 6.

[0094] The learning algorithm used in the model generation unit 112 of the inference device 10 may be deep learning, which learns to extract feature quantities themselves, or other known methods. For example, the model generation unit 112 may perform machine learning according to genetic programming, functional logic programming, support vector machines, or the like.

[0095] Although the inference device 10 described above uses supervised learning, it may also use known learning methods such as unsupervised learning, semi-supervised learning, or reinforcement learning. For example, when performing unsupervised learning, the inference device 10 may use only the individual data of the compressor 6 of the input 1 shown in FIG. 6 as the learning data 30. In the learning phase, the inference device 10 clusters the collected individual data to learn the characteristics or trends of the collected individual data. Then, in the utilization phase, the inference device 10 uses the trained model 20 to identify the class to which the input individual data belongs, and output the sound-related data of the compressor 6 corresponding to that class as the inference result.

[0096] <Summary> (Item 1) An inference device 10 according to one embodiment infers sound-related data relating to the sound of a compressor 6 that compresses a refrigerant. The inference device 10 includes a data acquisition unit 111 that acquires individual data indicating individual variations of the compressor 6 that are correlated with the sound-related data, and an inference unit 113 that infers the sound-related data based on the individual data acquired by the data acquisition unit 111, using a trained model 20 for inferring the sound-related data based on the individual data.

[0097] According to the above configuration, the inference device 10 can allow the user to easily check sound-related data related to the sound of the compressor 6 based on individual data that can be obtained before the compressor 6 is assembled, without having to inspect the compressor 6.

[0098] (Clause 2) In the inference device 10 according to clause 1, the compressor 6 includes a compression mechanism 62 for compressing the refrigerant, and an electric motor 61 for supplying power for compressing the refrigerant to the compression mechanism 62. The individual data indicates individual variations of at least one of the compression mechanism 62 and the electric motor 61.

[0099] According to the above configuration, the inference device 10 can allow the user to check the sound-related data of the compressor 6 based on the individual variation of at least one of the compression mechanism 62 and the electric motor 61 of the compressor 6.

[0100] (Item 3) In the inference device 10 according to item 2, the compression mechanism 62 includes a cylinder 621, a rolling piston 622 that rotates along the inner circumferential surface of the cylinder 621 based on power from the electric motor 61, a vane 625 that divides a compression chamber 630 formed by the inner circumferential surface of the cylinder 621 and the outer circumferential surface of the rolling piston 622 into a suction side and a compression side, a vane groove 624 in the cylinder 621 into which the vane 625 is inserted, an upper frame 623A that supports the cylinder 621 from above, a lower frame 623B that supports the cylinder 621 from below, and a shaft 613 that is inserted into the upper frame 623A and the lower frame 623B and rotates the rolling piston 622. The shaft 613 includes an upper shaft portion 613A inserted into the upper frame 623A and a lower shaft portion 613B inserted into the lower frame 623B. The individual data includes at least one of the concentricity between the upper shaft portion 613A and the lower shaft portion 613B and the size of the gap between the vane 625 and the vane groove 624.

[0101] According to the above configuration, the inference device 10 can allow the user to confirm sound-related data of the compressor 6 based on at least one of the concentricity between the upper shaft portion 613A and the lower shaft portion 613B, and the dimension of the gap between the vane 625 and the vane groove 624.

[0102] (4) In the inference device 10 according to the 2nd or 3rd paragraph, the electric motor 61 includes a stator 611, a winding 615 wound around the stator 611, and a rotor 612 provided inside the stator 611. The individual data includes at least one of the amount of magnetic flux of the rotor 612 interlinked with the winding 615, the inner diameter roundness of the stator 611, and the amount of eccentricity of the rotor 612.

[0103] According to the above configuration, the inference device 10 can allow the user to check sound-related data of the compressor 6 based on at least one of the amount of magnetic flux of the rotor 612 interlinked with the winding 615, the inner diameter roundness of the stator 611, and the eccentricity of the rotor 612.

[0104] (Item 5) In the inference device 10 according to items 1 to 4, the sound-related data includes at least one of the sound pressure level of the compressor 6 and the vibration level of the compressor 6.

[0105] According to the above configuration, the inference device 10 can allow the user to confirm at least one of the sound pressure level and the vibration level of the compressor 6 based on the individual data of the compressor 6.

[0106] (Item 6) In the inference device 10 according to items 1 to 5, the sound-related data includes at least one determination result of the sound pressure level of the compressor 6 and the vibration level of the compressor 6.

[0107] According to the above configuration, the inference device 10 can allow the user to confirm the determination result of at least one of the sound pressure level and the vibration level of the compressor 6 based on the individual data of the compressor 6.

[0108] (Clause 7) The inference device 10 according to clauses 1 to 6 further comprises a presentation unit 114 that presents a method for adjusting individual variations based on the sound-related data output by the inference unit 113.

[0109] According to the above configuration, the inference device 10 can present to the user a method for adjusting the individual variations of the compressor 6 based on the inferred sound-related data of the compressor 6, thereby improving the manufacturing efficiency of the compressor 6.

[0110] (Item 8) In the inference device 10 according to items 1 to 7, the trained model 20 is generated by performing machine learning using training data 30 including individual data and sound-related data to infer sound-related data based on the individual data.

[0111] According to the above configuration, the inference device 10 can infer sound-related data of the compressor 6 based on the individual data of the compressor 6, using a trained model 20 generated by machine learning using training data 30 including the individual data of the compressor 6 and sound-related data.

[0112] (Item 9) In the inference device 10 according to items 1 to 8, the individual data included in the learning data 30 includes data indicating individual variations outside the range of tolerance in the compressor 6.

[0113] According to the above configuration, the inference device 10 can improve the accuracy of inferring sound-related data of the compressor 6 even when individual data outside the tolerance range is input.

[0114] (Item 10) An inference method according to one aspect is an inference method for inferring sound-related data relating to the sound of a compressor 6 that compresses a refrigerant by a computer. The inference method includes a step (S11) of acquiring individual data indicating individual variations of the compressor 6 that are correlated with the sound-related data, and a step (S13) of inferring the sound-related data based on the individual data acquired in the acquiring step, using a trained model 20 for inferring the sound-related data based on the individual data.

[0115] According to the above configuration, the computer can allow the user to easily check sound-related data related to the sound of the compressor 6 based on individual data that can be obtained before assembling the compressor 6, without having to inspect the compressor 6.

[0116] (Item 11) A learning device 110 according to one embodiment performs supervised learning. The learning device 110 includes a data acquisition unit 111 that acquires learning data 30 including individual data indicating individual variations of the compressor 6 that are correlated with sound-related data related to the sound of the compressor 6 that compresses a refrigerant, and the sound-related data, and a model generation unit 112 that uses the learning data 30 to generate a trained model 20 for inferring the sound-related data based on the individual data.

[0117] According to the above configuration, the learning device 110 can generate a trained model 20 for inferring sound-related data of the compressor 6 based on the individual data of the compressor 6 by supervised learning using learning data 30 including individual data of the compressor 6 and sound-related data.

[0118] (Item 12) A learning device 110 according to one embodiment performs unsupervised learning. The learning device 110 includes a data acquisition unit 111 that acquires learning data 30 including individual data that indicates individual variations of the compressor 6 that are correlated with sound-related data related to the sound of the compressor 6 that compresses a refrigerant, and a model generation unit 112 that uses the learning data 30 to generate a trained model 20 for inferring sound-related data based on the individual data.

[0119] According to the above configuration, the learning device 110 can generate a trained model 20 for inferring sound-related data of the compressor 6 based on the individual data of the compressor 6 through unsupervised learning using learning data 30 including individual data of the compressor 6.

[0120] The embodiments disclosed herein should be considered to be illustrative in all respects and not restrictive. The scope of the present disclosure is defined by the claims, not by the description of the above embodiments, and is intended to include all modifications within the meaning and scope of the claims. [Explanation of symbols]

[0121] 6 Compressor, 10 Inference device, 11 Control unit, 12 Memory unit, 13 Input unit, 20 Trained model, 30 Training data, 40 Training program, 60 Housing, 61 Electric motor, 62 Compression mechanism unit, 63 Accumulator, 64 Suction pipe, 65 Supply pipe, 66 Discharge pipe, 67 Glass terminal, 110 Learning device, 111 Data acquisition unit, 112 Model generation unit, 113 Inference unit, 114 Presentation unit, 121 Training program memory unit, 122 Trained model memory unit, 611 Stator, 612 Rotor, 613 Shaft, 613A Upper shaft unit, 613B Lower shaft unit, 614 Slot, 615 Winding, 616, 627 Shaft hole unit, 617 Air hole unit, 618 Permanent magnet, 619 Central hole unit, 621 Cylinder, 622 rolling piston, 623A upper frame, 623B lower frame, 624 vane groove, 624A upper muffler, 624B lower muffler, 625 vane, 626 eccentric shaft portion, 628 back pressure chamber, 630 compression chamber.

Claims

1. An inference device that infers sound-related data related to a sound of a compressor that compresses a refrigerant, a data acquisition unit that acquires individual data indicating individual variations of the compressor that are correlated with the sound-related data; an inference unit that infers the sound-related data based on the individual data acquired by the data acquisition unit, using a trained model for inferring the sound-related data based on the individual data, the compressor includes a compression mechanism for compressing the refrigerant, and an electric motor for supplying power to the compression mechanism for compressing the refrigerant, the compression mechanism includes a cylinder, a rolling piston that rotates along the inner peripheral surface of the cylinder based on the power from the electric motor, a vane that divides a compression chamber formed by the inner peripheral surface of the cylinder and the outer peripheral surface of the rolling piston into a suction side and a compression side, a vane groove in the cylinder into which the vane is inserted, an upper frame that supports the cylinder from above, a lower frame that supports the cylinder from below, and a shaft that is inserted into the upper frame and the lower frame and rotates the rolling piston; the shaft includes an upper shaft portion inserted into the upper frame and a lower shaft portion inserted into the lower frame, An inference device, wherein the individual data includes at least one of the concentricity of the upper shaft portion and the lower shaft portion and the dimension of the gap between the vane and the vane groove.

2. An inference device that infers sound-related data related to the sound of a compressor that compresses a refrigerant, a data acquisition unit that acquires individual data indicating individual variations of the compressor that are correlated with the sound-related data; an inference unit that infers the sound-related data based on the individual data acquired by the data acquisition unit, using a trained model for inferring the sound-related data based on the individual data, the compressor includes a compression mechanism for compressing the refrigerant, and an electric motor for supplying power to the compression mechanism for compressing the refrigerant, The electric motor includes a stator, a winding wound around the stator, and a rotor provided inside the stator, The individual data includes at least one of the amount of magnetic flux of the rotor interlinked with the winding, the inner diameter roundness of the stator, and the eccentricity of the rotor.

3. The inference device according to claim 1 or 2, wherein the sound-related data includes at least one of a sound pressure level of the compressor and a vibration level of the compressor.

4. 3. The inference device according to claim 1, wherein the sound-related data includes at least one determination result of a sound pressure level of the compressor and a vibration level of the compressor.

5. The inference device according to claim 1 , further comprising a presentation unit that presents a method for adjusting the individual variations based on the sound-related data output by the inference unit.

6. The inference device according to claim 1, wherein the trained model is generated by performing machine learning using training data including the individual data and the sound-related data to infer the sound-related data based on the individual data.

7. The inference device according to claim 6 , wherein the individual data included in the learning data includes data indicating the individual variation outside a tolerance range in the compressor.

8. An inference method for inferring sound-related data related to a sound of a compressor that compresses a refrigerant by a computer, comprising: acquiring individual data indicating individual variations of the compressor correlated with the sound-related data; and inferring the sound-related data based on the individual data acquired by the acquiring step, using a trained model for inferring the sound-related data based on the individual data, the compressor includes a compression mechanism for compressing the refrigerant, and an electric motor for supplying power to the compression mechanism for compressing the refrigerant, the compression mechanism includes a cylinder, a rolling piston that rotates along the inner peripheral surface of the cylinder based on the power from the electric motor, a vane that divides a compression chamber formed by the inner peripheral surface of the cylinder and the outer peripheral surface of the rolling piston into a suction side and a compression side, a vane groove in the cylinder into which the vane is inserted, an upper frame that supports the cylinder from above, a lower frame that supports the cylinder from below, and a shaft that is inserted into the upper frame and the lower frame and rotates the rolling piston; the shaft includes an upper shaft portion inserted into the upper frame and a lower shaft portion inserted into the lower frame, An inference method, wherein the individual data includes at least one of the concentricity of the upper shaft portion and the lower shaft portion, and the dimension of the gap between the vane and the vane groove.

9. An inference method for inferring sound-related data related to the sound of a compressor that compresses a refrigerant by a computer, comprising: acquiring individual data indicating individual variations of the compressor correlated with the sound-related data; and inferring the sound-related data based on the individual data acquired by the acquiring step, using a trained model for inferring the sound-related data based on the individual data, the compressor includes a compression mechanism for compressing the refrigerant, and an electric motor for supplying power to the compression mechanism for compressing the refrigerant, The electric motor includes a stator, a winding wound around the stator, and a rotor provided inside the stator, An inference method, wherein the individual data includes at least one of the amount of magnetic flux of the rotor interlinked with the winding, the inner diameter roundness of the stator, and the eccentricity of the rotor.

10. A learning device for performing supervised learning, a data acquisition unit that acquires individual data indicating individual variations of the compressor that are correlated with sound-related data related to the sound of the compressor that compresses the refrigerant, and learning data including the sound-related data; a model generation unit that generates a trained model for inferring the sound-related data based on the individual data, using the training data; the compressor includes a compression mechanism for compressing the refrigerant, and an electric motor for supplying power to the compression mechanism for compressing the refrigerant, the compression mechanism includes a cylinder, a rolling piston that rotates along the inner peripheral surface of the cylinder based on the power from the electric motor, a vane that divides a compression chamber formed by the inner peripheral surface of the cylinder and the outer peripheral surface of the rolling piston into a suction side and a compression side, a vane groove in the cylinder into which the vane is inserted, an upper frame that supports the cylinder from above, a lower frame that supports the cylinder from below, and a shaft that is inserted into the upper frame and the lower frame and rotates the rolling piston; the shaft includes an upper shaft portion inserted into the upper frame and a lower shaft portion inserted into the lower frame, The individual data includes at least one of the concentricity of the upper shaft portion and the lower shaft portion, and the dimension of the gap between the vane and the vane groove.

11. A learning device for performing supervised learning, comprising: a data acquisition unit that acquires individual data indicating individual variations of the compressor that are correlated with sound-related data related to the sound of the compressor that compresses the refrigerant, and learning data including the sound-related data; a model generation unit that generates a trained model for inferring the sound-related data based on the individual data, using the training data; the compressor includes a compression mechanism for compressing the refrigerant, and an electric motor for supplying power to the compression mechanism for compressing the refrigerant, The electric motor includes a stator, a winding wound around the stator, and a rotor provided inside the stator, The individual data includes at least one of the amount of magnetic flux of the rotor interlinked with the winding, the inner diameter roundness of the stator, and the eccentricity of the rotor.

12. A learning device for performing unsupervised learning, a data acquisition unit that acquires learning data including individual data that indicates individual variations of the compressor that are correlated with sound-related data related to the sound of the compressor that compresses the refrigerant; a model generation unit that generates a trained model for inferring the sound-related data based on the individual data, using the training data; the compressor includes a compression mechanism for compressing the refrigerant, and an electric motor for supplying power to the compression mechanism for compressing the refrigerant, the compression mechanism includes a cylinder, a rolling piston that rotates along the inner peripheral surface of the cylinder based on the power from the electric motor, a vane that divides a compression chamber formed by the inner peripheral surface of the cylinder and the outer peripheral surface of the rolling piston into a suction side and a compression side, a vane groove in the cylinder into which the vane is inserted, an upper frame that supports the cylinder from above, a lower frame that supports the cylinder from below, and a shaft that is inserted into the upper frame and the lower frame and rotates the rolling piston; the shaft includes an upper shaft portion inserted into the upper frame and a lower shaft portion inserted into the lower frame, The individual data includes at least one of the concentricity of the upper shaft portion and the lower shaft portion, and the dimension of the gap between the vane and the vane groove.

13. A learning device for performing unsupervised learning, comprising: a data acquisition unit that acquires learning data including individual data that indicates individual variations of the compressor that are correlated with sound-related data related to the sound of the compressor that compresses the refrigerant; a model generation unit that generates a trained model for inferring the sound-related data based on the individual data, using the training data; the compressor includes a compression mechanism for compressing the refrigerant, and an electric motor for supplying power to the compression mechanism for compressing the refrigerant, The electric motor includes a stator, a winding wound around the stator, and a rotor provided inside the stator, The individual data includes at least one of the amount of magnetic flux of the rotor interlinked with the winding, the inner diameter roundness of the stator, and the eccentricity of the rotor.

Images