State detection device, state detection method, and computer program

What is AI technical title?
AI technical title is built by Patsnap AI team. It summarizes the technical point description of the patent document.
By designing a numerical acquisition unit and a predicted value determination unit in the state detection device, the difference between the predicted value and the measured value is used to determine the next predicted value using the predicted value or the measured value, the problem of difficulty in accurately determining the abnormal state in the prior art is solved, and accurate state detection is achieved.

CN119856126APending Publication Date: 2025-04-18SHINWA CONTROLS

View PDF 0 Cites 0 Cited by

Patent Information

Application Number: CN202380064688.7
Authority / Receiving Office: CN · China
Patent Type: Applications(China)
Current Assignee / Owner
Priority Date: 2022-11-10
Filing Date: 2023-11-10
Publication Date: 2025-04-18

AI Technical Summary

Technical Problem

In the prior art, when using machine learning-based prediction models for state detection, it is difficult to accurately judge the occurrence of abnormal states, especially when the measured value is used as the input value, it is difficult to effectively evaluate the degree of deviation between the predicted value and the measured value.

Method used

A state detection device and method are designed to obtain continuous specific values through the numerical acquisition unit, and to determine the predicted value of the next specific value based on a plurality of specific values. When the difference between the predicted value and the measured value reaches or exceeds the specified value, the device uses the predicted value to determine the next predicted value; conversely, the measured value to determine the next predicted value.

Benefits of technology

By accurately determining the predicted value for determining the state, the state can be accurately detected, and the occurrence of misjudgment of abnormalities or non-abnormalities can be avoided.

✦ Generated by Eureka AI based on patent content.

Smart Images

Figure CN119856126A_ABST

Patent Text Reader

Abstract

A state detection device (100) according to one embodiment is provided with: a numerical value acquisition unit (101) that acquires an actual opening degree and a target opening degree of a flow rate control valve (16) as specific numerical values that are continuously detected and / or determined; and a predicted value determination unit (102) that determines, on the basis of the plurality of specific values (the actual opening degree and the target opening degree) acquired by the value acquisition unit (101), a predicted value of the actual opening degree that is a specific value next to the plurality of specific values. Furthermore, when the difference between the predicted value corresponding to the actual opening degree and the actual opening degree corresponding to the predicted value is equal to or greater than a predetermined value, the predicted value specifying unit (102) replaces the actual opening degree corresponding to the predicted value with the predicted value, and specifies the next predicted value and the subsequent predicted value.

Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] Embodiments of the present invention relate to a state detection device, a state detection method, and a computer program. Background Art

[0002] Various technologies related to state detection have been proposed so far.

[0003] For example, in temperature control devices such as refrigeration devices and cooling devices that use fluid to control the temperature of a temperature control object, an electric valve that controls the flow of the fluid is sometimes provided. In such a valve, rust, wear, clogging with foreign matter, etc. may cause abnormal operation.

[0004] Detection of possible abnormal operation of the valve as described above has been performed in the past. Specifically, for example, the following method is sometimes used: abnormal operation of the valve is detected based on an abnormal increase in the value of the current supplied to the valve. This method basically detects the occurrence of abnormality by comparing the value of the current supplied to the valve with a preset threshold value, so it can be said to be a simple method.

[0005] On the other hand, in recent years, the technology of detecting the state by using a prediction model (learned model) based on machine learning has become popular. Such technology can be said to be effective in detecting the signs of abnormalities. That is, by detecting the signs before the actual occurrence of obvious abnormalities based on such a prediction model, it is expected that major abnormalities can be avoided. Summary of the invention

[0006] The above-mentioned prediction model is, for example, an algorithm that predicts future measured values using measured values as input values. When such a prediction model is used to determine abnormality, the occurrence of abnormality can be inferred based on the degree of deviation between the predicted value and the measured value.

[0007] However, in a prediction model that uses measured values as input values for prediction, using measured values when an abnormality is occurring or has occurred as input values may make it impossible to accurately evaluate the degree of deviation between the measured values and the predicted values used to infer the occurrence of the abnormality.

[0008] Specifically, for example, the predicted value determined by using the measured value in the case where an abnormality has occurred may become a value close to the corresponding measured value showing the abnormality next time. In this case, it may not be possible to accurately infer the occurrence of the abnormality that should be inferred. On the other hand, when the predicted value is determined based on the predicted value, the predicted value may deviate greatly from the measured value over time. In this case, it is possible to infer the occurrence of an abnormality that should not be inferred.

[0009] In view of the above circumstances, an object of the present invention is to provide a state detection device, a state detection method, and a computer program that can accurately detect a state by accurately determining a prediction value for determining a state such as an abnormality.

[0010] One embodiment of the present invention is related to the following aspects [1] to

[10] .

[0011] [1] A state detection device, comprising: a value acquisition unit, which acquires specific values that are continuously detected and / or determined; and a predicted value determination unit, which determines a predicted value of the next specific value of the multiple specific values based on the multiple specific values acquired by the value acquisition unit, wherein the state detection device detects a state based on the specific values acquired by the value acquisition unit and the predicted value determined by the predicted value determination unit, wherein the predicted value determination unit replaces the corresponding specific value with the predicted value to determine the next predicted value and subsequent predicted values when the difference between the determined predicted value and the specific value corresponding to the predicted value is greater than a specified value, and the predicted value determination unit does not replace the corresponding specific value with the predicted value, but uses the specific value to determine the next predicted value and subsequent predicted values when the difference between the determined predicted value and the specific value corresponding to the predicted value is less than the specified value.

[0012] [2] The state detection device according to [1], wherein the state detection device further comprises: an error evaluation unit, which determines the proportion of the number of times the difference between the predicted value determined by the predicted value determination unit and the specific numerical value corresponding to the predicted value is less than the specified value in multiple predictions performed by the predicted value determination unit within a specified time; and an abnormality notification unit, which notifies an abnormality when the proportion determined by the error evaluation unit is less than the specified proportion.

[0013] [3] The state detection device according to [2], wherein the state detection device further comprises a setting change unit, which changes the prescribed value and the prescribed ratio, and the setting change unit automatically adjusts the prescribed value and the prescribed ratio in an inverse proportional relationship.

[0014] [4] A state detection device according to any one of [1] to [3], wherein the numerical value acquisition unit acquires a command value continuously input to the mechanical element in order to make the mechanical element move and an evaluation value of the action state of the mechanical element continuously detected as the specific numerical value, and the predicted value determination unit determines a predicted value corresponding to the latest evaluation value as the predicted value based on a plurality of command values other than the latest command value as the specific numerical value and a plurality of evaluation values other than the latest evaluation value.

[0015] [5] A state detection device according to [4], wherein the predicted value determination unit replaces the latest evaluation value corresponding to it with the predicted value when the difference between the predicted value corresponding to the latest evaluation value and the latest evaluation value corresponding to the predicted value as a specific numerical value is greater than the specified value, and determines the next predicted value after the numerical value acquisition unit obtains the instruction value and the evaluation value that are next regarded as the latest instruction value and the latest evaluation value; and the predicted value determination unit does not replace the latest evaluation value corresponding to it with the predicted value when the difference between the predicted value corresponding to the latest evaluation value and the latest evaluation value corresponding to the predicted value as a specific numerical value is less than the specified value, but uses the evaluation value to determine the next predicted value after the numerical value acquisition unit obtains the instruction value and the evaluation value that are next regarded as the latest instruction value and the latest evaluation value.

[0016] [6] A state detection device according to [4] or [5], wherein the predicted value determination unit makes a prediction using a learned model, wherein the learned model is a model that completes learning by taking as a normal state a training instruction value continuously input to the mechanical element in order to make the mechanical element act in a normal state and a training evaluation value corresponding to the action state of the mechanical element that acts according to the training instruction value, and the learned model predicts a normal value corresponding to the next training evaluation value based on a plurality of variables corresponding to the training instruction value and a plurality of variables corresponding to the training evaluation value, and the predicted value determination unit determines a predicted value corresponding to the evaluation value in the learned model using a plurality of the instruction values and a plurality of the evaluation values as variables as the specific numerical values.

[0017] [7] A state detection method, comprising: a value acquisition process, which acquires specific values that are continuously detected and / or determined; and a predicted value determination process, which determines a predicted value of the next specific value of the multiple specific values obtained by the value acquisition process. The state is detected based on the specific values obtained by the value acquisition process and the predicted value determined by the predicted value determination process, wherein, in the predicted value determination process, when the difference between the determined predicted value and the specific value corresponding to the predicted value is greater than a specified value, the predicted value is used to replace the specific value corresponding to the predicted value to determine the next predicted value and subsequent predicted values, and in the predicted value determination process, when the difference between the determined predicted value and the specific value corresponding to the predicted value is less than the specified value, the predicted value is not used to replace the specific value corresponding to the predicted value, but the specific value is used to determine the next predicted value and subsequent predicted values.

[0018] [8] The state detection method according to [7], wherein the state detection method further comprises a learning model preparation step of preparing a learning model, wherein the learning model is learned using a training instruction value continuously input to the mechanical element in order to make the mechanical element in a normal state to act and a training evaluation value corresponding to the action state of the mechanical element that acts according to the training instruction value as a normal state, and predicting a normal value corresponding to the next training evaluation value based on a plurality of variables corresponding to the training instruction value and a plurality of variables corresponding to the training evaluation value, and in the value acquisition step, obtaining the instruction value continuously input to the mechanical element in order to make the mechanical element act and the evaluation value of the action state of the mechanical element that is continuously detected as the specific numerical value, and in the predicted value determination step, in the learning model, determining a predicted value corresponding to the evaluation value using a plurality of the instruction values and a plurality of the evaluation values as the specific numerical values as variables.

[0019] [9] A state detection device that detects the state of a valve whose opening changes according to an input command signal, wherein the state detection device comprises: a numerical acquisition unit that acquires information on the actual opening of the valve as a specific numerical value that is continuously detected and / or determined; and a predicted value determination unit that determines a predicted value of the next actual opening of the multiple actual openings based on the information on the multiple actual openings acquired by the numerical acquisition unit, wherein the state detection device detects the state based on the information on the actual opening of the valve acquired by the numerical acquisition unit and the predicted value of the actual opening determined by the predicted value determination unit, wherein the predicted value determination unit replaces the actual opening corresponding to the predicted value with the predicted value when the difference between the determined predicted value and the actual opening corresponding to the predicted value is greater than a specified value and determines the next predicted value and subsequent predicted values, and wherein the predicted value determination unit does not replace the actual opening corresponding to the predicted value with the predicted value and determines the next predicted value and subsequent predicted values using the actual opening when the difference between the determined predicted value and the actual opening corresponding to the predicted value is less than the specified value.

[0020]

[10] A computer program for causing a computer to execute the following steps: a numerical value acquisition step, acquiring a specific numerical value that is continuously detected and / or determined; and a predicted value determination step, determining a predicted value of the next specific numerical value of the multiple specific numerical values based on the multiple specific numerical values acquired through the numerical value acquisition step, detecting a state based on the specific numerical value acquired through the numerical value acquisition step and the predicted value determined through the predicted value determination step, wherein, in the predicted value determination step, when the difference between the determined predicted value and the specific numerical value corresponding to the predicted value is greater than a specified value, the predicted value is used to replace the specific numerical value corresponding to the predicted value to determine the next predicted value and subsequent predicted values, and in the predicted value determination step, when the difference between the determined predicted value and the specific numerical value corresponding to the predicted value is less than the specified value, the predicted value is not used to replace the specific numerical value corresponding to the predicted value, but the specific numerical value is used to determine the next predicted value and subsequent predicted values.

[0021] According to the present invention, it is possible to accurately detect a state by accurately determining a predicted value for judging a state. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a diagram schematically showing a state detection device according to an embodiment and a temperature control device to which the state detection device is applied.

[0023] Figure 2 It is shown Figure 1 A block diagram of the functional structure of a state detection device.

[0024] Figure 3 Yes Figure 1 A flowchart for explaining an example of the prediction operation of the state detection device.

[0025] Figure 4 Yes Figure 1 A flowchart for explaining an example of the state detection operation of the state detection device.

[0026] Figure 5 It is conceptually shown Figure 3 Graph of predicted actions.

[0027] Figure 6 Is to explain the decision used for Figure 3 and Figure 4 The diagram shows the process of a predetermined value evaluated as a difference between a predicted value calculated in the operation and a specific numerical value of an actual measurement value. DETAILED DESCRIPTION

[0028] Hereinafter, one embodiment will be described.

[0029] <Structure of the state detection device and the temperature control device>

[0030] Figure 1 This is a diagram schematically showing a state detection device 100 according to one embodiment and a refrigeration device 1 as a temperature control device to which the state detection device 100 is applied.

[0031] The refrigeration device 1 has a refrigeration circuit 10, which has a compressor 11, a condenser 12, an expansion valve 13, and an evaporator 14. In the refrigeration circuit 10, the compressor 11, the condenser 12, the expansion valve 13, and the evaporator 14 are connected in this order through a piping member (pipe) so that a refrigerant circulates. The refrigeration device 1 controls the temperature of a temperature-controlled object by exchanging heat between the refrigerant flowing in the evaporator 14 and the temperature-controlled object not shown.

[0032] The compressor 11 compresses the refrigerant in a substantially gaseous state flowing out of the evaporator 14 and supplies the refrigerant to the condenser 12 in a state of increased temperature and pressure. The condenser 12 cools and condenses the refrigerant compressed by the compressor 11 using cooling water, and makes the refrigerant a high-pressure liquid at a predetermined temperature and supplies it to the expansion valve 13. The expansion valve 13 expands the refrigerant from the condenser 12 so that the refrigerant is in a gas-liquid mixed phase state and flows into the evaporator 14. The refrigerant flowing into the evaporator 14 from the expansion valve 13 is cooled by expansion. Thus, the evaporator 14 cools the temperature-controlled object, for example, by exchanging heat between the refrigerant and the temperature-controlled object.

[0033] The condenser 12 is connected to a cooling water flow path 15. A flow regulating valve 16 is provided on the cooling water flow path 15. The cooling water flow path 15 is a flow path for cooling water to flow into the condenser 12 and for cooling water flowing out of the condenser 12 to flow. The flow regulating valve 16 regulates the flow rate of cooling water flowing from the cooling water flow path 15 into the condenser 12. In the present embodiment, the flow regulating valve 16 is provided at a portion of the cooling water flow path 15 on the upstream side of the condenser 12, but may also be provided at the downstream side of the condenser 12.

[0034] The flow control valve 16 is a proportional valve that adjusts the flow rate of cooling water flowing into the condenser 12 by adjusting the opening. The flow control valve 16 of this embodiment has a motor 16M, and the opening of the flow control valve 16 is adjusted by adjusting the position of the valve body by the motor 16M. In addition, the flow control valve 16 has an opening detection unit 16S, and the opening of the valve can be determined by using the opening detection unit 16S. The opening detection unit 16S can include an encoder that can identify the rotational change of the rotating shaft of the motor 16M from a reference position, and the opening of the valve can be determined based on the rotational change detected by the encoder. In addition, the motor 16M of the flow control valve 16 is a stepping motor in this embodiment, but it can also be a DC servo motor or an AC servo motor. In addition, the flow control valve 16 can also be a proportional solenoid valve.

[0035] In detail, the motor 16M of the flow control valve 16 is electrically connected to the controller 17, and the opening degree of the flow control valve 16 is controlled by the controller 17. In the present embodiment, a pressure sensor 18 is provided, and the pressure sensor 18 detects the pressure of the refrigerant flowing in the part downstream of the condenser 12 of the refrigeration circuit 10 and upstream of the expansion valve 13. Information on the pressure detected by the pressure sensor 18 is provided to the controller 17. In addition, the controller 17 of the present embodiment controls the flow control valve 16 in such a manner that the pressure detected by the pressure sensor 18 is consistent with a preset target pressure.

[0036] The controller 17 controls the flow control valve 16 by feedback control based on the difference between the pressure detected by the pressure sensor 18 and the target pressure so that the pressure detected by the pressure sensor 18 is consistent with the target pressure. The controller 17 controls the flow control valve 16 by PID control, but it can also be controlled by PI control, PD control, etc. or other methods. In addition, a temperature sensor can be provided instead of the pressure sensor 18, so that the controller 17 controls the flow control valve 16 in a manner that the temperature detected by the temperature sensor is consistent with the target temperature.

[0037] The controller 17 calculates the target opening (value in the range of 0% to 100%) as the command value and outputs it to the flow control valve 16. In the present embodiment, the controller 17 is electrically connected to the motor 16M as a stepping motor, so the target opening output by the controller 17 is actually converted into a group of pulse signals and output. In addition, the group of pulse signals corresponding to the target opening is input to the flow control valve 16 as a command signal. Then, the motor 16M changes the rotation position according to the output pulse signal to adjust the opening of the flow control valve 16.

[0038] The controller 17 calculates the target opening at a predetermined period, and then generates a group of pulse signals corresponding to the target opening. The controller 17 increases or decreases the target opening as a command value and performs calculations based on the difference between the pressure detected by the pressure sensor 18 and the target pressure. The group of pulse signals can also be said to be a signal waveform that generates one or more pulse signals in a time width corresponding to the predetermined period. The number of pulse signals included in the generated group of pulse signals increases or decreases based on the difference between the pressure detected by the pressure sensor 18 and the target pressure.

[0039] In more detail, the larger the difference is, the more pulse signals the controller 17 includes in the group of pulse signals. In this case, the more pulse signals are included in the group of pulse signals received by the motor 16M, the faster the opening of the flow control valve 16 changes toward the target opening. In this embodiment, when the flow control valve 16 receives a group of pulse signals from the controller 17, it is intended to operate in the time width corresponding to the group until the target opening is reached. However, the control method of the flow control valve 16 by the controller 17 is not limited to the above example, and other methods may also be used.

[0040] The controller 17 may be a computer or microcomputer including a CPU, a ROM, etc. In this case, the controller 17 executes various processes according to the program stored in the ROM. In addition, the controller 17 may also be another processor or circuit (for example, FPGA (Field Programmable Gate Array: Field Programmable Gate Array)).

[0041] The state detection device 100 is electrically connected to the controller 17. In addition, the state detection device 100 is provided with the command value (in this example, "target opening") output by the controller 17 to the motor 16M and the information of the current opening of the flow control valve 16. As described above, in this embodiment, the controller 17 calculates the target opening at a predetermined period. The state detection device 100 of this embodiment is provided with the target opening information each time the controller 17 calculates the target opening, and is provided with the information of the actual opening of the flow control valve 16 at the time when the target opening information is provided.

[0042] Then, the state detection device 100 predicts the next actual opening of the plurality of actual opening information based on the information of the opening of the plurality of targets as specific values that are continuously detected and / or determined and the information of the plurality of actual openings. Furthermore, the state detection device 100 detects the state of the flow control valve 16, specifically, detects abnormality, based on the predicted opening of the flow control valve 16 and the corresponding actual opening of the flow control valve 16.

[0043] The state detection device 100 may be composed of a computer or a microcomputer including a CPU, a ROM, etc. In this case, the state detection device 100 performs various processes according to the program stored in the ROM. In addition, the controller 17 may also be another processor or circuit (for example, FPGA (Field Programmable Gate Array)). In this embodiment, the controller 17 and the state detection device 100 are different devices, but they may also be integrated. The state detection device 100 is described below.

[0044] <Functional Structure of Status Detection Device>

[0045] Figure 2 1 is a block diagram showing the functional structure of the state detection device 100. The state detection device 100 includes a numerical value acquisition unit 101, a predicted value determination unit 102, an error evaluation unit 103, an abnormality notification unit 104, a setting change unit 105, and a storage unit 106. When the state detection device 100 is constituted by a computer, for example, the numerical value acquisition unit 101, the predicted value determination unit 102, the error evaluation unit 103, the abnormality notification unit 104, and the setting change unit 105 can be realized by executing a program stored in a ROM. On the other hand, the storage unit 106 may be constituted by a part of the ROM.

[0046] The numerical value acquisition unit 101 acquires information on the target opening of the flow control valve 16 calculated by the controller 17 and information on the actual opening of the flow control valve 16 detected by the opening detection unit 16S as specific numerical values that are continuously detected and / or determined.

[0047] Here, the target opening information corresponds to the command value continuously input to the flow control valve 16 in order to operate the flow control valve 16 as a mechanical element. The actual opening information corresponds to the evaluation value of the continuously detected operating state of the flow control valve 16. The target opening information and the actual opening information acquired by the value acquisition unit 101 are sequentially stored in the prediction variable memory 106A described later set in the storage unit 106.

[0048] The predicted value determination unit 102 is a part that determines the predicted value of the next actual opening of the multiple actual openings based on the information of the openings of the multiple targets and the information of the multiple actual openings acquired by the numerical value acquisition unit 101 and stored in the prediction variable memory 106A as described above. In detail, the predicted value determination unit 102 of the present embodiment determines the predicted value corresponding to the latest actual opening information based on the information of the openings of the multiple targets other than the latest target opening information and the information of the multiple actual openings other than the latest actual opening information in the prediction variable memory 106A. In the present embodiment, when the number of the opening information of the multiple targets used by the predicted value determination unit 102 is n, the number of the multiple actual opening information is also n. The above n is an integer and can be 4 to 10.

[0049] Furthermore, the prediction value determination unit 102 of the present embodiment is configured to, after determining the prediction value as described above, replace the actual opening information used in the next prediction with the prediction value as necessary based on the difference between the prediction value and the latest actual opening corresponding thereto.

[0050] Specifically, when the difference between the predicted value corresponding to the latest actual opening and the latest actual opening corresponding to the predicted value as an actual measurement value is greater than a predetermined value, the predicted value determination unit 102 replaces the latest actual opening corresponding to the predicted value with the predicted value and determines the next predicted value and subsequent predicted values. On the other hand, when the difference between the predicted value corresponding to the latest actual opening and the latest implemented opening corresponding to the predicted value as an actual measurement value is less than a predetermined value, the predicted value determination unit 102 does not replace the latest actual opening corresponding to the predicted value, but uses the actual opening to determine the next predicted value and subsequent predicted values. The predicted value replaced with respect to the actual opening is maintained in subsequent predictions.

[0051] The state detection device 100 detects an abnormality based on a state in which the difference between the predicted value of the detection opening and the corresponding measured value of the opening is large, which will be described later. Here, when the predicted value is determined based on an abnormal measured value, sometimes the predicted value no longer shows normal behavior, and it may not be possible to accurately detect the abnormality. Therefore, it is preferred that the predicted value shows normal behavior. Therefore, the predicted value determination unit 102 replaces the corresponding measured value with the predicted value as needed as described above and performs subsequent predictions.

[0052] In the present embodiment, as an example, the above-mentioned prediction value determination unit 102 is configured to perform prediction using a learned model. The learned model is learned by taking the training command value (target opening) continuously input to the flow control valve 16 in order to operate the flow control valve 16 in a normal state and the training evaluation value (actual opening) corresponding to the operating state of the mechanical element operated according to the training command value as a normal state (normal data group).

[0053] In detail, the above-mentioned learned model is sequentially input with training instruction values and training evaluation values, and a plurality of training instruction values other than the latest training instruction value and a plurality of training evaluation values other than the latest training evaluation value among the plurality of training evaluation values are used as explanatory variables, and the latest training evaluation value is used as a correct answer label to perform machine learning. Furthermore, the learned model is constructed to predict a normal value corresponding to the next training evaluation value based on a plurality of variables corresponding to the above-mentioned training instruction values and a plurality of variables corresponding to the above-mentioned training evaluation values.

[0054] Therefore, the learned model used in the present embodiment is constructed as follows: the target openings as multiple instruction values obtained by the numerical acquisition unit 101 and the actual openings as multiple evaluation values are used as explanatory variables, and the predicted value corresponding to the actual opening is determined as a value consistent with the correct answer label.

[0055] As an example, the learned model used by the prediction value determination unit 102 as described above is constructed by machine learning based on LightGBM (Light Gradient Boosting Machine). However, the learned model used by the prediction value determination unit 102 is not particularly limited, and can also be constructed by other machine learning such as neural networks and SVM. In addition, the prediction value determination unit 102 can also determine the prediction value by a method such as regression.

[0056] In the present embodiment, the predicted value of the actual opening determined by the predicted value determination unit 102 and the difference between the predicted value and the corresponding actual opening are sequentially stored in a storage memory 106B described later set in the storage unit 106. Such information can be used to improve the learned model used by the predicted value determination unit 102 later.

[0057] Furthermore, in the present embodiment, when the predicted value of the actual opening determined by the predicted value determination unit 102 and the difference between the predicted value and the corresponding actual opening are maintained in the storage memory 106B or after being maintained in the storage memory 106B, when the difference is less than a prescribed value, it is associated with a flag (for example, "1") indicating the situation (in other words, normal) and maintained; when the difference is greater than a prescribed value, it is associated with a flag (for example, "0") indicating the situation (in other words, abnormal) and maintained.

[0058] The error evaluation unit 103 is a part that determines the ratio of the number of times that the difference between the predicted value of the actual opening determined by the predicted value determination unit 102 and the actual opening corresponding to the predicted value is less than a predetermined value in the multiple predictions performed by the predicted value determination unit 102 within a predetermined time. Specifically, the error evaluation unit 103 determines the ratio of the number of differences associated with a flag (e.g., "1") indicating that the difference is less than the predetermined value in the predetermined number (e.g., 3600) of the information of the above-mentioned differences stored in the storage memory 106B as described above. Thus, the error evaluation unit 103 determines the ratio of the number of times that the difference between the predicted value and the measured value is less than the predetermined value in the multiple predictions performed by the predicted value determination unit 102 within a predetermined time.

[0059] When the ratio of the number of times the difference between the predicted value and the actual opening degree is smaller than a predetermined value in a plurality of predictions performed by the predicted value determination unit 102 within a predetermined time is smaller than a predetermined ratio, the error evaluation unit 103 determines that an abnormality has occurred.

[0060] When the error evaluation unit 103 determines that an abnormality has occurred, it provides alarm information to the abnormality notification unit 104. The abnormality notification unit 104 notifies the abnormality upon receiving the alarm information from the error evaluation unit 103. The notification of the abnormality may be performed by displaying a warning on a display (not shown) or by generating an alarm sound.

[0061] The setting change unit 105 is a part that changes the above-mentioned prescribed value used by the prediction value determination unit 102 to evaluate the difference and the above-mentioned prescribed ratio used by the prediction value determination unit 102 in determining whether to notify an abnormality. The setting change unit 105 of this embodiment changes the prescribed value and the prescribed ratio in conjunction with the user's operation. In detail, the setting change unit 105 automatically adjusts the prescribed value and the prescribed ratio according to an inverse proportional relationship.

[0062] In the present embodiment, the predicted value of the actual opening determined by the predicted value determination unit 102 and the actual opening corresponding to the predicted value are determined by any value between fully closed 0% and fully open 100%. Here, the above-mentioned prescribed value is set to 1% in the unit of the opening of the flow control valve 16, and the above-mentioned prescribed ratio for determining whether to notify an abnormality is set to 60%. In the case where the prescribed value is increased to 2% using the setting change unit 105 from this state, for example, the above-mentioned prescribed ratio is automatically reduced to, for example, 40%.

[0063] The combination of the prescribed value and the prescribed ratio set by the setting change unit 105 is predetermined as a combination of optimal values that can detect abnormalities with high reliability based on examples in which abnormalities actually occur. According to such a setting change unit 105, it is possible to flexibly perform desired state detection according to the situation. However, the prescribed value and the prescribed ratio may be set arbitrarily.

[0064] The storage unit 106 sets the above-mentioned prediction variable memory 106A and storage memory 106B in a part of the data holding part, and holds the information in each memory. Here, when the number of information on the openings of the plurality of targets used by the prediction value determination unit 102 is set to n, and the number of information on the plurality of actual openings is also set to n, the prediction variable memory 106A can be configured to have n+1 storage areas for the target openings and n+1 storage areas for the plurality of actual openings.

[0065] In the prediction variable memory 106A as described above, the target opening information initially acquired by the numerical acquisition unit 101 is stored in the n+1th storage area regarding the target opening, and the actual opening information initially acquired by the numerical acquisition unit 101 is stored in the n+1th storage area regarding the actual opening. Then, before acquiring the next target opening information, the target opening information stored in the n+1th storage area is moved to the nth storage area, and the new target opening information is stored in the n+1th storage area. The actual opening information is also stored in the same manner as the target opening information. Then, after the first to n+1th storage areas regarding the target opening are filled with the target opening information, when the next information is acquired, the information in the first storage area is discharged. The same discharge process is performed for the actual opening. In addition, the structure of the prediction variable memory 106A is not particularly limited, as long as it has at least a storage area capable of storing information used by the prediction value determination unit 102 in the prediction.

[0066] The storage unit 106 as described above may be constituted by a main storage device or an auxiliary storage device. The storage unit 106 stores various operation programs.

[0067] <Predicted Action>

[0068] Next, refer to Figure 3 The flowchart shown in the figure explains an example of the prediction operation of the state detection device 100. The prediction operation of the state detection device 100 starts when the operation of the refrigeration device 1 starts.

[0069] First, in step S31, the state detection device 100 obtains information on the target opening of the flow control valve 16 calculated by the controller 17 and information on the actual opening of the flow control valve 16 detected by the opening detection unit 16S from the controller 17. The information is obtained by the numerical acquisition unit 101. And the target opening information and the actual opening information obtained by the numerical acquisition unit 101 are sequentially stored in the prediction variable memory 106A.

[0070] Next, in step S32, the state detection device 100 determines whether the prediction variable memory 106A is filled with the target opening information and the actual opening information. Specifically, the state detection device 100 determines whether the target opening information is filled in the first to n+1 storage areas related to the target opening and the actual opening information is filled in the first to n+1 storage areas related to the actual opening in the prediction variable memory 106A.

[0071] Furthermore, when it is determined in step S32 that the prediction variable memory 106A is not full, the acquisition of information in step S31 is repeated. On the other hand, when it is determined in step S32 that the prediction variable memory 106A is full, in step S33, the state detection device 100 determines the predicted value of the information on the actual opening degree through the predicted value determination unit 102.

[0072] In step S33, the prediction value determination unit 102 determines the prediction value corresponding to the latest actual opening information based on the information of the openings of the plurality of targets other than the latest information of the openings of the target and the information of the plurality of actual openings other than the latest information of the actual opening. That is, the prediction value determination unit 102 determines the prediction value of the actual opening held (n+1) based on the information of the openings of the first to nth targets among the information of the openings of the target held in the first to n+1th storage areas of the prediction variable memory 106A and the information of the openings of the first to nth targets among the information of the actual openings held in the first to n+1th storage areas of the prediction variable memory 106A, and holds it in the storage memory 106B.

[0073] Then, in step S34, the predicted value determination unit 102 calculates the difference between the predicted value determined in step S33 and the actual opening corresponding to the predicted value as the actual measured value. That is, the difference between the predicted value determined in step S33 and the latest actual opening stored in the n+1th storage area of the prediction variable memory 106A is calculated. And the difference is stored in the storage memory 106B.

[0074] Next, in step S35, the information stored in the first storage area of the prediction variable memory 106A regarding the target opening and the actual opening is ejected, and the other information is moved from the current storage area to the storage area one level lower in order, leaving the n+1th storage area vacant.

[0075] Then, in step S36, the prediction value determination unit 102 determines whether the difference calculated in step S34 is greater than or equal to a predetermined value. Then, when the difference is greater than or equal to a predetermined value in step S36 (yes in S36), in step S37, the prediction value determination unit 102 replaces the latest actual opening degree stored in the current nth storage area of the prediction variable memory 106A corresponding to the latest actual opening degree with the prediction value corresponding to the latest actual opening degree determined in step S33 and stores it. Then, after the processing of step S37, the processing returns to step S31, and the information of the next target opening degree and the information of the actual opening degree are acquired.

[0076] On the other hand, when the difference is smaller than the predetermined value in step S36 (No in S36), the predicted value determination unit 102 does not replace the latest actual opening degree stored in the current n-th storage area of the corresponding prediction variable memory 106A with the predicted value corresponding to the latest actual opening degree determined in step S33. That is, the information of the actual opening degree as the measured value continues to be stored in the prediction variable memory 106A. After that, the process returns to step S31, and the information of the next target opening degree and the information of the actual opening degree are acquired.

[0077] <Status detection action>

[0078] Below, refer to Figure 4 An example of the state detection operation of the state detection device 100 will be described. The state detection operation of the state detection device 100 starts when the operation of the refrigeration device 1 starts, and is performed in parallel with the above-mentioned prediction operation.

[0079] As described above, in the prediction operation, the prediction value determination unit 102 calculates the difference between the prediction value determined in step S33 and the actual opening degree corresponding to the prediction value as the actual measured value, and stores it in the storage memory 106B (step S34). In the state detection operation, first in step S41, the state detection device 100 monitors whether the difference calculated in step S34 is stored in the storage memory 106B. And, the monitoring is continued until it is confirmed that the difference is stored.

[0080] Then, when the difference is held in the storage memory 106B, in step S42, the state detection device 100 determines whether the difference is normal based on whether it is equal to or greater than the predetermined value used in the determination in step S36.

[0081] Furthermore, when the difference is smaller than a predetermined value in the determination of step S42 (Yes in S42), in step S43, as an example, in this operation example, a flag indicating 1 (a flag indicating normality) is associated with the difference, and the flag indicating 1 is held together with the difference in the storage memory 106B. On the other hand, when the difference is greater than a predetermined value in the determination of step S42 (No in S42), in step S44, as an example, in this operation example, a flag indicating 0 (a flag indicating abnormality) is associated with the difference, and the flag indicating 0 is held together with the difference in the storage memory 106B.

[0082] Then, in step S45 after the processing of step S43 and step S44, the error evaluation unit 103 determines the ratio of the number of times that the difference between the predicted value of the actual opening determined by the predicted value determination unit 102 and the actual opening corresponding to the predicted value is less than the prescribed value (normal) in the multiple predictions performed by the predicted value determination unit 102 within the prescribed time. Specifically, the error evaluation unit 103 determines the ratio of the number of differences associated with the flag "1" indicating that the difference is less than the prescribed value among the information of the prescribed number (for example, 3600) of the above differences held in the storage memory 106B. In this way, the ratio of the number of times that the difference between the predicted value and the actual opening is less than the prescribed value in the multiple predictions performed by the predicted value determination unit 102 within the prescribed time can be determined.

[0083] Then, in step S46, the error evaluation unit 103 determines whether the ratio determined in step S45 is greater than or equal to a prescribed ratio, in other words, whether it is less than or equal to a prescribed ratio. Then, when the ratio determined in step S45 is greater than or equal to a prescribed ratio (yes in S46), it is considered that no abnormality has occurred, and the processing from step S41 is repeated. On the other hand, when the ratio determined in step S45 is less than the prescribed ratio (no in S46), an alarm is notified by the abnormality notification unit 104 in step S47 as an abnormality has occurred. Thereafter, the processing from step S41 is repeated.

[0084] In addition, in this embodiment, when the ratio of the normal state (the mark of 1) is less than the prescribed ratio, it is determined that an abnormality has occurred, but it may also be that when the ratio of the abnormal state (the mark of 0) is greater than the prescribed ratio, it is determined that an abnormality has occurred.

[0085] <An example of predicted action>

[0086] Hereinafter, the concept of the above-mentioned prediction operation will be described with reference to the drawings. Figure 5 is conceptually shown in Figure 3 As described above, when the difference between the predicted value corresponding to the latest actual opening and the latest actual opening corresponding to the predicted value as an actual measured value is greater than a specified value, the predicted value determination unit 102 replaces the latest actual opening corresponding to the predicted value with the predicted value to determine the next predicted value and subsequent predicted values.

[0087] exist Figure 5 In the graphs shown in (A) and (B) of FIG. 1 , the horizontal axis represents time, and the vertical axis represents the opening degree (%) of the flow rate control valve 16 . Figure 5 (A) shows the prediction operation when the prediction value is determined using only the actual opening degree, which is different from the present embodiment. Figure 5 (B) shows the prediction operation of this embodiment, and shows the prediction operation when the prediction value is used to determine the subsequent prediction value according to the situation. Figure 5 (A) and Figure 5 In (B), an abnormality occurs 2 hours or later from the reference time 0. Figure 5 (A) and Figure 5 In (B), the target opening (MV) and the actual opening (PV) are common values, but the states of the openings (P_PV) as their respective predicted values change in different states.

[0088] Specifically, when determining the predicted value using only the measured value, which is different from the present embodiment, as shown in FIG. Figure 5As shown in (A), the predicted value (P_PV) sometimes changes to a value that deviates greatly from the target opening (MV). When the flow control valve 16 is normal, the actual opening (PV) and the target opening (MV) are close to each other. Therefore, Figure 5 The predicted value determined 2 hours later and thereafter as shown in (A) does not show a normal state. In this embodiment, it is predetermined that the predicted value is calculated as a normal value, and when the deviation from the measured value is large, an abnormality is inferred. Figure 5 The predicted value determined in (A) is not preferable in the case of determining the detection of abnormality. Figure 5 In (A), the following situation occurs: although an abnormality occurs, the difference ΔD between the predicted value and the measured value 4 hours later is determined to be smaller than the predetermined value Th, and is determined to be a normal state.

[0089] On the other hand, in this embodiment, when the predicted value deviates greatly from the measured value, the predicted value is used instead of the measured value when determining the next predicted value. Figure 6 As shown in (B), the predicted value (P_PV) changes to a value close to the target opening (MV). As a result, the predicted value is maintained at a value indicating a normal state, and an abnormality can be accurately estimated based on the difference between the predicted value and the actual value.

[0090] <Generation process of predicted value determination unit>

[0091] Next, an example of the generation process of the prediction value determination unit 102 of the present embodiment is described. As described above, the prediction value determination unit 102 uses a learned model. The learned model is constructed by using multiple training instruction values other than the latest training instruction value among multiple training instructions and multiple training evaluation values other than the latest training evaluation value among multiple training evaluation values as explanatory variables, performing machine learning with the latest training evaluation value as the correct answer label, and predicting a normal value corresponding to the next training evaluation value based on multiple variables corresponding to the above training instruction values and multiple variables corresponding to the above training evaluation values.

[0092] The predicted value determination unit 102 is equipped with the above-mentioned learned model and is also equipped with the following algorithm: the difference between the predicted value and the measured value is evaluated using a predetermined value, and the measured value is replaced with the predicted value based on the evaluation to perform prediction. Here, in this embodiment, the above-mentioned set value is set in consideration of the model accuracy of the learned model and the mechanical repeatability accuracy of the flow control valve 16.

[0093] Specifically, before setting the above-mentioned prescribed value, a test data set (target opening and actual opening) of a normal state is input to the learned model, and the error between the predicted value and the actual opening is evaluated to determine the MAE (mean absolute error) as the error of the opening as the model accuracy. In addition, the same target opening command is input to the normal flow control valve 16 multiple times, and the error between the actual opening corresponding to the command and the target opening is measured to determine the repetitive accuracy distribution of the opening. Then, the maximum value of the error of the upper 95% of the trials counting from the trial with the smaller error in the trials based on the multiple command inputs is determined. Then, the value obtained by adding the maximum value of the error of the upper 95% of the trials counting from the trial with the smaller error in the repetitive accuracy distribution is determined, and the above-mentioned prescribed value is set within the range of 0.9 to 1.1 times of the value. When the prescribed value is set based on the value considering the MAE and the repetitive accuracy in this way, the abnormality can be accurately inferred.

[0094] here, Figure 6 (A) shows an example of the distribution of model accuracy of the learned model. Figure 6 In the example (A), 107764 predictions were performed, and the target opening was input in the range of 0 to 100 (%). Figure 6 In the example of (A), the MAE is calculated to be 0.83%. Figure 6 (B) shows an example of the normal distribution of the repeatability of the opening of the flow control valve 16. Figure 6 In the example (B), 3000 trials were performed. And the maximum value of the error of the upper 95% of the trials in the repeatability distribution, starting from the trials with the smallest error, is 0.2%. In this case, the above-mentioned prescribed value for evaluating the difference between the predicted value and the measured value can be set to, for example, 1% based on 0.83+0.2.

[0095] The state detection device 100 of the present embodiment described above includes: a numerical value acquisition unit 101, which acquires the actual opening and target opening of the flow control valve 16 as specific numerical values that are continuously detected and / or determined; and a predicted value determination unit 102, which determines the predicted value of the next actual measured value of the plurality of measured values based on the plurality of actual openings and target openings acquired by the numerical value acquisition unit 101. Then, the state detection device 100 detects the state based on the actual opening acquired by the numerical value acquisition unit 101 and the predicted value corresponding to the actual opening determined by the predicted value determination unit 102. And, when the difference between the determined predicted value corresponding to the actual opening and the actual opening corresponding to the predicted value is greater than a predetermined value, the predicted value determination unit 102 replaces the actual opening corresponding to the predicted value with the predicted value and determines the next predicted value and subsequent predicted values. Furthermore, when the difference between the predicted value corresponding to the actual opening determined by the predicted value determination unit 102 and the actual opening corresponding to the predicted value is smaller than a prescribed value, the predicted value determination unit 102 does not replace the actual opening corresponding to the predicted value, but uses the actual opening to determine the next predicted value and subsequent predicted values.

[0096] Thus, in this embodiment, the predicted value for determining the state (abnormality) can be accurately determined, and the state can be accurately detected. That is, in this embodiment, when the predicted value deviates greatly from the measured value, the predicted value is used instead of the measured value when determining the next predicted value. Thus, the predicted value is maintained as a value indicating a normal state, and the abnormality based on the difference between the predicted value and the measured value can be accurately inferred.

[0097] Furthermore, the numerical value acquisition unit 101 acquires the target opening as the command value continuously input to the flow control valve 16 and the actual opening as the evaluation value of the action state of the flow control valve 16 continuously detected as specific numerical values used in the prediction. Then, the prediction value determination unit 102 determines the prediction value corresponding to the latest actual opening as the prediction value based on multiple target openings other than the latest target opening and multiple actual openings other than the latest actual opening. In this case, the target opening and the actual opening are used as two variables, so the accuracy of the prediction value can be improved. However, the prediction value determination unit 102 can also use only multiple actual openings, in other words, use one variable to predict the actual opening.

[0098] Furthermore, the present embodiment determines the predicted value for abnormality determination, and does not actively control the command value based on the predicted value. In this case, by adopting a structure in which the predicted value determination unit 102 determines the predicted value corresponding to the latest actual opening based on multiple target openings other than the latest target opening and multiple actual openings other than the latest actual opening as described above, the predicted value can be efficiently determined based on the acquisition of simple information. That is, the calculation speed of determining the predicted value can be increased, and the processing load can be reduced.

[0099] An embodiment of the present invention has been described above, but the present invention is not limited to the embodiment described above, and various further changes can be made to the above embodiment. For example, in the above embodiment, the state detection device 100 is applied to the state detection of the flow control valve 16 of the refrigeration device 1, but it can also be applied to the state detection of the expansion valve 13 and the compressor 11. In addition, the state detection device 100 can also be applied to the valve provided in the cooling device, and can also be used for other purposes.

Claims

1. A state detection device, comprising: a value acquisition section that acquires specific values that are continuously detected and / or determined; and a predicted value determination unit, which determines a predicted value of a next specific value of the multiple specific values according to the multiple specific values acquired by the value acquisition unit, The state detection device detects the state according to the specific value acquired by the value acquisition unit and the predicted value determined by the predicted value determination unit, in, The predicted value determination unit determines the next predicted value and subsequent predicted values by replacing the corresponding specific numerical value with the predicted value when the difference between the determined predicted value and the specific numerical value corresponding to the predicted value is greater than a predetermined value. When the difference between the determined predicted value and the specific numerical value corresponding to the predicted value is smaller than the prescribed value, the predicted value determination unit does not replace the corresponding specific numerical value with the predicted value, but uses the specific numerical value to determine the next predicted value and subsequent predicted values.

2. The state detection device according to claim 1, wherein: The state detection device also has: an error evaluation unit that determines a ratio of the number of times a difference between a predicted value determined by the predicted value determination unit and a specific numerical value corresponding to the predicted value is less than the specified value among a plurality of predictions performed by the predicted value determination unit within a specified time; as well as The abnormality notifying unit notifies an abnormality when the ratio determined by the error evaluating unit is smaller than a predetermined ratio.

3. The state detection device according to claim 2, wherein: The state detection device further includes a setting change unit that changes the specified value and the specified ratio. The setting change unit automatically adjusts the predetermined value and the predetermined ratio in an inverse proportional relationship.

4. The state detection device according to claim 1, wherein: The numerical value acquisition unit acquires, as the specific numerical value, a command value continuously input to the mechanical element in order to cause the mechanical element to operate and an evaluation value of the operating state of the mechanical element continuously detected. The predicted value determination unit determines, as the predicted value, a predicted value corresponding to the latest evaluation value based on a plurality of the command values other than the latest command value as the specific numerical values and a plurality of the evaluation values other than the latest evaluation value.

5. The state detection device according to claim 4, wherein: The predicted value determination unit replaces the latest evaluation value corresponding to the predicted value with the predicted value when the difference between the predicted value corresponding to the latest evaluation value and the latest evaluation value corresponding to the predicted value as a specific numerical value is greater than the specified value, and determines the next predicted value after the numerical value acquisition unit acquires the instruction value and the evaluation value that are next regarded as the latest instruction value and the latest evaluation value. When the difference between the prediction value corresponding to the latest evaluation value and the latest evaluation value as a specific numerical value corresponding to the prediction value is less than the prescribed value, the prediction value determination unit does not replace the latest evaluation value corresponding to it with the prediction value, but uses the evaluation value to determine the next prediction value after the numerical acquisition unit acquires the instruction value and the evaluation value that are next regarded as the latest instruction value and the latest evaluation value.

6. The state detection device according to claim 4 or 5, wherein: The predicted value determination unit makes a prediction using a learned model, wherein the learned model is a model that completes learning by taking as a normal state a training instruction value that is continuously input to the mechanical element in order to make the mechanical element in a normal state to act, and a training evaluation value corresponding to the action state of the mechanical element that acts according to the training instruction value, and the learned model predicts a normal value corresponding to the next training evaluation value based on a plurality of variables corresponding to the training instruction value and a plurality of variables corresponding to the training evaluation value. The predicted value determination unit determines predicted values corresponding to the evaluation values in the learned model using the plurality of command values and the plurality of evaluation values as variables, which are the specific numerical values.

7. A state detection method, comprising: A value acquisition step of acquiring specific values that are continuously detected and / or determined; and a predicted value determination step, determining a predicted value of a next specific value of the multiple specific values according to the multiple specific values obtained by the value acquisition step, detecting a state based on the specific value acquired by the value acquisition step and the predicted value determined by the predicted value determination step, in, In the prediction value determination step, when the difference between the determined prediction value and the specific numerical value corresponding to the prediction value is greater than a predetermined value, the prediction value is used to replace the specific numerical value corresponding to the prediction value to determine the next prediction value and subsequent prediction values. In the prediction value determination process, when the difference between the determined prediction value and the specific numerical value corresponding to the prediction value is less than the specified value, the prediction value is not used to replace the corresponding specific numerical value, and the specific numerical value is used to determine the next prediction value and subsequent prediction values.

8. The state detection method according to claim 7, wherein: The state detection method further comprises a learning model making step of making a learning model, wherein the learning model is learned by taking a training instruction value continuously input to the mechanical element in order to make the mechanical element in a normal state to act and a training evaluation value corresponding to the action state of the mechanical element that acts according to the training instruction value as a normal state, and predicting a normal value corresponding to the next training evaluation value based on a plurality of variables corresponding to the training instruction value and a plurality of variables corresponding to the training evaluation value, In the numerical value acquisition step, a command value continuously input to the mechanical element in order to operate the mechanical element and an evaluation value of the operating state of the mechanical element continuously detected are acquired as the specific numerical value. In the prediction value determination step, in the learned model, prediction values corresponding to the evaluation values are determined using the plurality of command values and the plurality of evaluation values as variables.

9. A state detection device for detecting the state of a valve whose opening degree is changed according to an input command signal, wherein: The state detection device comprises: a value acquisition unit, which acquires information on the actual opening of the valve as a specific value that is continuously detected and / or determined; and a predicted value determination unit, which determines a predicted value of the next actual opening degree of the plurality of actual opening degrees based on the information of the plurality of actual opening degrees acquired by the value acquisition unit, The state detection device detects the state based on the information of the actual opening of the valve acquired by the numerical value acquisition unit and the predicted value of the actual opening determined by the predicted value determination unit. When the difference between the determined predicted value and the actual opening corresponding to the predicted value is greater than a predetermined value, the predicted value determination unit replaces the actual opening corresponding to the predicted value with the predicted value to determine the next predicted value and subsequent predicted values. When the difference between a determined predicted value and an actual opening corresponding to the predicted value is smaller than the prescribed value, the predicted value determination unit does not replace the corresponding actual opening with the predicted value, but uses the actual opening to determine the next predicted value and subsequent predicted values.

10. A computer program for causing a computer to execute the following steps: a value acquisition step, acquiring specific values that are continuously detected and / or determined; and a predicted value determining step, determining a predicted value of a next specific value of the multiple specific values obtained in the value obtaining step, detecting a state according to the specific value acquired by the value acquisition step and the predicted value determined by the predicted value determination step, in, In the prediction value determination step, when the difference between the determined prediction value and the specific numerical value corresponding to the prediction value is greater than a predetermined value, the prediction value is used to replace the specific numerical value corresponding to the prediction value to determine the next prediction value and subsequent prediction values. In the prediction value determination step, when the difference between the determined prediction value and the specific numerical value corresponding to the prediction value is less than the specified value, the prediction value is not used to replace the corresponding specific numerical value, and the specific numerical value is used to determine the next prediction value and subsequent prediction values.