System control unit

Abstract

To provide a system controller capable of implementing highly precise correction control using an estimation model which can cope with an immense change in an environment of a system, which is an object of control, in the event that the immense change may arise.SOLUTION: Virtual charging coefficient variation value estimation means 41 creates plural candidates for an estimation model to be used to estimate variation values of operation of an actual system 20 derived from a change in a variation factor, and then, based on each estimation model, estimates a charging coefficient variation value. Evaluation means 42 evaluates each estimation model. An estimation model capable of preferably implementing operation of the actual system 20 is selected, and the operation of the actual system 20 is corrected or controlled based on the selected estimation model.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to a system control device for specifying a target value for operation and operating an actual system, and relates to a system control device suitable for controlling a filling device that drives a filling operation member such as an auger shaft to fill a fixed amount of a filling material such as powder into a filling target. [Background technology]

[0002] Patent Document 1 discloses a system control device previously proposed by the applicants of the present application for controlling a filling device. The system control device for the filling device disclosed in the document 1 creates an estimation model for calculating the correction rotation amount of the auger shaft (12) in order to achieve highly accurate fixed-quantity filling, and calculates a filling coefficient fluctuation estimation value used in the estimation model in accordance with the state value of the filling amount fluctuation factor. Specifically, the rotation amount of the auger shaft (12) required to fill a specified amount is calculated based on an arithmetic expression including a filling coefficient whose value varies depending on the filling amount variation factors. Meanwhile, the state measurement unit (24) measures state values ​​of the filling amount variation factors, and the filling coefficient variation value estimation means (25) calculates a variation value of the filling coefficient due to the filling amount variation factors (filling coefficient variation value) based on a preset estimation model in response to the state values. Then, the rotation amount of the auger shaft (12) is corrected based on the calculated estimated value of the filling coefficient variation. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2019-43631 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the conventional system control device disclosed in Patent Document 1 calculates the correction rotation amount of the auger shaft (12) using only a single pre-set estimation model. Therefore, if there is a significant change in the environment of the device to be controlled, the pre-set estimation model may not be able to respond to the change, and there is a limit to the accuracy of estimating the fluctuation of the filling coefficient in response to the change in the environment.

[0005] The present invention has been made in consideration of the above circumstances, and aims to provide a system control device that can perform highly accurate correction control using an estimation model that can respond to changes in the environment, etc., even if there are major changes in the environment, etc., of the device to be controlled. [Means for solving the problem]

[0006] In order to achieve the above object, the present invention provides a system control device for controlling an actual system (real system) equipped with a measurement unit that measures output results, and for operating the actual system by specifying a target value for operation, the system control device comprising: A fluctuation factor data acquisition means for acquiring data on factors (fluctuation factors) that cause fluctuations in operation at the target value by the actual system; a fluctuation value estimation means for estimating a fluctuation value of the operation of the actual system due to a change in a fluctuation factor based on an estimation model created in advance; a control means for calculating a correction value based on the fluctuation value estimated by the fluctuation value estimation means so that the actual system operates at the target value, and correcting and controlling the operation of the actual system based on the correction value; a virtual fluctuation value estimation means for estimating a fluctuation value of the operation of the actual system due to a change in a fluctuation factor based on a candidate estimation model created in advance; a virtual system that simulates the output results of the real system based on the fluctuation values ​​estimated by the virtual fluctuation value estimation means, the output of the control means, and the measurement values ​​measured by the measurement unit of the real system; evaluation means for evaluating the operation of the real system and the output results of the virtual system; and an estimation model switching means for selecting one of the estimation models including the candidates based on the evaluation result by the evaluation means, and using the selected estimation model for operation control of the actual system.

[0007] The present invention having the above-described configuration creates a plurality of estimation models for estimating fluctuation values ​​of the operation of the actual system due to changes in the variable factors, and selects and uses an estimation model based on the evaluation results by the evaluation means. Therefore, even if there is a major change in the environment of the device to be controlled, highly accurate correction control can be performed using an estimation model that can respond to the change in the environment.

[0008] Here, when a plurality of candidate estimation models are prepared, it is preferable to provide the same number of virtual systems and virtual fluctuation value estimating means as the number of candidate estimation models, corresponding to the plurality of candidate estimation models.

[0009] The present invention may also be configured to include an estimation model candidate creation means for creating estimation model candidates based on data obtained from the variation factor data acquisition means and data measured from the actual system. This configuration is efficient because it is not necessary to create multiple estimation models in advance, and the estimation model candidate creation means can prepare usable estimation models in response to changes in the environment, etc., as needed.

[0010] Furthermore, the estimation model candidate generating means may be configured to include a plurality of estimation model candidate generating means that generate estimation model candidates using different methods. By creating estimation models using different methods in this way, various estimation model candidates are created, and the range of estimation model choices is expanded, making it possible to achieve highly accurate correction control that is more adapted to changes in the environment, etc. [Effects of the Invention]

[0011] As described above, according to the present invention, even if there is a major change in the environment of the device to be controlled, highly accurate correction control can be performed using an estimation model that can respond to the change in the environment, etc. [Brief explanation of the drawings]

[0012] [Figure 1] 1 is a configuration diagram showing an overview of a filling device that is a control target of a system control device according to an embodiment of the present invention. [Figure 2] 1 is a block diagram showing a configuration of a system control device according to an embodiment of the present invention; [Figure 3] FIG. 10 is a block diagram showing the configuration of a system control device according to another embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0013] Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings. In this embodiment, an example configuration will be described in which the present invention is applied to a system control device that controls a filling device for driving a filling operation member such as an auger shaft 12 to fill a fixed amount of a filling material such as powder into a filling target.

[0014] [Schematic structure of the filling device to be controlled] FIG. 1 is a configuration diagram showing an outline of a filling device that is a control target of a system control device according to an embodiment of the present invention. The filling device fills a fixed amount of powder or granular material (powder, granules, or a mixture thereof) into a filling target such as a packaging container 1.

[0015] As shown in the figure, the filling device has a cylindrical filling tube 11 connected to the center of the bottom of a funnel-shaped hopper 10, and an auger shaft 12 coaxially arranged within the hollow portion of this filling tube 11. This auger shaft 12 has a rod-shaped structure with spiral blades on its circumferential surface. Powdered or granular material (filling material) is supplied to the hopper 10 from an opening on the top surface. The powdered or granular material stored in the hopper 10 is then supplied from above into the hollow portion of the filling tube 11, and as the auger shaft 12 rotates, the powdered or granular material is transferred downward within the hollow portion of the filling tube 11 and is discharged from a discharge port 11a that opens at the bottom end of the filling tube 11. The packaging container 1 to be filled is positioned below the discharge port 11a, and a fixed amount of the powdered or granular material discharged from the discharge port 11a is filled into the packaging container 1. The weight of the powder or granular material filled in the packaging container 1 is measured by a measuring instrument 16 (measuring unit).

[0016] As described above, the auger shaft 12 functions as a filling member for filling a fixed amount of material into a filling target. The auger shaft 12 is rotationally driven by a driving force from a servo motor 13 serving as a driving means.

[0017] A spatula-shaped filling assist member called an agitator 14 is provided inside the hopper 10. This agitator 14 is driven by a driving force from another servo motor 15 and moves in a circular motion along the inner circumferential surface of the hopper 10. The powder and granular material inside the hopper 10 is agitated by the circular movement of the agitator 14, and is smoothly guided into the filling tube 11 without accumulating inside the hopper 10.

[0018] [Control principle of system control device] Next, the principle of control of the filling device by the system control device according to this embodiment will be described in detail. In this embodiment, the control system is constructed on the premise that the filling amount Wp is defined as shown in the following equation (1) based on the density of the powder (filling material) supplied into the hopper 10, the structure of the auger shaft 12, the driving conditions of the auger shaft 12, and the filling efficiency of the filling material by the auger shaft 12.

[0019]

number

[0020] Here, the product ηρ of η and ρ is defined as the filling coefficient K. Furthermore, the elements on the right side of equation (1) are defined as θset and V, respectively, as follows:

[0021]

number

[0022]

number

[0023] Here, the state of the filling device, the state of the powder and granular material, and the surrounding environment (humidity and temperature) are factors (hereinafter referred to as filling amount fluctuation factors) that affect the filling amount Wp of powder and granular material (filling material) filled by the auger shaft 12 (filling operation member). When the state values ​​of these filling amount fluctuation factors fluctuate, the filling amount Wp fluctuates. In other words, the filling coefficient K fluctuates depending on the surrounding environment (temperature and humidity), the state of the powder and granular material, and the state of the filling device. Therefore, equation (1)' is transformed into equation (2) which includes the relationship between the variation of the filling coefficient K and the variation of the filling amount Wp.

[0024]

number

[0025]

number

[0026]

number

[0027] The estimated filling coefficient fluctuation value Δ^K can be calculated using a state equation, various statistical methods, and machine learning methods. For example, the relationship between the state values ​​of the various filling amount fluctuation factors described above and the filling coefficient fluctuation value ΔK can be derived using simple regression analysis or multiple regression analysis, and the estimated filling coefficient fluctuation value Δ^K can be obtained from that relational equation (estimation model).

[0028] When the simple regression analysis method is used, if the variable in the following relational expression (5) is a state value (for example, the torque value τagi of the agitator 14), an estimation model can be created by determining the constants a and b in advance through experiments, etc. Note that the state value is not limited to the torque value of the agitator 14.

[0029]

number

[0030] When the multiple regression analysis method is used, if the variable χi in the following relational expression (6) is a plurality of state values, an estimation model can be created by determining the constants a0 and ai (N is the number of related state values) in advance through experiments, etc.

[0031]

number

[0032] It is also possible to obtain the estimated filling coefficient fluctuation value Δ^K using machine learning with the RF5 method, which is a multivariate polynomial regression method using a neural network (three-layer perceptron). In this case, too, the variables are state values, and an estimation model can be created by determining constants through experiments, etc.

[0033] [Structure and operation of system control device] Next, the structure and operation of the system control device according to this embodiment will be described in detail. 2 is a block diagram showing the configuration of a system control device according to this embodiment. Specifically, the system control device is configured with the following components using a computer, a program, and devices such as measuring instruments.

[0034] The system control device controls the servo motor 13 to adjust the rotation amount of the auger shaft 12 in order to fill a fixed amount of powder or granular material, with the filling device being the control target as the actual system 20. The control means 30 for this purpose includes the components of output calculation means 31, drive command means 32, and correction means 33.

[0035] Based on the above-mentioned formula (4), the output calculation means 31 calculates the amount of rotation of the auger shaft 12 (i.e., the output of the auger shaft 12) required to discharge the filling amount (target value) of powder or granular material input (specified) by the input means 36 such as a keyboard or touch panel. Furthermore, it calculates the output of the servo motor 13 required to drive the auger shaft 12 by the calculated amount of rotation. The drive command means 32 generates a drive command signal corresponding to the output value of the servo motor 13 calculated by the output calculation means 31 and outputs it to the servo motor 13 . The servo motor 13 operates in accordance with a drive signal input from the drive command means 32, and drives the auger shaft 12 to rotate by the required amount. As a result, the specified amount of powder or granular material is discharged from the discharge port 11a of the filling cylinder 11 and filled into the packaging container 1 in a fixed amount.

[0036] As described above, the filling coefficient fluctuation value ΔK in equation (4) fluctuates with fluctuations in the state values ​​of the filling amount fluctuation factors. In response to this fluctuation in the filling coefficient fluctuation value ΔK, the rotation amount of the auger shaft 12 needs to be corrected by Δθ from the theoretical value θset. This correction rotation amount Δθ is calculated by the correction means 33 .

[0037] Furthermore, in order to estimate a filling coefficient fluctuation value ΔK required for calculating the correction rotation amount Δθ, a fluctuation factor data acquisition means 34 and a filling coefficient fluctuation value estimation means 35 (fluctuation value estimation means) are provided. The fluctuation factor data acquisition means 34 is configured to measure the current state in accordance with a pre-specified filling amount fluctuation factor, and output the measurement data (state value).

[0038] There are various factors that can cause fluctuations in the filling amount, but it is preferable to select factors that have a close correlation with the filling coefficient fluctuation value ΔK. Furthermore, the filling amount fluctuation factors to be measured can be measured not only directly but also indirectly. For example, changes in the state of factors that cause fluctuations in the filling amount, such as the viscosity and inertia of the powder or granular material, can be measured indirectly by using the fluctuation factor data acquisition means 34 to measure the torque and starting angle of the agitator 14, which receives pressure from the powder or granular material inside the hopper 10, or the torque and starting angle that the auger shaft 12 receives from the powder or granular material. 2 shows an example of a configuration in which the torque of the auger shaft 12, the torque of the agitator 14, the ambient temperature and humidity are used as filling amount fluctuation factors, and the measurement data is output as a status value. However, the filling amount fluctuation factors are not limited to these.

[0039] Based on an estimation model of the state values ​​of the filling amount fluctuation factors and the filling coefficient fluctuation value ΔK created in advance, the filling coefficient fluctuation value estimation means 35 calculates a filling coefficient fluctuation estimated value Δ^K corresponding to the current state of the filling amount fluctuation factors measured by the fluctuation factor data acquisition means 34. The estimation is performed based on the output data (state values) of the filling amount fluctuation factors measured by the fluctuation factor data acquisition means 34, and outputs the filling coefficient fluctuation estimated value Δ^K.

[0040] The correction means 33 calculates the correction rotation amount Δθ of the auger shaft 12 based on the above equation (4) in accordance with the filling coefficient fluctuation value ΔK estimated by the filling coefficient fluctuation value estimation means 35.

[0041] The rotation amount of the auger shaft 12 calculated by the output calculation means 31 is corrected by the correction rotation amount Δθ calculated by the correction means 33 . That is, in the calculation process of the output calculation means 31 described above, the calculation result is based on the theoretical values ​​Wn, Kn, and θset in the above equation (3), and errors due to changes in the state of the filling amount fluctuation factors are not taken into consideration. Therefore, based on the estimation model, an estimated value Δ^K of the filling coefficient fluctuation due to changes in the state of the filling amount fluctuation factors is calculated, and a corrective rotation amount Δθ is calculated using this estimated value and reflected in the calculation process by the output calculation means 31. In this way, the output calculation means 31 calculates the required rotation amount θ(t) = (θset + Δθ) of the auger shaft 12, including this corrective rotation amount Δθ. As a result, the filling device of this embodiment controls the drive of the auger shaft 12, taking into account errors (fluctuations) associated with changes in the state of factors that cause fluctuations in the filling amount, and fills a fixed amount of powder or granular material into the packaging container 1 as specified.

[0042] The system control device of this embodiment is further configured to create multiple estimation models for estimating fluctuation values ​​in the operation of the actual system 20 due to changes in the fluctuation factors, and by selecting and using these estimation models, even if there is a large change in the filling volume fluctuation factors of the filling device to be controlled, it is possible to perform highly accurate correction control using an estimation model that can respond to the change in the filling volume fluctuation factors.

[0043] To this end, the system control device of this embodiment includes components including a virtual filling device 40 (virtual system), a virtual filling coefficient fluctuation value estimation means 41 (virtual fluctuation value estimation means), an evaluation means 42, and an estimation model switching means 43. These components are also specifically constructed by a computer and a program.

[0044] The configuration shown in FIG. 2 is configured to include a plurality of units each consisting of a virtual filling coefficient fluctuation value estimating means 41 and a virtual filling device 40.

[0045] The virtual filling device 40 is a virtual system that simulates the operation of a filling device, which is the real system 20. The virtual filling device 40 simulates the output results of the real system 20 based on the fluctuation value estimated by the virtual filling coefficient fluctuation value estimation means 41, the output of the control means 30 for correctively controlling the real system 20, and the measurement value measured by the measuring device 16 of the real system 20.

[0046] Based on candidates for estimation models of the state values ​​of the filling amount fluctuation factors and the filling coefficient fluctuation value ΔK created in advance, the virtual filling coefficient fluctuation value estimation means 41 calculates a filling coefficient fluctuation estimated value Δ^K corresponding to the current state of the filling amount fluctuation factors measured by the fluctuation factor data acquisition means 34. The estimation is performed based on the output data of the filling amount fluctuation factors measured by the fluctuation factor data acquisition means 34, and outputs the filling coefficient fluctuation estimated value Δ^K.

[0047] Here, different estimation models are created and incorporated in advance for use in the calculations by the plurality of virtual filling coefficient fluctuation value estimation means 41. For example, as described above, an estimation model of relational expression (5) can be created using a simple regression analysis technique. Also, an estimation model of relational expression (6) can be created using a multiple regression analysis technique. An estimation model can also be created using a technique that utilizes machine learning based on the RF5 method, which is a multivariate polynomial regression method using a neural network (three-layer perceptron). While two virtual filling coefficient fluctuation value estimation means 41 are shown in FIG. 2, any number of virtual filling coefficient fluctuation value estimation means 41 (41-A, 41-B, ...) can be incorporated as needed.

[0048] Each virtual filling device 40 calculates a correction rotation amount Δθ based on the filling coefficient fluctuation estimated value Δ^K output from the virtual filling coefficient fluctuation value estimating means 41. (1) Then, the correction rotation amount Δθ (1) and the required rotation amount θ which is the output from the control means 30. (t) = (θset + Δθ) and the measurement value W measured by the measuring instrument 16 of the actual system 20 (t)Based on this, the virtual filling amount W^ is calculated using the following equations (7) and (8). (t) is the measurement value W measured by the measuring instrument 16 of the actual system 20 (t) and the required rotation amount θ (1) It is a value calculated from W^=K (t) ×(θset+Δθ (1) ) ···(7) K (t) =W (t) / θ (t) ···(8)

[0049] The evaluation means 42 evaluates the operation of the real system 20 and the operation of the virtual filling device 40. That is, the output result (measured value W (t) ) and the output result from the virtual filling machine 40 (virtual filling amount W^).

[0050] The estimation model switching means 43 selects one of the estimation models including the candidates based on the evaluation result by the evaluation means 42, and uses the selected estimation model for operation control of the filling device, which is the actual system 20. Here, when the estimation model that has been used to control the real system 20 (the estimation model that has been incorporated into the filling coefficient fluctuation value estimation means 35) is switched to a candidate estimation model that has been incorporated into the virtual filling coefficient fluctuation value estimation means 41, the candidate estimation model is incorporated into the filling coefficient fluctuation value estimation means 35 and used. Then, the estimation model that has been used to control the real system 20 until now is incorporated into the virtual filling coefficient fluctuation value estimation means 41 and is simulated by the virtual filling device 40.

[0051] For example, the target filling amount (target value) of powder or granular material is set to 100 g. It is also assumed that the filling coefficient fluctuation estimation means 35 used in the operational control of the actual system 20 calculates the filling coefficient fluctuation estimated value Δ^K based on the estimation model of the above relational expression (5) created by the simple regression analysis technique. It is also assumed that when correction control is performed using the output of this filling coefficient fluctuation estimation means 35 and the filling device of the actual system 20 is operated, the measurement result of the filling amount of powder or granular material is 98 g. On the other hand, it is assumed that one of the virtual filling coefficient fluctuation value estimation means 41 used for the operation control of the virtual filling device calculates the filling coefficient fluctuation estimated value Δ^K based on the estimation model of the above relational expression (6) created by the multiple regression analysis technique. When correction control is performed using the output of this virtual filling coefficient fluctuation value estimation means 41 and the operation of the virtual filling device is simulated, it is assumed that the measurement result of the filling amount of powder or granular material is 99 g. In this case, the evaluation means 42 switches from the filling coefficient fluctuation value estimation means 35 to the virtual filling coefficient fluctuation value estimation means 41 using the estimation model of the above relational equation (6) created by the multiple regression analysis technique, in which the filling amount measurement result is close to the target filling amount, and connects it to the control means 30 of the actual system 20.

[0052] By selecting and using an estimation model in this manner, the system control device of this embodiment can perform highly accurate correction control using an estimation model that can respond to changes in the filling volume fluctuation factors, even if there are significant changes in the filling volume fluctuation factors of the filling device to be controlled.

[0053] FIG. 3 is a block diagram showing the configuration of a system control device according to another embodiment of the present invention. In the system control device shown in FIG. 3, the same components as those in the system control device according to the previous embodiment shown in FIG. 2 are given the same reference numerals, and detailed description thereof will be omitted. The system control device shown in FIG. 3 includes a new model candidate generating unit 50 (estimation model candidate generating means) that generates candidates for estimation models.

[0054] The new model candidate generation unit 50 inputs the output (filling amount) from the weighing device 16 in the filling device of the actual system 20 and the output (state value of the filling amount fluctuation factors) from the fluctuation factor data acquisition means 34, and based on these output data, searches for a suitable estimation model that will bring the output from the weighing device 16 closer to the target value. When the system control device is equipped with a plurality of virtual filling coefficient variation value estimation means 41, a new model candidate generation unit 50 is provided corresponding to each virtual filling coefficient variation value estimation means 41, and the estimation model candidates generated by each new model candidate generation unit 50 are incorporated into the corresponding virtual filling coefficient variation value estimation means 41. Note that, although Fig. 3 shows two pairs of virtual filling coefficient variation value estimation means 41 and new model candidate generation unit 50, any number of pairs of these components can be incorporated as necessary.

[0055] By configuring in this manner, even if the factors that cause fluctuations in the filling volume of the filling device to be controlled change significantly, it is possible to continuously perform highly accurate correction control by appropriately creating an estimation model that can respond to those changes.

[0056] The estimation model candidate generating means 70 can be constructed in a virtual space of a computer using, for example, artificial intelligence (AI) software.

[0057] The present invention is not limited to the above-described embodiment, and it goes without saying that various modifications and applications are possible within the scope of the gist of the invention as defined in the claims. For example, in the above-described embodiment, the present invention is applied to a filling device configured to fill a fixed amount of powder or granular material (powder, granules, or a mixture thereof) into a filling target such as a packaging container 1, but the present invention is not limited to this and can be applied to the operation control of an actual system 20 used in various industrial and technical fields. [Explanation of symbols]

[0058] 10: Hopper, 11: Filling tube, 11a: Discharge port, 12: Auger shaft, 13, 15: Servo motor, 14: Agitator, 16: Measuring instrument (measuring unit), 20: Actual system, 30: Control means, 31: Output calculation means, 32: Drive command means, 33: Correction means, 34: Fluctuation factor data acquisition means, 35: Filling coefficient fluctuation value estimation means, 36: Input means, 40: Virtual filling device, 41 (41-A, 41-B): Virtual filling coefficient fluctuation value estimation means, 42: Evaluation means, 43: Estimation model switching means, 50: New model candidate generation unit (estimation model candidate creation means)

Claims

1. A system control device for operating an actual system (real system) by specifying a target value for operation, the system control device comprising: a fluctuation factor data acquisition means for acquiring data on factors (fluctuation factors) that cause fluctuations in the operation of the actual system at the target value; a fluctuation value estimation means for estimating a fluctuation value of the operation of the actual system due to a change in the fluctuation factor based on an estimation model created in advance; a control means for calculating a correction value based on the fluctuation value estimated by the fluctuation value estimation means so that the actual system operates at the target value, and correcting and controlling the operation of the actual system based on the correction value; a virtual fluctuation value estimation means for estimating a fluctuation value of the operation of the actual system due to a change in the fluctuation factor based on a candidate estimation model created in advance; a virtual system that simulates an output result of the real system based on the fluctuation value estimated by the virtual fluctuation value estimation means, the output of the control means, and a measurement value measured by a measurement unit of the real system; evaluation means for evaluating the operation of the real system and the output results of the virtual system; an estimation model switching means for selecting one of the estimation models including the candidate models based on the evaluation result by the evaluation means, and using the selected estimation model for operation control of the actual system; an estimation model candidate creation means for creating a virtual estimation model candidate from the data acquired by the variation factor data acquisition means and the data measured by the measurement unit; A system control device comprising:

2. The actual system is an auger screw type filling device, 2. The system control device according to claim 1, wherein the measuring unit is a weighing device that measures the weight of the filled article. The system control device described in claim 1, wherein the estimation model candidate creation means inputs the filling amount output from the weighing device in the auger screw type filling device of the actual system and the state value of the filling amount fluctuation factor output from the fluctuation factor data acquisition means, and based on these output data, searches for an estimation model that brings the output from the weighing device closer to the target value.

3. A plurality of the virtual fluctuation value estimating means are provided, an estimation model candidate creating means for creating estimation model candidates by different methods corresponding to the plurality of virtual fluctuation value estimating means; 3. The system control device according to claim 1, wherein the estimation model candidates generated by the estimation model candidate generating means are incorporated into the virtual variation value estimating means associated with the estimation model candidates.

Images