Computation device, pressure measurement device, computation method, and computation program

Abstract

The present invention comprises: a reception unit 10 that receives an input value from a sensor S; a data group storage unit 40 that stores a data group comprising a plurality of input values acquired in advance; a statistical distance calculation unit 50 that calculates the statistical distance, with respect to the data group, of the input value received by the reception unit 10; a smoothing processing unit 20 that performs smoothing processing on the input value received by the reception unit 10; a determination unit 70 that determines whether or not the statistical distance satisfies a predetermined condition; and an expression change unit 80 that, if the statistical distance does not satisfy the predetermined condition, changes an arithmetic expression used for the smoothing processing from a first arithmetic expression used when the statistical distance satisfies the predetermined condition to a second arithmetic expression having higher followability.

Description

Calculation device, pressure measurement device, calculation method, and calculation program

[0001] The present invention relates to a computing device, a computing method, and a computing program.

[0002] For example, as shown in Patent Document 1, when monitoring values ​​detected by various sensors, such as installing a pressure sensor in a semiconductor process chamber to monitor the internal pressure, the raw data output from the sensor contains noise, so the raw data may be subjected to smoothing processing such as applying a moving average before being displayed.

[0003] The arithmetic formula used in such smoothing processing is required to have a stable displayed output value when the input value, which is the raw data, is stable. From this perspective, it is desirable to use an arithmetic formula that can guarantee the stability of the output value, such as one with a large number of moving averages.

[0004] On the other hand, if an arithmetic formula with a large number of moving averages is used as described above, when the input value fluctuates, for example due to a change in settings during the process, the displayed output value will not quickly follow the input value, and the required tracking ability cannot be guaranteed.

[0005] In this way, for the arithmetic formula used in the smoothing process, there is a trade-off between stability and tracking ability of the output value.

[0006] Japanese Patent Application Laid-Open No. 2023-54358

[0007] In this situation, the inventors of the present application have considered a method for improving both stability and tracking capability by monitoring the amount of fluctuation in the input value, and when the amount of fluctuation is below a threshold, performing smoothing processing using a highly stable arithmetic formula, and when the amount of fluctuation exceeds the threshold, switching to a arithmetic formula with high tracking capability.

[0008] However, when the input value fluctuates slowly or only slightly, such as the internal pressure of a semiconductor process chamber, it is difficult to set a threshold value that can capture the fluctuation, and the above-mentioned calculation formula cannot be switched appropriately.

[0009] Therefore, the present invention has been made to solve the above problems, and its main objective is to improve both the stability and tracking ability of the output value obtained by smoothing the input value, even when the input value fluctuates slowly or only slightly.

[0010] That is, the arithmetic device according to the present invention is characterized by having a receiving unit that receives input values ​​from a sensor, a data group storage unit that stores a data group consisting of a plurality of input values ​​acquired in advance, a statistical distance calculation unit that calculates the statistical distance of the input value received by the receiving unit from the data group, a smoothing processing unit that smoothes the input value received by the receiving unit, a determination unit that determines whether or not the statistical distance satisfies a predetermined condition, and a formula modification unit that, if the statistical distance does not satisfy the predetermined condition, changes the arithmetic formula used in the smoothing processing from a first arithmetic formula that is used when the statistical distance satisfies the predetermined condition to a second arithmetic formula that has better tracking ability.

[0011] According to the arithmetic device configured in this manner, the receiving unit calculates the statistical distance of the input value received and increases the feature amount for fluctuations in the input value, thereby making it possible to capture gradual or slight fluctuations in the input value. Furthermore, if the statistical distance does not satisfy a predetermined condition, the arithmetic formula used for the smoothing process is changed from the first arithmetic formula used when the statistical distance satisfies the predetermined condition to the second arithmetic formula with higher tracking capability. Therefore, even if the fluctuations in the input value are gradual or slight, the fluctuations can be captured while improving both the stability and tracking capability of the output value obtained by smoothing the input value.

[0012] However, if the sensor noise is large and the statistical distance calculated by the statistical distance calculation unit fluctuates significantly with each calculation, when the input value fluctuates and the second arithmetic formula should be switched to, the statistical distance may suddenly decrease in value, satisfying a predetermined condition, making it impossible to appropriately switch the arithmetic formula. Therefore, it is preferable that the apparatus further includes a second smoothing processing unit that smoothes the statistical distance calculated by the statistical distance calculation unit, and the determination unit determines whether the statistical distance smoothed by the second smoothing processing unit satisfies the predetermined condition. A statistical distance smoothed in this manner will not suddenly decrease in value when the input value fluctuates, making it possible to appropriately change the arithmetic formula.

[0013] Here, if, for example, the set pressure of the chamber in which the sensor is installed differs between when the value detected by the sensor is monitored and when the input values ​​constituting the data group are acquired, there is a concern that most of the input values ​​accepted by the accepting unit will deviate from the data group, resulting in an overestimated statistical distance. Therefore, it is preferable that the statistical distance calculation unit calculates the statistical distance of the corrected input value for the data group by performing a predetermined correction operation on the input value accepted by the accepting unit. In this case, by appropriately setting the correction value, it is possible to prevent the statistical distance from being overestimated.

[0014] When the statistical distance does not satisfy the predetermined condition, it is preferable that the formula modification unit modify the first arithmetic formula to the second arithmetic formula by correcting a coefficient included in the first arithmetic formula. With this configuration, the arithmetic process can be simplified compared to when separate arithmetic formulas that are unrelated to each other are used as the first arithmetic formula and the second arithmetic formula.

[0015] When the statistical distance does not satisfy the predetermined condition, the formula modification unit preferably modifies the first arithmetic formula to the second arithmetic formula by correcting the coefficient included in the first arithmetic formula to a different value according to the statistical distance. With this configuration, the coefficient is increased as the statistical distance increases, thereby further improving tracking performance.

[0016] Preferably, the smoothing unit performs Kalman filter processing on the input value, and the coefficient is a weighting coefficient multiplied by the difference between the input value input to the smoothing unit and the output value output from the smoothing unit before the input. With this configuration, by correcting the weighting coefficient, it is possible to improve tracking performance without changing, for example, a Kalman gain used in the Kalman filter processing.

[0017] A specific example of the statistical distance calculation unit is one that calculates, as the statistical distance, a Mahalanobis distance for the data group of the input value received by the reception unit.

[0018] However, capacitance-type pressure sensors measure the displacement of a diaphragm, which deforms in response to changes in pressure, as a change in capacitance and convert it into pressure. This makes them susceptible to noise caused by diaphragm vibration, making the smoothing process described above even more difficult. Therefore, the pressure measuring device according to the present invention is characterized by comprising the above-described arithmetic unit and a pressure sensor that outputs pressure as the input value. With this pressure measuring device configured in this manner, both the stability and tracking of the output value obtained by smoothing the input value from the pressure sensor can be improved, making the pressure measuring device according to the present invention useful for monitoring the internal pressure of, for example, a semiconductor process chamber.

[0019] A calculation method according to the present invention is a method characterized by comprising: a receiving step of receiving input values ​​from a sensor; a data group storage step of storing a data group consisting of a plurality of input values ​​acquired in advance; a statistical distance calculation step of calculating a statistical distance for the data group of the input values ​​received in the receiving step; a smoothing processing step of smoothing the input values ​​received in the receiving step; a determination step of determining whether or not the statistical distance satisfies a predetermined condition; and an equation changing step of changing, if the statistical distance does not satisfy the predetermined condition, the calculation formula used in the smoothing processing from a first calculation formula used when the statistical distance satisfies the predetermined condition to a second calculation formula with better tracking ability.

[0020] The arithmetic program of the present invention is characterized in that it causes a computer to function as a reception unit that receives input values ​​from a sensor, a data group storage unit that stores a data group consisting of a plurality of input values ​​acquired in advance, a statistical distance calculation unit that calculates the statistical distance of the input value received by the reception unit from the data group, a smoothing processing unit that smoothes the input value received by the reception unit, a determination unit that determines whether the statistical distance satisfies a predetermined condition, and an equation modification unit that, if the statistical distance does not satisfy the predetermined condition, changes the arithmetic formula used in the smoothing processing from a first arithmetic formula that is used when the statistical distance satisfies the predetermined condition to a second arithmetic formula with better tracking ability.

[0021] Such a calculation method and calculation program can achieve the same effects as those of the above-mentioned calculation device.

[0022] According to the present invention configured in this manner, even when the input value fluctuates slowly or only slightly, it is possible to improve both the stability and tracking ability of the output value obtained by smoothing the input value.

[0023] FIG. 1 is a schematic diagram showing an example of use of a calculation device according to an embodiment of the present invention; FIG. 2 is a functional block diagram showing the functions of the calculation device according to the embodiment; FIG. 3 is a flowchart showing the operation of the calculation device according to the embodiment; FIG. 4 is a mathematical formula showing a calculation formula used in a statistical distance calculation unit according to the embodiment; FIG. 5 is a schematic diagram explaining the function of a second smoothing processing unit according to the embodiment; FIG. 6 is a mathematical formula showing an arithmetic formula used in a smoothing processing unit according to the embodiment;

[0024] An embodiment of a computing device according to the present invention will be described below with reference to the drawings.

[0025] 1, the arithmetic device 100 of this embodiment is used in, for example, a semiconductor manufacturing process, and is used to monitor values ​​such as pressure detected by, for example, a capacitance-type pressure sensor S provided in a semiconductor process chamber CH. The arithmetic device 100 and the pressure sensor S constitute a pressure measurement device that detects and outputs the pressure of the chamber CH, etc.

[0026] However, the use of the arithmetic device 100 is not limited to semiconductor manufacturing processes, and the detected value may be pressure output from a pressure sensor S of a type other than the capacitance type, or may be various physical quantities output from various sensors S, such as a temperature sensor, humidity sensor, distance sensor, or speed sensor, without being limited to pressure.

[0027] This arithmetic device 100 smoothes the input values ​​from the sensor S (in other words, the output values ​​output from the sensor S, i.e., the raw data output from the sensor S) and displays and outputs them, and is specifically a computer equipped with a CPU, memory, an input / output interface, etc.

[0028] More specifically, the arithmetic device 100 of this embodiment is configured such that, by the CPU and its peripheral devices working together in accordance with the arithmetic program stored in the memory, it is provided with a reception unit 10 that receives input values ​​from a sensor S, a smoothing processing unit 20 that smoothes the input values ​​received by the reception unit 10, and an output unit 30 that displays the input values ​​after smoothing processing by the smoothing processing unit 20 as output values ​​on a display or the like, as shown in FIG.

[0029] The arithmetic device 100 is configured to use at least two arithmetic expressions (hereinafter referred to as a first arithmetic expression and a second arithmetic expression) as arithmetic expressions used in the smoothing process.

[0030] Here, when focusing on the stability and tracking ability of the arithmetic formula, the first arithmetic formula is more stable than the second arithmetic formula, but conversely has less tracking ability than the second arithmetic formula; in other words, the second arithmetic formula is more tracking ability than the first arithmetic formula, but conversely has less stability than the first arithmetic formula.

[0031] The stability of the arithmetic formula is an index showing the stability of the output value when the input value is stable, and the tracking ability of the arithmetic formula is an index showing the response speed of the output value when the input value is fluctuating. Note that when the input value is stable, for example, there are no changes in various settings and the amount or rate of fluctuation of the input value is below a predetermined value, and conversely, when the input value is fluctuating, for example, there is a change in at least one of the various settings and the amount or rate of fluctuation of the input value exceeds a predetermined value.

[0032] The arithmetic device 100 of this embodiment further includes functions as a data group storage unit 40, a statistical distance calculation unit 50, a second smoothing processing unit 60, a determination unit 70, and an equation modification unit 80, as shown in FIG. 2, by the CPU and its peripheral devices working together in accordance with the arithmetic program stored in the memory.

[0033] In the following, the operation of the arithmetic device 100 will be described with reference to the flowchart of FIG. 3, along with an explanation of these functions.

[0034] The data group storage unit 40 is set in a predetermined area of ​​the memory, and acquires in advance a large number of input values ​​output from the sensor S, and stores a data group consisting of these large number of input values ​​(S1).

[0035] This data group is made up of input values ​​when stable, and is acquired, for example, when the arithmetic device 100 is shipped from a factory or when it is calibrated at the site of use. The data group here is made up of a large number of input values ​​(in other words, a large number of output values ​​from the sensor S) acquired when the pressure inside the chamber CH is being maintained at a predetermined value, such as 0 Pa.

[0036] In this embodiment, the receiving unit 10 sequentially receives input values ​​from the sensor S (S2).

[0037] The statistical distance calculation unit 50 calculates the statistical distance of the input value for the data group received by the receiving unit 10. The statistical distance indicates the degree to which the input value deviates from the data group based on the average value and variance (degree of variation) of the data group described above.

[0038] The statistical distance calculation unit 50 of this embodiment sequentially calculates the Mahalanobis distance, which represents a measure between the input value received by the receiving unit 10 and the distribution of the input values ​​that make up the above-mentioned data group, as the statistical distance (S3).

[0039] The statistical distance calculation unit 50 is configured to calculate the statistical distance using a predetermined statistical distance calculation formula. Examples of such a statistical distance calculation formula include a formula that includes the average value and variance of the data group described above as parameters. This statistical distance calculation formula is pre-stored in a calculation formula storage unit (not shown) that is set in a predetermined area of ​​the memory. However, the specific calculation method of the statistical distance calculation unit 50 is not limited to the above, and may be a calculation method that uses, for example, differential values ​​and / or integral values ​​of input values ​​at regular intervals.

[0040] Incidentally, the above-mentioned data group was acquired when the pressure in the chamber CH was being maintained at 0 Pa. If the pressure in the chamber CH is to be maintained at a pressure other than 0 Pa, most of the input values ​​at that time will deviate from the data group, and the Mahalanobis distance will continue to be calculated as an excessive value.

[0041] Therefore, the statistical distance calculation unit 50 may calculate the statistical distance for the data group of the input value itself, but in order to prevent the Mahalanobis distance from being calculated as an excessive value, in this embodiment, the input value is corrected and the statistical distance for the data group of the corrected input value is calculated.

[0042] In other words, the statistical distance calculated by the statistical distance calculation unit 50 is a concept that includes not only the statistical distance for the data group of the input values ​​themselves, but also the statistical distance for the data group of the corrected input values ​​after arithmetic operations with a predetermined correction value.

[0043] Specifically, as shown in the formula in FIG. 4, a correction is made by subtracting the output value x output by the statistical distance calculation unit 50 prior to the input of the input value y to the statistical distance calculation unit 50, and the Mahalanobis distance for the data group of the corrected input value Δ is calculated.

[0044] More specifically, the statistical distance calculation unit 50 of this embodiment calculates a value obtained by subtracting the output value x from the smoothing processing unit 20 at time k-1 from the input value y at time k, and calculates the Mahalanobis distance for the data group of this corrected input value Δ.

[0045] Next, in this embodiment, the second smoothing processing unit 60 smoothes the statistical distance calculated by the statistical distance calculation unit 50 (S4).

[0046] This second smoothing processing unit 60 smoothes the statistical distances output from the statistical distance calculation unit 50 and outputs the smoothed statistical distances sequentially to the determination unit 70 described later. In this example, as shown in FIG. 5 , the second smoothing processing unit 60 performs Kalman filter processing on the statistical distances and outputs the logarithmic statistical distances obtained after the processing to the determination unit 70.

[0047] However, the specific form of the smoothing process is not necessarily limited to Karma filter processing, and various forms such as moving averages may be used, and it is not necessarily necessary to take the logarithm of the statistical distance after processing; the smoothed statistical distance may be output directly to the judgment unit 70.

[0048] Next, the determining unit 70 determines whether the statistical distance output by the second smoothing processing unit 60 satisfies a predetermined condition (S5).

[0049] The determination unit 70 of this embodiment determines whether the statistical distance output by the second smoothing processing unit 60 exceeds a predetermined threshold value, and determines that the predetermined condition is satisfied if the statistical distance is below the threshold value, and determines that the predetermined condition is not satisfied if the statistical distance exceeds the threshold value. Note that this threshold value is stored in advance in a threshold value storage unit (not shown) set in a predetermined area of ​​the memory.

[0050] Specifically, the judgment unit 70 sequentially compares the statistical distance output from the second smoothing processing unit 60 with a threshold value, and sequentially outputs a signal indicating that the statistical distance is below the threshold value or a signal indicating that the statistical distance exceeds the threshold value to the equation modification unit 80 as the judgment result.

[0051] The formula change unit 80 changes the arithmetic formula used by the smoothing processing unit 20 to the first arithmetic formula or the second arithmetic formula based on the determination result of the determination unit 70 .

[0052] Specifically, the formula modification unit 80 receives a signal indicating the judgment result from the judgment unit 70, and if the statistical distance output by the second smoothing processing unit 60 is below a threshold value (i.e., if the statistical distance satisfies a predetermined condition), selects the first arithmetic formula as the arithmetic formula to be used by the smoothing processing unit 20 (S6).

[0053] On the other hand, the formula change unit 80 receives a signal indicating the judgment result from the judgment unit 70, and when the statistical distance output by the second smoothing processing unit 60 exceeds a threshold value (i.e., when the statistical distance does not satisfy a predetermined condition), it changes the arithmetic formula used in the smoothing processing unit 20 from the first arithmetic formula to a second arithmetic formula with higher tracking ability (S7).

[0054] To explain in more detail, when the statistical distance exceeds a threshold value, the formula modification unit 80 corrects the coefficients contained in the first formula, thereby changing the formula used in the smoothing processing unit 20 from the first formula to the second formula.

[0055] Here, the smoothing processing unit 20 of this embodiment is configured to perform Kalman filter processing on the input value received by the receiving unit 10 .

[0056] This Kalman filter processing is a smoothing process that outputs an estimate of an input value at time k, in other words, a smoothing process that outputs an estimate of an output value at time k output from the sensor S described above.

[0057] Specifically, the smoothing processing unit 20 outputs an estimated value of the input value at time k using a prior estimated value at time k estimated using data up to time k-1 and the input value at time k.

[0058] More specifically, in the process of calculating an estimate of the input value at time k (i.e., in the process of performing Kalman processing on the input value at time k), the smoothing processing unit 20 calculates the above-mentioned prior estimate at time k using the calculation formula shown in FIG. 6 , and outputs the estimate of the input value at time k using this prior estimate, the Kalman gain, etc.

[0059] This prior estimate x ｍ As can be seen from the calculation formula of FIG. 6, the value x is obtained by adding the difference u between the current input value input to the smoothing processing unit 20 and the previous output value output from the smoothing processing unit 20 to the previous output value x (i.e., the previous output value x obtained by the Kalman filter processing of the smoothing processing unit 20). ｍ is.

[0060] The difference u is not limited to the difference between the current input value and the previous output value, but may be, for example, the difference between the current input value and the output before last or even earlier, or may be the difference between the input value input to the smoothing processing unit 20 and the output value output from the smoothing processing unit 20 prior to that input.

[0061] The calculation formula of this embodiment includes a weighting coefficient A that weights the previous output value x and a weighting coefficient B that weights the difference u between the current input value and the previous output value, and the output value is the sum of the value obtained by multiplying the previous output value x by the weighting coefficient A and the value obtained by multiplying the difference u between the current input value and the previous output value by the weighting coefficient B. The weighting coefficients A and B are values ​​between 0 and 1.

[0062] In such a configuration, the formula modification unit 80 modifies the arithmetic formula used in the smoothing processing unit 20 to either the first arithmetic formula or the second arithmetic formula by correcting the weighting coefficient B that is multiplied by the difference u between the input value input to the smoothing processing unit 20 and the output value output from the smoothing processing unit 20 prior to that input.

[0063] Specifically, the formula modification unit 80 is configured to modify the first arithmetic formula to the second arithmetic formula by correcting the weighting coefficient B included in the first arithmetic formula to another value according to the statistical distance when the statistical distance exceeds a threshold value (i.e., when the statistical distance does not satisfy a predetermined condition).

[0064] In other words, when the statistical distance is below the threshold, an arithmetic expression including a weighting coefficient B as a predetermined fixed value is used as the first arithmetic expression, and when the statistical distance exceeds the threshold, an arithmetic expression in which the weighting coefficient B is corrected to a value greater than the fixed value as the statistical distance increases is used as the second arithmetic expression. Note that, as described above, the weighting coefficient B is a value between 0 and 1.

[0065] Then, the smoothing processing unit 20 smoothes the input value using the first arithmetic formula selected by the formula modification unit 80 or the second arithmetic formula switched by the formula modification unit 80 (S8), and the output unit 30 displays the smoothed value as an output value on a display or the like (S9).

[0066] The effects of the arithmetic device 100 of this embodiment are shown in the graphs of Figures 7 and 8. According to the arithmetic device 100 of this embodiment, as compared to smoothing processing using, for example, 32 moving averages, the ability of the output value to follow fluctuations in the input value is improved, as can be seen from Figure 7, and the stability of the output value when the input value is stable is also improved, as can be seen from Figure 8.

[0067] According to the arithmetic device 100 configured in this manner, the receiving unit 10 calculates the statistical distance of the input value received and increases the feature amount for fluctuations in the input value, thereby making it possible to capture gradual or slight fluctuations in the input value. Furthermore, when the statistical distance exceeds a threshold (i.e., when the statistical distance does not satisfy a predetermined condition), the arithmetic formula used for the smoothing process is changed from the first arithmetic formula used when the statistical distance is below the threshold (i.e., when the statistical distance satisfies the predetermined condition) to the second arithmetic formula with higher tracking capability. Therefore, even when the fluctuations in the input value are gradual or slight, the fluctuations can be captured while improving both the stability and tracking capability of the output value obtained by smoothing the input value.

[0068] Furthermore, when the statistical distance exceeds a threshold value, the formula modification unit 80 corrects the coefficients contained in the formula to change from the first formula to the second formula, thereby simplifying the calculation process compared to when these formulas are unrelated to each other.

[0069] Furthermore, when the statistical distance exceeds a threshold value, the equation modification unit 80 increases the weighting coefficient B included in the above-mentioned arithmetic equation as the statistical distance increases, thereby further improving tracking performance.

[0070] In addition, since the equation modification unit 80 is configured to correct the weighting coefficient that is multiplied by the difference between the input value input to the smoothing processing unit 20 and the output value output from the smoothing processing unit 20 prior to that input, it is possible to improve tracking performance without changing the calculation equation (e.g., Kalman gain, etc.) used in the Kalman filter processing.

[0071] As shown in the graph on the left of Fig. 5 , the statistical distance calculated by the statistical distance calculation unit 50 fluctuates, and when the input value fluctuates and the calculation formula should be changed to the second calculation formula, there is a risk that the statistical distance will suddenly decrease below the threshold (the area surrounded by the dotted line in the graph on the left of Fig. 5 ), making it impossible to appropriately change the calculation formula. In contrast, the calculation device 100 of this embodiment has a second smoothing processing unit 60 that smoothes the statistical distance calculated by the statistical distance calculation unit 50, so that the smoothed statistical distance will not suddenly decrease when the input value fluctuates, making it possible to appropriately change the calculation formula.

[0072] Furthermore, since the arithmetic unit 100 of this embodiment smooths the pressure output from the pressure sensor S and outputs the smoothed pressure, it is useful for monitoring the internal pressure of the semiconductor process chamber CH.

[0073] The present invention is not limited to the above-described embodiment.

[0074] For example, in the above embodiment, the equation modification unit 80 modifies the first arithmetic equation to the second arithmetic equation by correcting the coefficients included in the arithmetic equation, but it may also modify the first arithmetic equation to the second arithmetic equation by modifying the Kalman gain, or it may modify the first arithmetic equation to a completely different second arithmetic equation (having no common terms).

[0075] Furthermore, the correction of the coefficients included in the arithmetic formula is not limited to the above embodiment, and the formula modification unit 80 may be configured to correct the coefficients included in the first arithmetic formula to another fixed value when the statistical distance exceeds a threshold. A specific example of this configuration is to make the fixed value of the weighting coefficient B included in the second arithmetic formula larger than the fixed value of the weighting coefficient B included in the first arithmetic formula. With this configuration, the load of the arithmetic processing can be reduced compared to the above embodiment in which the coefficients are corrected according to the statistical distance.

[0076] In the above embodiment, the smoothing processing unit 20 performs Kalman processing on the input value, but may also perform smoothing by taking a moving average.

[0077] The calculation device 100 may not have the function of the second smoothing processing unit 60 and may output the statistical distance calculated by the statistical distance calculation unit 50 to the determination unit 70 as is.

[0078] Furthermore, in the above embodiment, a determination as to whether the statistical distance satisfies a predetermined condition is made by determining whether the statistical distance is below a threshold value. However, it is also acceptable to use a determination as to, for example, whether the amount of fluctuation or rate of fluctuation of the statistical distance is below a threshold value.

[0079] Furthermore, although the statistical distance has been described using Mahalanobis distance as an example in the above embodiment, it may also be an index obtained by machine learning such as a support vector machine (SVM), or an index such as AIC (Akaike Information Criterion) or BIC (Bayes Information Criterion).

[0080] Furthermore, the present invention is not limited to the above-described embodiment, and it goes without saying that various modifications are possible without departing from the spirit of the present invention.

[0081] According to the present invention, even when the input value fluctuates slowly or only slightly, it is possible to improve both the stability and tracking ability of the output value obtained by smoothing the input value.

[0082] REFERENCE SIGNS LIST 100: Calculation device CH: Chamber S: Sensor 10: Reception unit 20: Smoothing processing unit 30: Output unit 40: Data group storage unit 50: Statistical distance calculation unit 60: Second smoothing processing unit 70: Determination unit 80: Formula modification unit

Claims

1. a reception unit that receives an input value from a sensor; a data group storage unit for storing a data group consisting of a plurality of input values ​​acquired in advance; a statistical distance calculation unit that calculates a statistical distance between the input value received by the reception unit and the data group; a smoothing processing unit that smoothes the input value received by the receiving unit; a determination unit that determines whether the statistical distance satisfies a predetermined condition; and a formula change unit that changes the arithmetic formula used in the smoothing process from a first arithmetic formula used when the statistical distance satisfies the predetermined condition to a second arithmetic formula with higher tracking ability when the statistical distance does not satisfy the predetermined condition.

2. a second smoothing unit that smoothes the statistical distance calculated by the statistical distance calculation unit; The arithmetic device according to claim 1 , wherein the determining unit determines whether the statistical distance smoothed by the second smoothing unit satisfies the predetermined condition.

3. 3. The arithmetic device according to claim 1, wherein the statistical distance calculation unit calculates the statistical distance of a corrected input value for the data group, the corrected input value being calculated by performing a predetermined correction operation on the input value received by the reception unit.

4. 3. The arithmetic device according to claim 1, wherein when the statistical distance does not satisfy the predetermined condition, the formula modification unit changes the first arithmetic formula to the second arithmetic formula by correcting a coefficient included in the first arithmetic formula.

5. 5. The arithmetic device according to claim 4, wherein, when the statistical distance does not satisfy the predetermined condition, the formula modification unit modifies the first arithmetic formula to the second arithmetic formula by correcting the coefficient included in the first arithmetic formula to another value corresponding to the statistical distance.

6. the smoothing processing unit performs Kalman filter processing on the input value, 5. The arithmetic device according to claim 4, wherein the coefficient is a weighting coefficient by which the difference between the input value input to the smoothing processor and the output value output from the smoothing processor prior to the input is multiplied.

7. The arithmetic device according to claim 1 , wherein the statistical distance calculation unit calculates, as the statistical distance, a Mahalanobis distance of the input value received by the reception unit to the data group.

8. A computing device according to claim 1 or 2; a pressure sensor that outputs pressure as the input value.

9. a receiving step of receiving an input value from a sensor; a data group storage step of storing a data group consisting of a plurality of input values ​​acquired in advance; a statistical distance calculation step of calculating a statistical distance of the input value received in the receiving step with respect to the data group; a smoothing processing step of smoothing the input value received in the receiving step; a determining step of determining whether the statistical distance satisfies a predetermined condition; and an equation changing step of changing the arithmetic equation used in the smoothing process from a first arithmetic equation used when the statistical distance satisfies the predetermined condition to a second arithmetic equation with higher tracking ability when the statistical distance does not satisfy the predetermined condition.

10. a reception unit that receives an input value from a sensor; a data group storage unit for storing a data group consisting of a plurality of input values ​​acquired in advance; a statistical distance calculation unit that calculates a statistical distance between the input value received by the reception unit and the data group; a smoothing processing unit that smoothes the input value received by the receiving unit; a determination unit that determines whether the statistical distance satisfies a predetermined condition; A calculation program that causes a computer to function as a formula change unit that changes the calculation formula used in the smoothing process from a first calculation formula that is used when the statistical distance satisfies the specified condition to a second calculation formula that has better tracking performance when the statistical distance does not satisfy the specified condition.