JP6863393B2 - Parameter value determination method, parameter value determination program and parameter value determination device - Google Patents

Description

The present invention relates to a parameter value determination method, a parameter value determination program, and a parameter value determination device.

Conventionally, in mass production of products that output signals such as optical transmission devices, a process of setting the parameter value of the firmware that controls the signal output to an appropriate value so that the signal output falls within a predetermined allowable range is performed. It has been.

On the other hand, in Patent Document 1 and the like, in order to shorten the adjustment time of the adjustment target (for example, the electric motor drive device), the past adjustment targets having the same or similar adjustment specifications as the constants required for the adjustment of the new adjustment target are extracted. However, a technique for setting an initial value of adjustment based on the extracted past adjustment target is disclosed.

However, in the case of a product that outputs a signal, the output signal is greatly affected by the variation in the element value (individual feature amount) of the device mounted on the product, so the parameter value set in another product in the past is diverted. It's difficult. Therefore, in practice, it is necessary to set an appropriate initial value of the parameter for each device, observe the output signal, and change the parameter value so that the output falls within the permissible range.

Japanese Unexamined Patent Publication No. 6-339295

When searching for the parameter value of a product that outputs a signal as described above, it takes time M (for example, M = 1 minute) for the signal output to stabilize when the parameter value is changed. Therefore, if the parameter value is changed N times before the signal output falls within the permissible range, an adjustment time of M × N or more is required. At present, the adjustment time is about 20 to 30 minutes.

In one aspect, it is an object of the present invention to provide a parameter value determination method, a parameter value determination program, and a parameter value determination apparatus capable of shortening the adjustment time of the device that outputs a signal.

In one embodiment, the parameter value determination method is a parameter value determination method for determining a parameter value to be input to an adjustment target device that outputs a signal according to the parameter value, and is a device that has been adjusted in the past. New adjustment based on the data of the combination of the parameter value input to the device when the signal output of the device indicates a value near the boundary of the predetermined allowable range and the individual characteristic amount of the device. A model for estimating a parameter value such that the signal output of the target device indicates a value near the boundary of the permissible range from the individual characteristic amount of the new adjustment target device is set, and the new adjustment target device is set. The correction value for correcting the parameter value estimated using the model in order to make the probability that the signal output of the above is within the allowable range exceeds a predetermined range is based on the change in the signal output with respect to the parameter value predicted within the allowable range. The parameter value is estimated based on the individual feature amount of the new adjustment target device and the model, and the estimated parameter value is corrected by the correction value to obtain the new adjustment target device. This is a parameter value determination method in which a computer executes a process for calculating a parameter value to be input.

The adjustment time of the device that outputs the signal can be shortened.

It is a figure which shows roughly the structure of the adjustment system which concerns on one Embodiment. It is a figure which shows the hardware composition of the adjustment device. It is a functional block diagram of an adjustment device and a product. It is a figure which shows an example of the data structure of the history DB. It is a figure for demonstrating the adjustment of the product to be adjusted in the past. It is a figure which shows the example when the behavior at the time of converging to the boundary of an allowable range is assumed to be a straight line. It is a figure for demonstrating the prediction error. 8 (a) and 8 (b) are diagrams for explaining the probability that the signal output is included in the permissible range when the parameter value is used as the boundary predicted value. 9 (a) and 9 (b) are diagrams for explaining the probability that the signal output is included in the permissible range when the parameter value is corrected by the correction value. It is a figure which shows the example of the case where the behavior when converging to the boundary of the allowable range is assumed to be non-linear.

Hereinafter, the adjustment system according to the embodiment will be described in detail with reference to FIGS. 1 to 10.

FIG. 1 schematically shows the configuration of the adjustment system 100 according to the present embodiment. The adjustment system 100 of FIG. 1 includes an adjustment device 10 as a parameter value determining device connected to the product 50 to be adjusted, and a signal detection device 40 for detecting a signal output from the product 50. The product 50 to be adjusted is, for example, a device such as an optical transmission device that outputs a signal.

As shown in FIG. 3, the product 50 has a signal generator control unit 52 and a signal generator 54. The signal generator control unit 52 incorporates firmware that generates a signal in the signal generator 54 according to a set parameter value. The signal generator 54 generates a signal based on an instruction from the signal generator control unit 52.

The adjusting device 10 searches for a parameter value set in the signal generator control unit 52 of the product 50. Specifically, the adjusting device 10 searches for a parameter value that can include the output signal of the product 50 detected by the signal detecting device 40 in the allowable range.

FIG. 2 shows the hardware configuration of the adjusting device 10. As shown in FIG. 2, the adjusting device 10 includes a CPU (Central Processing Unit) 90, a ROM (Read Only Memory) 92, a RAM (Random Access Memory) 94, and a storage unit (here, HDD (Hard Disk Drive)) 96. It includes an input / output interface 97, a portable storage medium drive 99, and the like. Each component of the adjusting device 10 is connected to the bus 98. In the adjusting device 10, a program (including a parameter value determining program) stored in the ROM 92 or HDD 96, or a program (including a parameter value determining program) read from the portable storage medium 91 by the portable storage medium drive 99 is loaded. By executing the CPU 90, the functions of each part shown in FIG. 3 are realized. Note that FIG. 3 also shows a history DB (database) 22, a model parameter DB 24, a correction value DB 26, and an allowable range DB 28 stored in the HDD 96 or the like of the adjusting device 10.

FIG. 3 shows a functional block diagram of the adjusting device 10 and the product 50. As shown in FIG. 3, the adjustment device 10 has functions as a model setting unit 12, a correction value calculation unit 14, an initial value determination unit 16, and a parameter value output unit 18 when the CPU 90 executes a program. ing.

The model setting unit 12 sets the model parameters of the model used to determine the initial values (before correction) of the parameter values to be input to the new adjustment target product 50. Specifically, the model setting unit 12 sets the model parameters based on the adjusted product adjustment data (final parameter value (final adjustment value) and individual feature amount) stored in the history DB 22. Then, the model setting unit 12 stores the set model parameter in the model parameter DB 24. FIG. 4 shows an example of the data structure of the history DB 22. In the history DB 22 of FIG. 4, the individual feature amounts (X1, X2, ...) Of each adjusted product are stored, and the final adjustment value (y) of each adjusted product is stored. The individual feature amount is also called an element value. Here, in the present embodiment, in the past adjustment, a predetermined parameter value is set as an initial value, and as shown in FIG. 5, a process of repeating signal output while changing the parameter value at predetermined time intervals is performed. Then, when the signal output enters the permissible range, the parameter value at that time is saved as the final adjustment value, and the adjustment is completed. In this way, the adjustment is completed when the signal output is within the permissible range in order to shorten the manufacturing time of the product. Therefore, in the present embodiment, the parameter value when the signal that is close to the permissible range (converged to the boundary of the permissible range) is output is stored in the history DB 22 of FIG. 4 as the final adjustment value. The final adjustment value y varies from product to product as shown in FIG. It should be noted that this variation is due to the difference in the amount of individual characteristics.

The model setting unit 12 is a linear model represented by the following equation (1) based on the individual feature quantities (X1, X2, ...) Stored in the history DB 22 and the final adjustment value (y) of each adjusted product. The constants a ₀ and the coefficients a ₁ , a ₂ , ... Of the model are calculated, and the calculation results are stored in the model parameter DB 24 as model parameters.
y = a ₀ + a ₁ x X1 + a ₂ x X2 + ... (1)

By substituting the individual feature quantities (X1, X2, ...) Of the new product 50 to be adjusted into the above equation (1), it is most likely that the output signal of the product 50 shows a value just within the permissible range. A high parameter value y can be determined. In the following, the parameter value y will also be referred to as a “boundary predicted value”.

The correction value calculation unit 14 calculates a correction value for correcting the parameter value (boundary prediction value) y obtained by using the model set by the model setting unit 12 based on the data of the history DB 22. As described above, the boundary prediction value y is a parameter value at which the output signal of the product 50 is most likely to show a value just within the permissible range. Therefore, the correction value calculation unit 14 calculates the correction value g for correcting the boundary prediction value y to the parameter value at which the signal output is most likely to be included in the allowable range. Here, the corrected parameter value y ₀ can be expressed as y ₀ = y + g. The correction value calculation unit 14 stores the calculated correction value g in the correction value DB 26.

The details of the method of calculating the correction value g by the correction value calculation unit 14 will be described later.

The initial value determination unit 16 reads the model parameters of the model set by the model setting unit 12 from the model parameter DB 24, and reads the correction value calculated by the correction value calculation unit 14 from the correction value DB 26. Then, the initial value determination unit 16 calculates the boundary prediction value y of the product 50 using the read model parameter and the above equation (1), and corrects the boundary prediction value y with the correction value g to correct the boundary prediction value y with the correction value g. _{Calculate the initial value y 0} of the parameter value appropriate for setting to. The initial value determination unit 16 _{transmits the initial value y 0} of the determined parameter value to the parameter value output unit 18. The initial value determining unit 16 can acquire the individual feature amount of the product 50 when the product 50 is connected to the adjusting device 10. The initial value determination unit 16 calculates the initial value of the parameter value based on the model acquisition unit that acquires the model from the model parameter DB 24, the correction value acquisition unit that acquires the correction value from the correction value DB 26, and the model and the correction value. It functions as a calculation unit.

The parameter value output unit 18 outputs the initial value of the parameter value received from the initial value determination unit 16 to the signal generator control unit 52 of the product 50. Further, it is determined whether or not the value of the output signal from the product 50 detected by the signal detection device 40 is within the permissible range stored in the permissible range DB 28, and if it is not within the permissible range, it is determined. The parameter value is changed and the signal is output to the signal generator control unit 52 again. The permissible range DB 28 stores a value indicating the permissible range ([z ₀ , z ₁ ] in the example of FIG. 7).

In the present embodiment, as described above, in the initial value determination unit 16, the product 50 to be adjusted is determined by using the model set by the model setting unit 12 and the correction value g calculated by the correction value calculation unit 14. The initial value of the parameter value at the time of adjustment is determined. In this case, the initial value of the determined parameter value is most likely to be included in the permissible range. Therefore, the parameter value output unit 18 outputs the initial value of the determined parameter value to the product 50, so that the time for searching for the parameter value included in the permissible range can be shortened. This makes it possible to adjust the product 50 in a short time.

Next, a method of calculating the correction value by the correction value calculation unit 14 of the adjustment device 10 of the present embodiment will be described in detail.

In this embodiment, it is assumed that the curve (convergence curve) showing the relationship between the signal output and the parameter value in the vicinity of the permissible range has linearity. That is, the behavior within the permissible range is predicted by assuming that the behavior when converging to the boundary of the permissible range becomes a straight line as shown in FIG. In FIG. 6, the behavior within the permissible range can be expressed by the following equation (2).
z = c ₁ × y + c ₂ … (2)

Note that c ₁ means the sensitivity (predicted value) to the parameter value, and c ₂ means the constant (intercept).

Here, the sensitivity c ₁ has a prediction error, and the boundary prediction value y calculated by using the above-mentioned equation (1) also has a prediction error. FIG. 7 illustrates each prediction error. It is assumed that each prediction error has a normal distribution as shown in FIG. In this case, when the behavior of the new product 50 to be adjusted is estimated using the model of the equation (1), a prediction error shown with hatching will occur.

In the present embodiment, the correction value calculation unit 14 calculates the correction value g for calculating the _{value y 0} that maximizes the probability that the signal output is included in the allowable range.

Specifically, the signal output z is defined as the prediction error of the parameter value (boundary prediction value) y that converges to the boundary of the allowable range is σ _e ² , the average of the sensitivity c _{1 is cave} , and the sensitivity prediction error is σ _c ^2. The existence probability of the product 50 to be adjusted with respect to is expressed as a function of the correction value g. Then, the correction value g is calculated so that the probability of entering the allowable range is maximized. FIG. 8B is a diagram showing the probability that the signal output is included in the permissible range as the area of the hatched range when the parameter value is set as the boundary predicted value as shown in FIG. 8A. .. Further, FIG. 9B shows a hatched range of the probability that the signal output is included in the allowable range when the parameter value is deviated from the boundary predicted value by the correction value g as shown in FIG. 9A. It is a figure which shows as the area of. Here, the existence probability of the product 50 can be expressed by the area of the normal distribution as shown in FIGS. 8 (b) and 9 (b). In the present embodiment, the correction value g is calculated so that the area indicating the existence probability of the product 50 within the permissible range is maximized. Here, the permissible range is [z ₀ , z ₁ ].

In this case, if c _{1 to} N ( _cave , σ _c ² ) and Δz = z ₁ −z ₀ , then the average E [Δz] and the variance Var [Δz] of Δz are given by the following equations (3) and (4). ) Can be expressed.

Then, g can be obtained by substituting the above equations (3) and (4) into the following equation (5).

In the present embodiment, the correction value calculation unit 14 can calculate the correction value g that can maximize the possibility that the signal output is included in the permissible range by solving the above equation (5).

(Modification example)
In the above example, the curve (convergence curve) showing the relationship between the signal output and the parameter value in the vicinity of the permissible range is assumed to be linear, but it may be assumed to be non-linear. Assuming non-linearity, as shown in FIG. 10, using the slope of the convergence curve z = f (y) at _{the boundary prediction value μ y of the newly adjusted product 50, it was assumed to be approximately the above-mentioned linearity.} The method of the case may be used.

Specifically, the following equation (6) in _{μ y} is approximated to z = f (y).

Then, the average _{cave of c 1} and the variance σ _c ² of the above equation (6) may be obtained, and the correction value g may be obtained by using the above-mentioned method assuming linearity.

( _{How to find the mean cave} and variance σ _c ² (when the convergence curve is a quadratic polynomial))
Here, when the convergence curve is a quadratic polynomial, the mean _{cave of c 1} and the variance σ _c ² can be obtained as follows.

Let the two-dimensional polynomial be z = f (y) = (1/2) ay ² + by + c, and the parameters (a, b, c) have already been estimated, and the mean μ _a , μ _b , variance σ _a ² , σ _b. ² is known. In this case, if the parameters (a, b, c) of the two-dimensional polynomial f (y) are set as θ, f (y) = f (y; θ), and f (y; θ) is differentiated by y, then f '(Y ; θ ) = ay + b. Therefore, the average _{cave of c 1} and the variance σ _c ² can be obtained from the following equations (7) and (8). Note that μ _y is the average of y.

( _{How to find the mean cave} and variance σ _c ² (when the convergence curve is a high-dimensional polynomial))
When the convergence curve is a high-dimensional polynomial, the mean _{cave of c 1} and the variance σ _c ² can be obtained as follows.

Let θ be the parameter of the function f (y) of the high-dimensional polynomial, and _{estimate the mean μ θ} and variance σ _θ ² of θ from the following equation (9) using nearby data.
f (y; θ) = Σ _n θ _n y ⁿ … (9)

Further, f (y; θ) is differentiated by y to calculate f'(y; θ). In this case, f'(y; θ) is expressed as _{the polynomial Σ n n} θ n y ^{n-1 of y.}

In this case, the average _{cave of c 1} can be expressed by the following equation (10).
c _ave = E [f'(μ _y ; θ)] = Σ _n nE [θ _n y ^n-1 ]… (10)

In this case, assuming the independence of θ and y,
E [θ _n y ^n-1 ] = E [θ _n ] E [y ^n-1 ] = μ θ _n E [y ^n-1 ]… (11)
Will be.

Here, ^{by Taylor-expanding y n-1} to the _{second order with μ y} , the average E [y ^n-1 ] is _{approximated by μ y} and σ _e ² (Delta Method), and the following equation (12) Will be.

Further, in order to obtain ^{E [y n-1 ], y n-1} is expanded by T y _{lor around μ y} , and ^{the variance of y n-1} (Var (y ^n-1 )) is μ _y , σ. _{Approximating with e} ² , it becomes as shown in the following equation (13).

By using the above equations (12) and (13), the average and variance of the power expression of any random variable can be obtained.

Further, since σ _c ² can also be expressed as a polynomial of y, it can be expressed as in the following equation (14) by using _{μ y} and σ _e ^{2 in the same manner as described above.} It is _{assumed that θ n} and θ _k (n ≠ k) are independent.

The mean and variance of the power expression of θ _n can also be calculated in the same manner as in the above equations (12) and (13).

( _{How to find the mean cave} and variance σ _c ² (when the convergence curve is a non-linear curve that is not a polynomial))
Even when the convergence curve is a non-linear curve that is not a polynomial, it can be approximately expanded to a polynomial by Taylor expansion and applied to the above-mentioned method. For example, when the equation of the convergence curve is f (y) = Ae ^ay , it may be treated as a polynomial by Taylor expansion as in the following equation (15).

As described in detail above, according to the present embodiment, when the signal output of the product to be adjusted in the past shows a value near the boundary of the predetermined allowable range, the model setting unit 12 is said to be the same. To estimate the boundary prediction value y of the product 50 to be newly adjusted from the individual feature amount based on the combination data (data of the history DB 22) of the parameter value input to the product and the individual feature amount. After setting the model, the correction value calculation unit 14 sets the correction value g that corrects the boundary prediction value y in order to maximize the probability that the signal output of the product 50 falls within the allowable range, and sets the correction value g that is predicted within the allowable range. It is calculated based on the change of the signal output with respect to. Then, the initial value determining unit 16 calculates (estimates) the boundary predicted value y based on the individual feature amount of the product 50 and the model, and corrects the calculated boundary predicted value y with the correction value g, so that the product 50 Calculate _{the initial value y 0} of the parameter value to be input to. As a result, even if there is only a parameter value when the output signal indicates a value near the boundary of the allowable range as the data of the product adjusted in the past, the parameter to be input to the product 50 to be newly adjusted. As the initial value of the value, the parameter value having the maximum probability that the signal output falls within the allowable range can be calculated. Therefore, it is possible to shorten the time for searching the parameter value in which the signal output falls within the permissible range in the product 50, so that the adjustment time for the product 50 can be shortened. As a result, the manufacturing lead time can be shortened and the production efficiency can be improved. Further, since the number of products that can be adjusted can be increased without increasing the number of adjusting devices 10, it is possible to reduce the cost.

Further, in the present embodiment, it is assumed that the change in the signal output with respect to the parameter value predicted within the permissible range is a linear change or a non-linear change, and the correction value calculation unit 14 responds to the change in the signal output. Since the correction value is calculated by the method, an appropriate value can be calculated as the correction value according to the behavior within the permissible range.

Further, in the present embodiment, the correction value calculation unit 14 calculates the correction value by linear approximation when the change in the signal output with respect to the parameter value shows a non-linear change within the permissible range. As a result, the correction value can be calculated by a simple calculation.

Further, in the present embodiment, the correction value calculation unit 14 calculates the correction value based on the error (dispersion) σ _e ² of the boundary prediction value and the sensitivity prediction error (dispersion) σ _c ² , so that each error It is possible to calculate an appropriate correction value in consideration of.

In the above embodiment, the correction value calculation unit 14 has described the case where the correction value g is calculated as the correction value g in order to maximize the probability that the signal output of the product 50 falls within the permissible range, but the present invention is limited to this. is not it. For example, the correction value calculation unit 14 may calculate the correction value g whose probability is equal to or higher than a predetermined value.

The model setting unit 12 and the correction value calculation unit 14 may execute (update) the model setting and the correction value calculation at predetermined period intervals, or each time a predetermined number of data are newly added to the history DB 22. You may do it.

In the above embodiment, the case where the adjustment device 10 has the model setting unit 12 and the correction value calculation unit 14 has been described, but the present invention is not limited to this, and the model setting unit 12 and the correction value calculation unit 14 may be provided by another device. You may have. That is, the model may be set and the correction value may be calculated by another device, and the adjustment device 10 may acquire the model and the correction value from the other device. The separate device may be a device directly connected to the adjusting device 10 or a server (cloud server) connected via a network.

The above processing function can be realized by a computer. In that case, a program that describes the processing content of the function that the processing device should have is provided. By executing the program on a computer, the above processing function is realized on the computer. The program describing the processing content can be recorded on a computer-readable recording medium (however, the carrier wave is excluded).

When a program is distributed, it is sold in the form of a portable recording medium such as a DVD (Digital Versatile Disc) or a CD-ROM (Compact Disc Read Only Memory) on which the program is recorded. It is also possible to store the program in the storage device of the server computer and transfer the program from the server computer to another computer via the network.

The computer that executes the program stores, for example, the program recorded on the portable recording medium or the program transferred from the server computer in its own storage device. Then, the computer reads the program from its own storage device and executes the processing according to the program. The computer can also read the program directly from the portable recording medium and execute the processing according to the program. In addition, the computer can sequentially execute processing according to the received program each time the program is transferred from the server computer.

The embodiments described above are examples of preferred embodiments of the present invention. However, the present invention is not limited to this, and various modifications can be made without departing from the gist of the present invention.

10 Adjustment device (parameter value determination device)
12 Model setting unit 14 Correction value calculation unit 16 Initial value setting unit (model acquisition unit, correction value acquisition unit, calculation unit)

Claims (9)

It is a parameter value determination method that determines the parameter value to be input to the device to be adjusted that outputs a signal according to the parameter value.
The combination of the parameter value input to the device when the signal output of the device to be adjusted in the past shows a value near the boundary of the predetermined allowable range and the individual characteristic amount of the device. Based on the data, a model is set for estimating the parameter value such that the signal output of the device to be adjusted newly indicates the value near the boundary of the allowable range from the individual characteristic amount of the device to be adjusted. ,
A correction value that corrects a parameter value estimated using the model in order to increase the probability that the signal output of the new device to be adjusted falls within the allowable range is a parameter value predicted within the allowable range. Calculated based on the change in signal output to
A parameter value is estimated based on the individual feature amount of the new adjustment target device and the model, and the estimated parameter value is corrected by the correction value to input the parameter value to the new adjustment target device. To calculate,
A method for determining parameter values, characterized in that the processing is performed by a computer. The parameter value determination according to claim 1, wherein in the process of calculating the correction value, a correction value that maximizes the probability that the signal output of the new device to be adjusted falls within the permissible range is calculated. Method. Assuming that the change in signal output with respect to the parameter value predicted within the permissible range is a linear change or a non-linear change.
The parameter value determination method according to claim 1 or 2, wherein in the process of calculating the correction value, the correction value is calculated by a method according to a change in the signal output. When the change in the signal output with respect to the parameter value predicted within the permissible range is a non-linear change,
The parameter value determination method according to claim 3, wherein in the process of calculating the correction value, the non-linear change is linearly approximated to calculate the correction value. In the process of calculating the correction value, the correction value is calculated based on the estimation error of the parameter value estimated using the model and the prediction error of the change in the signal output with respect to the parameter value predicted within the permissible range. The parameter value determination method according to any one of claims 1 to 4, wherein the parameter value is calculated. A parameter value determination program that determines the parameter value to be input to the device to be adjusted that outputs a signal according to the parameter value.
The combination of the parameter value input to the device when the signal output of the device to be adjusted in the past shows a value near the boundary of the predetermined allowable range and the individual characteristic amount of the device. Based on the data, a model is set for estimating the parameter value such that the signal output of the device to be adjusted newly indicates the value near the boundary of the allowable range from the individual characteristic amount of the device to be adjusted. ,
A correction value that corrects a parameter value estimated using the model in order to increase the probability that the signal output of the new device to be adjusted falls within the allowable range is a parameter value predicted within the allowable range. Calculated based on the change in signal output to
A parameter value is estimated based on the individual feature amount of the new adjustment target device and the model, and the estimated parameter value is corrected by the correction value to input the parameter value to the new adjustment target device. To calculate,
A parameter value determination program characterized by having a computer execute processing. It is a parameter value determining device that determines the parameter value to be input to the device to be adjusted that outputs a signal according to the parameter value.
The combination of the parameter value input to the device when the signal output of the device to be adjusted in the past shows a value near the boundary of the predetermined allowable range and the individual characteristic amount of the device. A model for estimating a parameter value set based on the data such that the signal output of the new device to be adjusted indicates a value near the boundary of the permissible range from the individual characteristic amount of the device to be adjusted. Model acquisition department to acquire and
Using the model to make the probability that the signal output of the new device to be adjusted falls within the permissible range, which is calculated based on the change in the signal output with respect to the parameter value predicted within the permissible range. A correction value acquisition unit that acquires a correction value that corrects the estimated parameter value,
A parameter value is estimated based on the individual feature amount of the new adjustment target device and the model, and the estimated parameter value is corrected by the correction value to input the parameter value to the new adjustment target device. And the calculation unit that calculates
A parameter value determining device comprising. The parameter value determining device according to claim 7, further comprising a model setting unit for setting the model. The parameter value determining device according to claim 7 or 8, further comprising a correction value calculation unit for calculating the correction value.

Images