JP2006037245A - Width-direction position correspondence identification method and sheet-like product manufacturing apparatus using the same - Google Patents

Abstract

PROBLEM TO BE SOLVED: To measure and switch the position correspondence between the operation end and the measurement point for each brand, but it takes time, and if the position correspondence shifts during operation, measure the step response and measure the position correspondence again. There was a problem in quality control.
SOLUTION: An operation amount at an operation end is input to a process model, and the positional correspondence relationship, interference width, and process gain of the process model are corrected so that the deviation between the process model and the measurement profile is minimized, and this positional correspondence is also achieved. The relation is set to the width direction controller that outputs the operation amount. Since the positional correspondence relationship can be corrected to the optimum value sequentially during operation, the controllability does not deteriorate even if the positional correspondence relationship is shifted, and the trouble of measuring and switching the positional correspondence relationship for each brand is eliminated.
[Selection] Figure 1

Description

  The present invention relates to a width direction position correspondence identification method capable of automatically identifying a position correspondence between an operation end for controlling a width direction profile and a measurement point for measuring the profile in an apparatus for manufacturing a sheet-like product, and The present invention relates to a sheet-like product manufacturing apparatus using the same.

  The prior art for identifying the positional correspondence in the width direction includes the following.

JP 09-316791 A JP 09-049185 A JP 09-132892 A

  FIG. 14 shows a configuration of a portion for controlling the profile in the width direction in an apparatus for manufacturing a sheet-like product such as paper. Hereinafter, a paper manufacturing apparatus will be described. In FIG. 14, pulp as a raw material is discharged from the gap between the slice lips 41 to the wire part 42 and formed into a sheet shape. The width of the gap between the slice lips 41 is adjusted by the slice bolt 43. The pulp in sheet form is removed of moisture while being conveyed on the wire part 42 in the direction of the arrow 44, and the profile of its thickness is measured by the BM frame 45.

  While the interval between the slice bolts 43 is 35 to 100 mm, the measurement interval of the BM frame 45 is about 5 mm, so that a plurality of measurement points correspond to one slice bolt. Which measurement point corresponds to which slice bolt cannot be determined only by the geometric relationship. Therefore, tuning of these position correspondences is performed at startup.

  FIG. 15 shows the flow of position correspondence tuning. First, an automatic step response test is performed in which a step-like operation amount is given to the operation end (slice bolt 43) in (15-1), and a change in the width direction profile corresponding to the operation amount is measured. For this, for example, the technique disclosed in Patent Document 1 can be used.

  Next, in (15-2), the result of this step response test is analyzed, and the position of the measurement point corresponding to each operation end is determined individually. For this determination method, for example, the technique disclosed in Patent Document 2 can be used. Then, the individual position correspondence is smoothly interpolated to finally determine the overall position correspondence for determining the position correspondence of all the operation ends (15-3). For this, for example, the technique disclosed in Patent Document 3 can be used.

  During the execution of the automatic step response test, the number of operations and the operation amount at the operation end are limited, so it is difficult to maintain a good profile, and it is difficult to perform the automatic step response test during operation for quality control. . Therefore, the position correspondence between the operation end and the measurement point is determined based on the flow of FIG. 15 at the start of control, and this position correspondence is fixedly used during operation.

  However, the method for determining the positional correspondence between the operation end and the measurement point has the following problems.

  When performing an automatic step response test, the tuning operator must be monitoring. However, when the number of operation ends is large, the number is 100 to 200. Therefore, there has been a problem that a great load is applied to the tuning worker.

  In addition, when the brand changes, the position correspondence also changes. For this reason, the position correspondence relationship is individually stored for each brand, and when the brand is changed, the stored position correspondence relation is called and reset. However, an automatic step response test must be performed for each brand to identify the positional correspondence, and there is a problem that a great amount of man-hour is required for the tuning work.

  Furthermore, even if the brand is the same, the positional relationship changes if the operating conditions such as basis weight and paper making speed, or other production conditions change, but it is not possible to cope with such a case and the controllability deteriorates. There was also a problem of producing defective products.

  Accordingly, an object of the present invention is to provide a width direction positional correspondence identification method capable of automatically determining and resetting the positional correspondence between an operation end and a measurement point during operation, and a sheet-like product manufacturing apparatus using the same. It is to provide.

In order to achieve such a problem, the invention according to claim 1 of the present invention is:
A plurality of operation ends for controlling the width direction profile of the sheet-like product, and a width direction control unit for outputting a target profile and a measurement profile in the width direction of the sheet-like product and operating amounts for operating the plurality of operation ends; In an identification method for identifying a positional correspondence between an operation end for controlling the width direction profile and a measurement point for measuring the profile, for a sheet-shaped product manufacturing apparatus comprising:
Receiving an operation amount output by the width direction control unit, performing a model calculation by a process model that simulates a process including the plurality of operation ends;
Receiving a deviation between the model calculation output and the measurement profile, and calculating an optimum value of a position correspondence relationship, an interference width and a process gain such that the deviation is minimized;
Setting the optimum values of the position correspondence, interference width and process gain to the position correspondence, interference width and process gain in the process model;
Setting the optimum value of the position correspondence relationship to the position correspondence relationship in the width direction control unit,
These settings are made in the same or longer cycle as the control timing in the width direction control unit. The position correspondence can be corrected during the width direction control.

The invention according to claim 2 is the invention according to claim 1,
The process model uses a normalized distribution function as a width direction profile response corresponding to an input operation amount. Empirically, the profile response can be approximated accurately.

The invention according to claim 3 is the invention according to claim 1,
As the profile response in the width direction, a time delay and a first-order lag response are added to the normalized distribution function. More realistic processes can be modeled.

The invention according to claim 4 is the invention according to claim 3,
As the profile response in the width direction, the following equation (5) is used. Model more real processes

ここで、Ｋはプロセスゲイン、ｎは操作量が入力されてからそのプロファイル応答が出力されるまでのサンプリング周期の数、Ｔ_Ｏはサンプリング周期、Ｌはむだ時間、Ｔは１次遅れの時定数、Ｕ_ｊ（ｎ）はｎサンプリング周期前に入力された操作量である。

Here, K is the process gain, the number of sampling periods from when n is inputted by the operation amount until the profile response is outputted, T _O is the sampling period, time's L ham, T is the first-order-delay time constant , U _j (n) is an operation amount input before n sampling periods.

The invention according to claim 5 is the invention according to any one of claims 1 to 4,
In the calculation of the optimum value, the correction amount of the position correspondence is obtained by the steepest descent method. A solution can be obtained easily and quickly.

The invention according to claim 6 is the invention according to claim 5,
In the calculation of the optimum value, the correction amount obtained by the steepest descent method is interpolated by an interpolation operation using a neural network. Variations in the width direction of the position-corresponding correction amount can be smoothly interpolated.

The invention according to claim 7 is the invention according to claim 5 or claim 6,
A predetermined limit value is provided, and the correction amount is corrected so that the correction amount does not exceed the limit value. Optimization stability can be achieved.

The invention according to claim 8 is the invention according to any one of claims 5 to 7,
The average value of the absolute value of the correction amount of the position correspondence relationship is added for a predetermined time, and the correction step width of the position correspondence relationship is corrected by the added value. The correction step width can be automatically corrected.

The invention according to claim 9 is the invention according to claim 8,
A new correction step width is obtained based on the following equation (6). The correction step width can be automatically corrected.
D _m = D _m ^′ × F _S / F (6)
Here, D _m is the new correction step width, D _m ^′ is the previous correction step width, F _S is the position corresponding change amount setting value, and F is the addition value.

The invention according to claim 10 is:
A plurality of operation ends for controlling the width direction profile of the sheet-like product;
A width direction control unit that outputs a target profile and a measurement profile in the width direction of the sheet-like product, and outputs an operation amount for operating the plurality of operation ends;
A process model that receives the manipulated variable and simulates the process;
Optimize the positional correspondence between the plurality of operating ends and the plurality of measurement points for measuring the measurement profile and the interference width of the process model so that the deviation between the output of the process model and the measurement profile is minimized, A position correspondence optimization unit that sets the position correspondence relationship and the interference width in the process model, and sets the position correspondence relationship in the width direction control unit;
Is provided. The position correspondence can be corrected during the width direction control.

The invention of claim 11 is the invention of claim 10,
The process model uses a normalized distribution function as a width direction profile response corresponding to an input operation amount. Empirically, the profile response can be approximated accurately.

The invention of claim 12 is the invention of claim 10,
As the profile response in the width direction, a time delay and a first-order lag response are added to the normalized distribution function. More realistic processes can be modeled.

The invention of claim 13 is the invention of claim 12,
The following formula (7) is used as the profile response in the width direction. More realistic processes can be modeled.

ここで、Ｋはプロセスゲイン、ｎは操作量が入力されてからそのプロファイル応答が出力されるまでのサンプリング周期の数、Ｔ_Ｏはサンプリング周期、Ｌはむだ時間、Ｔは１次遅れの時定数、Ｕ_ｊ（ｎ）はｎサンプリング周期前に入力された操作量である。

Here, K is the process gain, the number of sampling periods from when n is inputted by the operation amount until the profile response is outputted, T _O is the sampling period, time's L ham, T is the first-order-delay time constant , U _j (n) is an operation amount input before n sampling periods.

The invention according to claim 14 is the invention according to any one of claims 10 to 13,
The position correspondence optimization unit obtains the correction amount of the position correspondence by the steepest descent method. A solution can be obtained easily and quickly.

The invention according to claim 15 is the invention according to claim 14,
The position correspondence optimization unit is configured to interpolate the correction amount obtained by the steepest descent method by an interpolation operation using a neural network. Smoothly interpolate the variation in the width direction of the position correction amount

The invention according to claim 16 is the invention according to claim 14 or claim 15,
A predetermined limit value is provided, and the correction amount is corrected so that the correction amount does not exceed the limit value. Optimization stability can be achieved.

The invention according to claim 17 is the invention according to any one of claims 14 to 16,
The average value of the absolute value of the correction amount of the position correspondence relationship is added for a predetermined time, and the correction step width of the position correspondence relationship is corrected by the added value. The correction step width can be automatically corrected.

The invention according to claim 18 is the invention according to claim 17,
A new correction step width is obtained based on the following equation (8). The correction step width can be automatically corrected.
D _m = D _m ^′ × F _S / F (8)
Here, D _m is the new correction step width, D _m ^′ is the previous correction step width, F _S is the position corresponding change amount setting value, and F is the addition value.

As is apparent from the above description, the present invention has the following effects.
According to the first, second, third, fourth, sixth, seventh, eighth, ninth, tenth, eleventh, twelfth, thirteenth, fourteenth, fifteenth, seventeenth, and seventeenth and seventeenth aspects of the invention, The model calculation is performed by a process model that simulates the process including, and the optimum value of the position correspondence relationship, the interference width, and the process gain that minimizes the deviation between the output of the model calculation and the actual measurement profile is calculated. The optimum values of the positional correspondence, interference width, and process gain are set in the process model and the width direction controller.

  Since the positional correspondence can be sequentially corrected during operation, there is an effect that the controllability does not deteriorate even if the operation condition changes. In addition, since it is not necessary to identify the position correspondence by performing step response during operation, there is an effect that the profile is not changed and a great burden is not imposed on the tuning work.

  Further, by correcting the correction step width from the correction amount of the position correspondence relationship, there is an effect that the tuning of the correction step width, which has conventionally relied on intuition and experience, can be automated.

  Hereinafter, the present invention will be described in detail with reference to the drawings.

  FIG. 1 is a block diagram showing an embodiment of a sheet-like product manufacturing apparatus according to the present invention. In FIG. 1, 1 is a width direction controller, and the deviation between the target profile SV and the measurement profile P is input. The width direction controller 1 calculates and outputs an optimum operation change amount U from this deviation. 2 is a process for producing a sheet-like product, and an operation change amount U is inputted. Process 2 produces a sheet-like product and outputs a measurement profile P in the width direction.

  Reference numeral 3 denotes a position correspondence identifier, which includes a position correspondence optimization unit 31 and a model 32. The model 32 simulates the process 2, and the operation change amount U is input and the profile response Y is output. The deviation between the profile response Y and the measurement profile P output by the process 2 is input to the position correspondence optimization unit 31.

  The position correspondence optimization unit 31 corrects the model 32 by performing an optimization calculation of the position correspondence relation, the interference width, and the process gain from the deviation between the input profile response Y and the measurement profile P. Further, the position correspondence relationship of the width direction controller 1 is reset.

  FIG. 2 is a flowchart showing the operation of the position correspondence identifier 3. This flow is executed at every width direction control timing. After waiting until the control timing in the width direction (2-1), model calculation is performed by the model 32 (2-2). Next, the position correspondence optimization unit 31 optimizes the position correspondence, the interference width, and the process gain (2-3).

  Next, the position correspondence relationship, the interference width and the process gain are set in the model 32, and the position correspondence relationship is set in the width direction controller 1. These settings are the same as or longer than the width direction control timing. Perform in cycles. Therefore, the counter n is incremented in (2-4), and if the count value is smaller than a predetermined adjustment cycle, the model calculation is repeated after waiting for the next width direction control timing (2-5).

  When the count value of the counter n is equal to or greater than a predetermined adjustment cycle, the position correspondence relationship optimized by the position correspondence optimization unit 31 is interpolated by a neural network in (2-6), and this interpolation is performed in (2-7). The position correspondence relationship is reset to the model 32. In addition, the interference width and process gain are reset in the model 32. Then, in (2-8), the count value of the counter n is set to 0 and waits until the next width direction control timing (2-1).

  Next, each operation of the flowchart of FIG. 2 will be described in detail. For the sake of simplicity, it is assumed that the target profile SV = 0 and the measurement profile P is equal to the deviation input to the width direction controller 1. Further, the number of operation ends is M, and the measurement point of the profile in the width direction is N point.

First, the model calculation will be described. Assume that the position correspondence relationship of the j-th operation end is m (j). This m (j) represents a profile measurement point and actually takes a natural number from 1 to N, but expands the range to a real number and takes a real value up to 1 ≦ m (j) ≦ N. . Further, a model of a profile response to the operation change amount U _j of the width direction controller 1 corresponding to the j-th operation end is considered, but in order to simplify the description, the response delay in the time axis direction is ignored and one sample is taken. Assume that a complete profile response is obtained with a period.

The profile response after one sample period with respect to the operation change amount U _j at the j-th operation end is represented by a normalized distribution function S (i; m (j), σ) shown in the following equation (9). Here, the scale of the normalized distribution function is normalized by the number of measurement points N / M of the profile per zone of the operation end.

  FIG. 3 shows the normalized distribution function S (i; m (j), σ). Note that m (j) = 50, σ = 8, N = 300, and M = 30. This normalized distribution function empirically approximates the profile response well. It can be said that the dispersion σ represents the size of the interference width in the width direction.

Using the normalized distribution function S (i; m (j), σ) in the above equation (9), the profile response Y (i) () by the operation change amount U _j (j = 1,..., M) The model of i = 1,..., N) can be expressed by the following equation (10).

ここで、Ｋはプロセスゲインである。

Here, K is a process gain.

  When the profile does not respond 100% in one sample period, the model can be extended as follows. The response delay of the profile is basically determined by the transport delay of the sheet from the operation end to the measurement point and the low-pass filter installed in the measurement unit in order to attenuate short cycle fluctuations. As the low-pass filter, a primary filter is often used.

Therefore, this time delay can be modeled by a system in which dead time and first-order delay are combined. The dead time and the time constant of the first-order lag may be known. Assuming that the sampling period is T _O , the dead time is L, and the time constant of the first-order delay is T, the current profile response can be expressed by the following equation (11).

  From this equation (11), the profile response Y (i) including the axial direction is expressed by the following equation (12).

ｎに対する総和は、例えばｎ×Ｔ_Ｏ＞Ｌ＋２Ｔとなる最小のｎまで行えば十分である。

It is sufficient that the total sum for n is up to the minimum n, for example, n × T _O > L + 2T.

  Next, optimization of position correspondence, interference width, and process gain will be described. Note that the profile response Y (i) may use either the expression (10) or the expression (12).

  Assuming that the profile response of the actual process 2 is P (i) (i = 1,..., N), the error between the actual process 2 and the model 32 is expressed by the square deviation function J of the following equation (13).

最適化を行うためには、このＪを最小化する位置対応関係ｍ（ｊ）、干渉幅σ、プロセスゲインＫを求めなければならない。

In order to perform optimization, it is necessary to obtain a position correspondence m (j), an interference width σ, and a process gain K that minimize this J.

  Here, an optimum value is obtained by using a steepest descent algorithm which is a well-known optimization technique. For this purpose, first, a differential for each variable of the normalized distribution function S (i; m (j), σ) is calculated by the following equations (14) to (16).

  Using these formulas (14) to (16), the differential for each variable of the square deviation function J is calculated by the following formulas (17) to (19).

  Using the equations (17) to (19), the correction amounts Δm (j) (j = 1,..., M), Δσ, ΔK of the variables one step ahead by the steepest descent method are the following (20) It can be calculated by the formula (22). Here, the correction step width of each variable is set to Dm, Dσ, and DK.

  Next, the neural network interpolation of the position correspondence will be described. The correction amount Δm (j) (j = 1,..., M) of each position correspondence m (j) (j = 1,..., M) can be obtained by the above equation (20). The following problem occurs only with this correction operation. That is, the amount of model correction increases at locations where control operations have been performed greatly, i.e., where the profile deviation is large, but there is a large difference between the process and the model at locations where control operations are small, i.e. where the profile deviation is small. Even so, the amount of modification of the model is small.

  That is, the correction amount is influenced not only by the magnitude of the difference between the process and the model but also by the amount of control operation, and as a result, the correction amount of the position correspondence relationship varies. In order to solve this problem, an algorithm using a neural network is applied.

As a neural network algorithm, a known method as disclosed in Patent Document 3 can be used. FIG. 4 shows an interpolation flow using a neural network. In FIG. 4, Δm (j) obtained by the equation (20) is substituted into the position corresponding deviation H (Nj) in (4-1), and an interpolation operation using a neural network is executed (4-2). . Then, the maximum likelihood position correspondence deviation function Y (i) obtained by this interpolation calculation is added to the position correspondence correction amount Δm ^* (j) (4-3) and m (j) (4-4).

Assuming that the current position correspondence relationship of the model 32 is m (j) (j = 1,..., M),
The positional correspondence after correction is
m (j) = m (j) + m ^* (j) (23)
become. Since the maximum likelihood position correspondence deviation function Y (i) is a function that changes smoothly, the position correspondence m (j) also changes smoothly.

  Next, the updated position correspondence m (j), the interference width σ, and the process gain K are set in the model 32, and the updated position correspondence m (j) is converted into an integer and set in the axial direction controller 1. . In these settings, in order to avoid sudden changes and to ensure the stability of optimization, a limit value is provided for the change amount so that the change amount does not exceed this limit value.

That is, the limit value of the position corresponding correction amount Δm ^* (j) as ± MBand, and Δm ^* (j) is + MBand larger than Delta] m ^* a (j) + mBand, and -MBand smaller than -MBand. Then, this position corresponding correction amount Δm ^* (j) is added to m (j). When expressed in mathematical formulas, the following formulas (24) to (26) are obtained.
Δm ^* (j)> mBand → Δm ^* (j) = mBand (24)
Δm ^* (j) <− mBand → Δm ^* (j) = − mBand (25)
m (j) → m (j) + Δm ^* (j) (j = 1,..., M) (26)
In addition, when setting to the width direction controller 1, it sets after m (j) is converted into an integer.

Limit values are similarly set for the change amount Δσ of the interference width and the change amount ΔK of the process gain. When the limit value of the change amount of the interference width is ± σBand, the change amount Δσ is + σBand if the change amount Δσ is greater than + σBand, and the change amount Δσ is −σBand if the change amount Δσ is less than −σBand, is added to the interference width σ. When expressed in mathematical formulas, the following formulas (27) to (29) are obtained.
Δσ> σBand → Δσ = σBand (27)
Δσ <−σBand → Δσ = −σBand (28)
σ → σ + Δσ (29)

Similarly, the process gain is added to the process gain K as + KBand when the change amount ΔK is larger than + KBand and −KBand when the change amount ΔK is larger than −KBand. When expressed by mathematical formulas, the following formulas (30) to (32) are obtained.
ΔK> KBand → ΔK = KBand (30)
ΔK <−KBand → ΔK = −KBand (31)
K → K + ΔK (32)

  FIG. 5 shows a simulation result of this example. Each column in the table of FIG. 5 is the number of the operation end from the left, the position correspondence output by the model 32, the position correspondence of the process 2, the difference in the position correspondence between the process 2 and the model 32, and the position correspondence optimization unit 31. This is a position corresponding change amount that is an output of. In this simulation, the simulation was performed with 30 operation ends, 300 profile measurement points, 200 control cycles, and 20 adjustment cycles. The step width corresponding to the position is 200.

  As can be seen from FIG. 5, the difference between the position correspondence of the process 2 and the position correspondence of the model 32 at most operation ends coincides with the position correspondence change amount with an error of 0.1 or less. Therefore, it can be understood that the control can be performed assuming that the value corresponding to the position of the process 32 corresponds to the position corresponding to the position of the process 2 by adding the position correspondence change amount to the position correspondence of the model 32.

  FIG. 6 shows simulation results of the interference width and process gain. The difference between the interference width of the process and the interference width of the model 32 is 2, whereas the interference width change amount is 1.737, which is almost the same. Further, while the difference between the process gain 2 and the model 32 is 0.3, the change amount of the process gain is 0.272, which is well matched.

  FIG. 7 is a plot of the position-corresponding change amount every 20 control times in the same simulation, and the horizontal axis is the number of the operation end. The symbol ▲ represents the difference between the position correspondence process 2 and the model 32, and the symbol ■ represents the position correspondence change amount at the control count of 200 times. It can be seen that both agree well.

  FIG. 8 shows the initial and final values of the profile in the same simulation. The horizontal axis is the number of the measurement point, the ● mark is the initial profile, and the line is the final profile. Since random noise having an amplitude of 0.2 is added for each control cycle, the final value includes noise, but it can be seen that the initial value and the final value are almost the same.

  FIG. 9 is a graph showing the operation change amount at that time. The horizontal axis is the number of the operation end. The operation ends 1, 4, and 7 have a small operation change amount and hardly perform any operation, but it can be seen that the position correspondence relationship is accurately identified.

  The correction step width Dm described in the above (20) to (22) is determined empirically, and a specific method for tuning has not been established. FIG. 10 shows a tuning flow conventionally used. In FIG. 10, first, a temporary gain, that is, a correction step width is set in (10-1), and control for position correspondence identification is started in (10-2).

  Then, the state of the paper made in (10-3) and the position correspondence correction state by the position correspondence identifier 3 are monitored to determine whether the state is good (10-4). If the state is not good, the gain (correction step width) is changed based on sense and intuition (10-5), and monitoring is continued (10-3). When the state becomes good, the tuning is finished (10-6).

  As described above, since tuning is performed with a sense and intuition, there are problems that tuning is difficult, individual differences are large, and a large number of man-hours are required. Moreover, since the basis weight of the paper and the composition of the raw materials differ depending on the brand, and the correction step width differs accordingly, there is a problem that more man-hours are required for tuning.

  FIG. 11 shows a tuning method for solving such a problem. This method enables automatic tuning of the correction step width.

A few hours after setting the position correspondence obtained by the identification by the automatic step response described in FIG. 15, or when the product quality such as 2σ or R in the width direction of the paper to be produced is stabilized the calculating the correction step width D _m done.

In the flow of FIG. 2, the position correspondence identifier 3 calculates the position correspondence correction amount Δm ^* (j) for each adjustment period and optimizes the position correspondence relationship. Here, the position correspondence correction amount Δm ^* ( The calculation of j) is performed every control cycle.

First, f (i) is calculated by the following equation (33). Δm ^* (j) is the position-corresponding correction amount described in the equation (23) and the like, and M is the number of operation ends 43.

Further, the number n of times of control of the width direction controller per time is calculated using the following equation (34), and the position corresponding change amount F is obtained by equation (35). Then, the position corresponding correction step width _Dm is calculated by the equation (36). Here F _S position corresponding change amount of the set value (points / operating end number / time), when D _m ^'data collection, namely the previous position corresponding correction step width, i is is a control number in the width direction controller .

F _S is an average value of the position-corresponding correction amount allowed per hour for each operation end, and for example, a value of 20% of the number of measurement points included between the operation end is set. For example, when the number of measurement points per operation end is 5, F _S = 1.000.

Figure 11 shows a flowchart of the automatic tuning correction step width D _m. This auto-tuning is performed several hours after setting the position correspondence identified by the conventional automatic step response test as described above, or when the product quality is stabilized.

  In FIG. 11, i and F are initialized to 0 in (11-1). Then, the number n of times of control of the width direction controller is calculated using the above equation (34) (11-2). In synchronization with the control timing in the width direction in (11-3), f (i) calculated by the equation (33) is added to F (11-4). Next, i is incremented in (11-5), the magnitude relationship between i and n calculated in (11-2) is checked in (11-6), and if i is smaller than n, (11-3) Return.

When i is equal to or larger than n, the correction step width _Dm is calculated using the above equation (36) in (11-7). The correction step width _Dm is displayed on the operation screen. The operator checks the value of the correction step width D _m, is set to a position corresponding identifier 3.

FIG. 12 shows an example of actual data for calculation using the equations (34) to (36). In addition, the meaning of a symbol is the same as said Formula (34)-(36). Since the control cycle T _O = 300 seconds and the position corresponding change amount setting value F _S = 1.000, the following equation is established.

FIG. 13 shows actual data when the correction step width D _m ^′ during data collection is changed to 150.000. The meaning of the symbols is the same as in FIG. Further, the control period T _O to the position corresponding change amount set value F _S was used the same value. The calculation result at this time is as follows.

From these results, it was confirmed that substantially the same corrected step width D _m can be obtained even if the corrected step width D _m ^′ at the time of data collection is changed. Since Dm = 150.000 was closer to 120.000 than Dm = 240.000, it was confirmed that the product quality was better than that of 240.000.

  In the equations (34) to (36), the position correction correction step width is tuned by adding the average of the absolute values of the position correction correction amount for one hour, but not necessarily one hour. Alternatively, it may be arbitrarily set according to the control state.

  In these embodiments, the paper manufacturing apparatus has been described. However, the present invention can also be applied to other sheet-shaped product manufacturing apparatuses such as plastic films. Further, the interpolation calculation by the neural network shown in FIG. 4 is not necessarily performed.

It is a block diagram which shows one Example of this invention. It is a flowchart which shows one Example of this invention. It is a characteristic view showing an example of a normalized distribution function. It is a flowchart which shows the other Example of this invention. It is a characteristic table | surface which shows the effect of one Example of this invention. It is a characteristic table | surface which shows the effect of one Example of this invention. It is a characteristic table | surface which shows the effect of one Example of this invention. It is a characteristic table | surface which shows the effect of one Example of this invention. It is a characteristic table | surface which shows the effect of one Example of this invention. It is a flowchart which shows the tuning of the conventional correction step width. It is a flowchart of the automatic tuning of the correction step width which is one Example of this invention. It is a characteristic table | surface which shows the effect of an automatic correction step width. It is a characteristic table | surface which shows the effect of an automatic correction step width. It is a block diagram which shows an example of arrangement | positioning of an operation end and a measurement point. It is a flowchart which shows the setting method of the conventional position correspondence.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Width direction controller 2 Process 3 Position corresponding identifier 31 Position corresponding optimization part 32 Model

Claims (18)

A plurality of operation ends for controlling the width direction profile of the sheet-like product, and a width direction control unit for outputting a target profile and a measurement profile in the width direction of the sheet-like product and operating amounts for operating the plurality of operation ends; In an identification method for identifying a positional correspondence between an operation end for controlling the width direction profile and a measurement point for measuring the profile, for a sheet-shaped product manufacturing apparatus comprising:
Receiving an operation amount output by the width direction control unit, performing a model calculation by a process model that simulates a process including the plurality of operation ends;
Receiving a deviation between the model calculation output and the measurement profile, and calculating an optimum value of a position correspondence relationship, an interference width and a process gain such that the deviation is minimized;
Setting the optimum values of the position correspondence, interference width and process gain to the position correspondence, interference width and process gain in the process model;
Setting the optimum value of the position correspondence relationship to the position correspondence relationship in the width direction control unit,
A method for identifying a width direction position correspondence relationship, characterized in that these settings are made at the same or longer cycle as the control timing in the width direction control unit.   2. The width direction position correspondence identification method according to claim 1, wherein the process model uses a normalized distribution function as a width direction profile response corresponding to an input operation amount.   2. The width direction position correspondence identification method according to claim 1, wherein a dead time and a first-order lag response are added to the normalized distribution function as the profile response in the width direction. 4. The width direction position correspondence method according to claim 3, wherein the following formula (1) is used as the profile response in the width direction.

Here, K is the process gain, the number of sampling periods from when n is inputted by the operation amount until the profile response is outputted, T _O is the sampling period, time's L ham, T is the first-order-delay time constant , U _j (n) is an operation amount input before n sampling periods.   The method of identifying a width direction position correspondence according to any one of claims 1 to 4, wherein the optimum value is calculated by calculating a correction amount of the position correspondence by a steepest descent method.   6. The method of claim 5, wherein the optimum value is calculated by interpolating a correction amount obtained by the steepest descent method by an interpolation operation using a neural network.   The width direction position correspondence identification method according to claim 5 or 6, wherein a predetermined limit value is provided, and the correction amount is corrected so that the correction amount does not exceed the limit value. .   The average value of the correction values of the position correspondence relationship is added for a predetermined time, and the correction step width of the position correspondence relationship is corrected by the added value. The identification method of the width direction position corresponding relationship in any one of. 9. The method of identifying a width-direction position correspondence relationship according to claim 8, wherein a new correction step width is obtained based on the following equation (2).
D _m = D _m ^′ × F _S / F (2)
Here, D _m is the new correction step width, D _m ^′ is the previous correction step width, F _S is the position corresponding change amount setting value, and F is the addition value. A plurality of operation ends for controlling the width direction profile of the sheet-like product;
A width direction control unit that outputs a target profile and a measurement profile in the width direction of the sheet-like product, and outputs an operation amount for operating the plurality of operation ends;
A process model that receives the manipulated variable and simulates the process;
Optimize the positional correspondence between the plurality of operating ends and the plurality of measurement points for measuring the measurement profile and the interference width of the process model so that the deviation between the output of the process model and the measurement profile is minimized, A position correspondence optimization unit that sets the position correspondence relationship and the interference width in the process model, and sets the position correspondence relationship in the width direction control unit;
An apparatus for producing a sheet-like product, comprising:   The sheet-like product manufacturing apparatus according to claim 10, wherein the process model uses a normalized distribution function as a width direction profile response corresponding to an input operation amount.   11. The sheet-like product manufacturing apparatus according to claim 10, wherein a dead time and a first-order lag response are added to the normalized distribution function as the profile response in the width direction. The sheet-shaped product manufacturing apparatus according to claim 12, wherein the following formula (3) is used as the profile response in the width direction.

Here, K is the process gain, the number of sampling periods from when n is inputted by the operation amount until the profile response is outputted, T _O is the sampling period, time's L ham, T is the first-order-delay time constant , U _j (n) is an operation amount input before n sampling periods.   14. The sheet-like product manufacturing apparatus according to claim 10, wherein the position correspondence optimization unit obtains a correction amount of the position correspondence relationship by a steepest descent method.   15. The sheet-like product manufacturing apparatus according to claim 14, wherein the position correspondence optimization unit interpolates a correction amount obtained by the steepest descent method by an interpolation operation using a neural network.   16. The sheet product manufacturing apparatus according to claim 14, wherein a predetermined limit value is provided, and the correction amount is corrected so that the correction amount does not exceed the limit value.   The average value of the correction values of the position correspondence relationship is added for a predetermined time, and the correction step width of the position correspondence relationship is corrected by the added value. The sheet-like product manufacturing apparatus according to any one of the above. 18. The sheet-like product manufacturing apparatus according to claim 17, wherein a new correction step width is obtained based on the following equation (4).
D _m = D _m ^′ × F _S / F (4)
Here, D _m is the new correction step width, D _m ^′ is the previous correction step width, F _S is the position corresponding change amount setting value, and F is the addition value.

Images