KR930007378B1 - Apparatus for adjusting shape of strip shaped metal material or plate-shaped metal material - Google Patents

Abstract

No content.

Description

Apparatus for adjusting the widthwise shape of a strip-shaped or plate-shaped metal material while running

1 is a schematic diagram showing a system arrangement of an aluminum foil rolling target shape adjusting device according to an embodiment of the present invention.

2 is a block diagram showing a processing flow of the copper foil rolling target shape adjusting device.

FIG. 3 (a) is an explanatory diagram showing main portions of real shape data showing elongation distribution in the width direction of foil; FIG.

3 (b) is an explanatory diagram showing the actual shape classification items of the aluminum foil patterned.

FIG. 4A is a flowchart showing a processing procedure for determining a change trend of the actual shape data.

4 (b) is an explanatory diagram showing an action of correcting the number of levels of the actual shape classification items in addition to the trend of shape change.

5 (a) and 5 (b) are graphs showing the correlation of the number of levels of two real shape classification items.

6 is a three-dimensional graph showing a three-dimensional pattern of the aluminum foil by the change of the actual shape data over time;

7 is a schematic diagram conceptually showing a neural network.

8 is an explanatory diagram showing an example of a relationship between a shape change target for a real shape classification item and an action candidate corresponding thereto.

FIG. 9 is a graph showing the relationship between the importance of shape change targets and the confidence level when the actual shape category is extended.

FIG. 10 is an explanatory diagram showing an example in which an action candidate inference unit is used for inferring and an example of changing a target shape using the inference.

FIG. 11 is an explanatory diagram showing a routine of a routine for checking the validity of an applied action by a check tree; FIG.

Fig. 12A is an explanatory diagram showing target shape adjustment parameters used to change the target shape.

Fig. 12B is a state diagram showing the image change of a3 of the parameter.

FIG. 12C is a state diagram showing a situation where the central portion of the target shape adjusted by a4 of the parameter is a forward pattern.

Fig. 12 (d) is a state diagram showing a situation where the central portion is an inverse pattern.

13 is a schematic explanatory diagram showing an example of inference execution for changing the target shape.

14 is a flowchart showing a processing flow of target shape adjustment.

15 is a view showing an input menu displayed on the screen of the rolling mill side terminal.

16 is a display diagram showing an example of a target shape displayed on the screen.

17 is a flowchart showing a method of correcting a real shape obtained asymmetrically.

18 is a schematic explanatory diagram showing a situation of correcting the asymmetric real shape.

19 is a schematic configuration diagram showing a continuous annealing apparatus applied as another example of the present invention.

20A to 21B are graphs showing the actual shape of the coil member in the width direction of the continuous annealing plant of FIG.

Figure 22 is a schematic perspective view showing a roll rolling mill as an example of the background of the present invention.

23 is an external view showing the surface shape of aluminum foil after rolling.

FIG. 24 is an explanatory diagram showing the correlation between the cross-sectional shape of the rolled roll, the actual shape of the aluminum foil, and the target shape for controlling the actual shape.

FIG. 25 is a graph showing simultaneously the target shape set in the operation phase of the aluminum foil and the control phase.

The present invention relates to an apparatus for adjusting the widthwise shape of a belt-shaped or plate-shaped metal material while traveling, and in particular, a shape control unit for controlling the shape in the widthwise direction of the belt-shaped or plate-shaped metal material while traveling by performing a predetermined treatment. The present invention relates to a target shape adjusting device for a metal material which gives target shape data to and adjusts the shape.

으로 감소된다. 그리고 압연된 알루미늄박(53)은 출구측 코일(64)(제1도)의 구동축의 회전구동에 의하여 발생하는 일정한 장력에 의해서 화살표 K방향으로 반송되어, 상기 출구측 코일(64)에 감겨진다. 예컨대 두께 수백 ㎛의 원료알루미늄박(51)을 최종적으로 두께수 ㎛의 알루미늄박(53)으로 압연하는 경우에는, 압연공정이 수회 반복되게 되고, 이 압연횟수는 패스 횟수라고 칭한다.22 shows a roll mill 2 for rolling aluminum foil as an example of the background of the present invention. In aluminum foil rolling, the raw aluminum foil 51 having a width of about 700 to 1700 mm and a thickness of several to several hundreds of micrometers wound on the inlet coil 50 is a pair of rolling rolls 52 at a speed of about 300 to 1200 m / min. Rolled) and its thickness is about

Is reduced. And the rolled aluminum foil 53 is conveyed in the arrow K direction by the constant tension which arises by the rotational drive of the drive shaft of the exit side coil 64 (FIG. 1), and is wound up by the said exit side coil 64. . For example, when the raw material aluminum foil 51 of several hundred micrometers in thickness is finally rolled into the aluminum foil 53 of several micrometers in thickness, a rolling process is repeated several times, and this rolling frequency is called a pass frequency.

In the above-described micro-micron metal rolling, the aluminum foil 53 is "stretched" in the width direction (arrow L) of the foil, although the thickness thereof is the same as shown in FIG. There are markedly "sites" and "extended" sites. In other words, the stretched portion 54 is formed with a crest portion 56 and a valley 57 portion along the conveying direction (arrow K) of the aluminum foil 53, and the stretched portion 55 is approximately flat. It has a shape. Therefore, the aluminum foil 53 shown in the drawing is in a state where the center portion of the thin width direction (arrow L) is stretched and the end portion thereof is extended.

The distribution of the elongated state and elongated state in the thin width direction (arrow L) is hereinafter referred to as the surface shape or the real shape of the aluminum foil 53.

The surface shape has a considerable influence on the quality of the foil product, and in some cases, a large tension is applied to the stretched portion 55 to cause breakage of the foil. In addition, the stretched portion 54 causes wrinkles. And it is natural that the flat shape which extended | stretched and expanded equally about the aluminum foil 53 as a final product is desired, but it is not necessarily a flat thread shape for every path | pass, The aluminum foil 53 in the middle path | pass There are many different shapes.

The surface shape of the aluminum foil 53 as described above can be controlled by changing the shape of the rolling roll 52.

As shown in Figs. 22 to 24, the rolling roll 52 has an expansion called a thermal crown due to the heat generation during rolling and its heat conduction characteristics.

In the example shown in FIG. 24, the quarter portion a is inflated. Such expanded portion, that is, the hot crown, is rolled at the rolled portion of the aluminum foil 53 rolled at the place where its appearance and the degree of expansion are large. Therefore, the surface shape of the aluminum foil 53 is the width of the aluminum foil 53 by the temperature or the injection amount of the coolant 58 (FIG. 1) injected toward the rolling roll 52 to cool the rolling roll 52. It can be controlled by changing in the direction (arrow L, FIG. 23).

The shape control of such aluminum foil 53 is performed by the shape control part 3 provided adjacent to the roll mill 2. That is, the shape control part 3 is made of aluminum foil from the inspection roll 4 composed of 36 elements 4e divided in the thin width direction (arrow mark L) by a rotational movement freely installed at the exit side of the rolling roll 52. The stretched / extended real shape data of (53) is input. The piezoelectric element (not shown) of (1) is embedded in each element 4e, and acts as a sensor which detects the piezoelectric force applied to the outer peripheral surface of the element 4e.

Then, the aluminum foil 53 pressed on the element 4e and drawn in the conveying direction (arrow K) by a constant tension is the element when its stretching portion 54 passes on the element 4e. The pressing force with respect to 4e is small, and largely detected when the reversely extended portion 55 passes.

Therefore, as shown in FIG. 24, the actual shape of the aluminum foil 53 is displayed as distribution (real data) of the elongation rate of the elongation rate which converted the pressure-contacting day difference detected from each element 4e. In the case of illustration, since the quarter part a of the rolling roll 52 is expanded, the target shape is set so as to accelerate the cooling of the portion and to accumulate the central portion of the rolling roll 52 and both ends thereof.

The shape control unit 3 compares the real shape data with the target shape data input in advance, and the coolant is injected toward the site of the rolling roll corresponding to the element 4e having a high elongation. Increase the amount of 58). The coolant 58 is arranged at the inlet side of the rolling roll 52 and is injected from the injection pipe 59 which is divided and sprayed in the thin width direction (arrow L). Thereby, the thermal crown of the rolling roll 52 is alleviated, and the site | part of the aluminum foil 53 corresponding to the quarter part a deforms toward an extended state. If the elongation is low in the real data, reverse operation is performed. In addition, in the said target shape data, the elongation obtained from the element 4e corresponding to the site | part is often set to 0 so that the quarter part a of the rolling roller 52 may be cooled. Here, the actual shape data may be an output value from a sensor or a concept indicating an elongated / extended shape as shown in the following practical example.

As described above, in foil rolling, the final product is obtained by repeating a plurality of rollings (passes). The plan of how many thicknesses or surfaces to finish the final product is called the operation policy. This operation policy sets the target shape stretched and extended in the middle of the path at the same time as the thickness target in order to finish the final product in a stretched and elongated and uniform flat shape. In actual operation, the extension | stretching and extension | distribution distribution in a path | route along the way mentioned above differ. This is because the operating conditions called deformation by heat (heat crown) of the rolling roll as described above are different for each pass. Therefore, the above-mentioned operating policy is to define a target shape that is elongated or extended in each pass, in addition to the expected operating conditions to some extent among these actual operating conditions.

However, there are many cases in which the stretched / extended target shape in the above-described operating policy does not coincide with the actual calculated / extended target shape by the shape control unit 3, that is, the target shape data. For example, when rolling a furnace with a rolling mill of a pass to target a target shape in the operating policy such as the broken line in FIG. 25, target shape data set in the shape control unit 3 is shown by solid lines. For example, setting them together gives good results.

In this way, the target shape in the operation policy and the target shape in the actual control do not coincide even when the amount of coolant for the rolling roll 52 is divided and adjusted to correspond to the element 4e of the inspection roll 4. This is because the thermal crown moves due to the heat conduction of the roll, and the mode of movement thereof also varies depending on the type of material, the target shape, thermal equilibrium, temperature, rolling speed, and foil shape of the rolling roll. .

The influence of such operating conditions and the types of materials is often a know-how that cannot be expressed by mathematical models. Therefore, in order to be close to the operation policy, the target direction data must be adjusted in correspondence with various actual shapes appearing in the middle of each path. Therefore, in the conventional thin-rolling and other metal zinc control apparatuses, the target shape data cannot be set automatically in each pass, and only the experience of the experienced person can be relied upon, and timely and accurate control is performed. It is not always possible to produce aluminum foil according to the above operating policy. In addition, the actual shape determination made to reflect the target shape as mentioned above is entrusted to the judgment etc. which were seen from the eyes of the operator 5 conventionally, and it takes into consideration from the human error factors, such as individual difference and skill level difference. It is not suitable for ordinary shape management and shape control.

Further, in the roll presser 2, not only the magnitude of actual shape data regarding elongation from individual elements of the inspection roll 4 but also the pattern of the entire surface shape, that is, the shape state and the degree thereof, is determined. It has great meaning in controlling the processor. By the way, according to the conventional shape detection method, since the judgment of the pattern of the whole surface shape depends on the experience and feeling of the operator, skill and lack of appropriateness were required.

In the rolling mill 2 as described above, even when rolling under the same conditions, the surface shape of the aluminum foil 53 may change while having a constant tendency. For example, when both ends of the rolling roll 52 start to retain heat, as shown in FIG. 24, an expansion part (a) called a hot crown is formed at both ends of the rolling roll 52 or the degree of expansion thereof is significant. The surface shape of the aluminum foil 53 may become the tendency to extend | stretch an edge by starting below.

Such a change in surface shape is shown in Table-1.

TABLE 1

For example, when it is specified by "end drawing" of the current surface shape and the degree is level 3, the case shown in (1) of Table-1 and the case shown in (2) In some cases, the adjustment of the target shape to be coped with in the future should be different. If there is a tendency to be displayed in the same (3), it is necessary to reduce the control gain when setting the target shape data given to the shape control unit 3 for controlling the surface shape.

On the other hand, it may not be possible to express all the rolling states only by the real shape data at this time and the tendency of the shape change derived by calculation from this real shape data.

In other words, statistical characteristic information that cannot be obtained without regular data collection and using real-shape data from several past types of time, such as the above-described trend of change of the collected data, or average, variance, interrelation between data, and three-dimensional There is a lot of pattern recognition. This statistical characteristic information becomes important for adjusting the surface shape of the aluminum foil 53.

The present invention has been made to solve the problems of the prior art as described above.

Therefore, the first object of the present invention is to automatically and appropriately change and set the target shape data to be given to the shape control unit in response to changes in the operating conditions so as to stably produce a desired strip-shaped or plate-shaped metal. An object of the present invention is to provide a target shape adjusting device for metal.

In order to appropriately change and set such target shape data, it is necessary to appropriately determine the pattern of the entire surface shape in the width direction of the belt-shaped or plate-shaped metal material during traveling.

Therefore, the second object of the present invention is to appropriately judge the pattern of the entire surface shape in the width direction of the belt-shaped or plate-shaped metal material of the traveling circumference, and to the process or the operator handling the metal material. The present invention provides a shape detecting device for a band metal material or a plate metal material that can give objective judgment information.

In the conventional shape control unit 3, the target shape or the control gain is adjusted based only on the determination result of the actual shape at any one point of time, so that the target good men are controlled in some cases, and the surface as described above. It was not always possible to carry out appropriate control with a change in shape.

Therefore, the third object of the present invention is to determine the tendency of the change in the shape of the metal produced by the rolling mill even if the rolling characteristics of the roll mill are greatly changed, and to predict the change tendency, SUMMARY OF THE INVENTION An object of the present invention is to provide a metal rolling shape adjusting device capable of accurately adjusting the shape based on the latest actual shape information.

On the other hand, not all the rolling states can be expressed only by the change tendency of the shape, and the past statistical characteristic information about the shape may become important in expressing the rolling state as a qualified one.

It is therefore a fourth object of the present invention to provide a metal rolling shape adjusting apparatus which can calculate past statistical characteristic information of a metal shape and adjust the shape with high accuracy based on the statistical characteristic information. .

While this specification particularly points out the subject matter of the present invention and ends with the claims that are clearly claimed, the present invention is believed to be well understood as the following description with reference to the accompanying drawings.

The aluminum foil rolling target shape adjustment apparatus 1 which concerns on a present Example is applied to process online control as an expert system which systemized the control know-how of an operator.

Prior to the detailed description of this system, an outline of the system configuration of the aluminum foil rolling target shape adjusting device 1 will be described using FIG. 2.

As shown in the drawing, the aluminum foil rolling target shape adjusting device 1 includes a rolling data collection unit 7, a rolling situation analysis unit 8, a control target generation unit 9, an action candidate reasoning unit 11, Mainly composed of the target shape generating unit 12 and the action effect evaluation unit 10, the rolling situation analysis knowledge base (D ₁ ), the control target setting knowledge base (D ₂ ) and action reasoning knowledge containing the operational knowledge A base D ₃ (including knowledge of maintaining the consistency and action of the loop) and work memories M ₁ , M ₂ , M ₃ , M ₄ , and M ₅ for temporarily storing various data are provided. The outline of the processing contents in the respective sections will be described below.

In the aluminum foil rolling target shape adjusting device 1, inference processing is started by a key input from the terminal 6 on the aluminum foil rolling mill 2 (same as the roll mill 2). The operating condition data from the foil rolling mill 2 is input through the shape control unit 3.

① Rolling data collector

The rolling data collection unit 7 writes the operating condition data including the actual shape data from the shape control unit 3 into the received work memory M ₁ .

② Rolling situation analysis part

The rolling situation analysis part 8 analyzes the said real shape data, and determines the rolling state of the aluminum foil 53. As shown in FIG.

That is, it is determined whether the room shape data is, to some degree suitable for each of the thread-like pattern which is stored in the rolled condition analysis knowledge base (D ₁₎ is divided into several types of patterns in advance. At the same time, the current target shape data is analyzed.

③ Control target generator

The control target generation unit 9 determines the actual shape of the aluminum foil 53 based on the analysis result of the actual shape data by the rolling situation analysis unit 8 or the input from the terminal 6 by the operator 5. Set the control target to convert in the direction.

In addition, based on both the analysis result by the said rolling situation analysis part 8, and the input by the said operator 5, you may set the said control objective with reference to these.

④ Action Candidate Reasoning Department

④- (1) Ruul Reasoning

The action candidate reasoning unit 11 stores a loop stored in an action reasoning knowledge base D ₃ of IF-THEN type, which concludes an action for realizing the control goal under the condition of the control goal and operation conditions. Based on the loop reasoning applied, the action determined to be valid is written into the working memory M ₅ . At this writing, the following processing is executed.

④- (2) Resolving Contradiction and Hassle

Two or more kinds of control targets are selected, and two or more kinds of contradictory action candidates are considered to be more important than the actions of the control targets, or the selected attributes, or two of the following attributes of a bet e: The action determined based on the combination of the above attributes is applied.

a. An attribute (majority) indicating the importance of the control method measurement set in advance for each real shape classification item (Fig. 3 (b)) described later.

b. The attribute (confidence) which shows the extent to which the real shape data of the aluminum foil 53 belongs to the said real shape classification port to some extent.

c. An attribute (subject) indicating a subject that has determined whether the real data belongs to two or more of the above real item categories.

d. An attribute indicating the degree of execution of the determined action.

e. Attribute (importance) indicating the degree of necessity to change each real shape classification item.

④- (3) Learning Invalid Actions

In the loop, when there are a plurality of action candidates for a control target, the action candidates are given priority, and only the action having the highest priority is applied. Actions that have been executed to achieve a certain control goal and have no effect are not applied repeatedly even if the same control goal is set next time.

⑤ Target shape generator

The target shape generation unit 12 generates a new target shape data based on the applied action and outputs it to the shape control unit 3. The shape control part 3 controls the aluminum foil rolling mill 2 based on this target shape data.

⑥ Action effect evaluation part

The action effect evaluation unit 10 evaluates whether or not the control of the aluminum foil rolling mill 2 based on the applied action is effective by the result of the data analysis and the inquiry to the operator 5. At this time, the action evaluated as invalid is stored in the work memory M ₄ , and is referred to by the action candidate inference unit 11 during the next action candidate inference. As shown in FIGS. 1 and 2, the aluminum foil rolling target shape adjusting device 1 includes a shape control unit for controlling the injection amount or the temperature of the coolant 58 so as to adjust the actual shape of the aluminum foil 53. The target shape data, which is the standard of the control, is output to 3), and the operating condition data is input by the shape control unit 3.

As a method of adjusting the actual shape of the aluminum foil 53, the rolling force of the rolling roll 60 which presses the lower rolling roll 52 upward toward the other rolling roll 52 is controlled. There may be a method, but in the present embodiment, only the control of the coolant 58 will be described below.

① Collection of operating condition data

In the aluminum foil rolling target shape adjusting device 1, the inspection roll 4 shows the stretched portion 54 and the stretched portion 55 (Fig. 23) of the aluminum foil 53 at the time of rolling. As an aggregate of the elements 4e provided with a sensor for detecting real shape data, an aggregate of the rolling rolls 52 is provided downstream of the conveying direction (arrow display K) of the rolling rolls 52 to form the shape control unit 3. The actual shape data is outputted to the rolling data collection unit 7 (FIG. 2) through.

The rolling data collecting unit 7 writes the operating condition data (Table-2) transmitted by the shape control unit every predetermined time interval in the work memo (M ₁ ).

TABLE 2

At the same time, the rolling status analysis unit 8 is started.

② Analysis of real shape data

Rolling situation analysis knowledge base (D ₁ ) for calculating the stretched state or elongated state (each real shape classification item) and their degree of the aluminum foil 53 during rolling from the real shape data detected from the element 4e (sensor). ), As shown in FIG. 3 (b), for example, "end extension" to "round elongation" are real shape classification items and real shape data showing the state of real shape detected by the element 4e. The program stores a specific program for each item for specifying the real item category.

Here, the specific method of the real shape classification item is demonstrated. The real shape data detected as the tension distribution in the element 4e is obtained in the form of an elongation distribution in the thin width direction as shown in FIG. The real data shown in the figure is composed of an end portion, a quarter portion, a central portion A, and a central portion B from the outside, and the whole central portion is formed from the central portion A and the central portion B again. In this case, the extension part 55 is located in the central part B and the extension part 54 in the quarter stub parts on both sides.

The actual shape classification items stored in the rolling situation analysis knowledge base D ₁ are mainly classified into five types as shown in FIG.

(1) "tip height"… When the elongation at the end is lower than the end, it is considered to be the end elongation. ① It is judged by whether the value of the end is the minimum value of the whole and ② the degree of elongation difference between the end and the quarter.

(2) "end extension"… Contrary to the case of end height, the value of the end is significantly higher than other parts.

As described above, the actual shape of which the end is stretched to some extent is preferable, but when the extension is excessive, it is regarded as a problem shape.

(3) "quarter extension"… It is judged that the elongation value of the portion that is most extended near the portion corresponding to the zero point set in the target shape is somewhat larger than that of the end portion. As described above, the actual shape is most likely to be elongated in the quarter portion, and almost this type appears in the actual actual shape.

(4) "Medium height". The extension state of the center (low elongation) and the end are compared.

(5) "Intermediate stretching"…. If the difference between the elongation value of the most elongated part in the center and the most elongated part (in most cases, the end or quarter part) is small or negative, it is judged to be medium elongation.

The intermediate stretching is classified into two types: 1) general intermediate stretching in which the central portion is stretched, 2) stretching between quarter portions and the central portion, and extending and rounding the central portion.

Other unusual real shape classification items include the following asymmetry and zero point inadequacy.

Asymmetry… Usually, the target shape and the real shape are close to the left and right symmetry in the thin width direction at the same time, but the shape whose symmetry is broken is called an asymmetric shape. As the criterion of determination, the difference in elongation of the maximum value or the difference in elongation of the minimum value at both the left and right ends is large. (2) Asymmetry shall be considered when one of the ends is elongated and the other is specified as extending the end.

Of course, none of them holds true.

"Zero point inappropriateness"… It means the case where the site | part of the zero point set in the target shape and the site | part which showed the elongation maximum value in a real shape do not correspond. Usually, heat is easily accumulated in the quarter portion a of the rolling roll 52, and the actual quarter portion corresponding to the quarter portion a is most easily stretched. Therefore, when setting the target shape, 0 point is set at the most extended part of the quarter of the real shape. And when these parts are shifted, it is necessary to adjust so that they match.

From the detected real data, which of the real shape classification items corresponds to the actual shape of the aluminum foil 53, the judgment is made according to the method shown in the section "Specific methods" in FIG. do. This technique is stored as a program in the rolling situation analysis knowledge base D ₁ as described above.

As described above, one or two or more kinds of real shape classification items which are the cause of the real shape data input from the working memory M ₁ are calculated by the specific program in the rolling situation analysis section 8. Becomes

Usually, only one real shape classification item which is determined to be in a state is selected. Since the real shape data is obtained as a result of complicated intertwined operating conditions, it is often judged to be in the state of two or more types of real shape classification items. In this case, there are a strong real category category and a weak real category category in the causal relationship with real data. This strength and weakness of the causal relationship, that is, the degree of the state of shape, is called confidence (property).

The rolling situation analysis section 8 clamps the real shape classification item and its confidence level into one or two or more real shape classification items in any degree of confidence in deriving the real shape data as an appropriate function. ₂ ) to remember. For example, at the end where the aluminum foil 53 is placed, the elongation difference (α ₂ ) between the site and the tip with the highest elongation in the range of the real shape data input from the four elements 4e and the whole real data When the ratio (β ₂ / α ₂ ) between the difference between the maximum value and the minimum value of the elongation (β ₂ ) exceeds a predetermined set value, it is explained that the real shape at this time includes the real shape classification item "end stretching". The confidence level from 0 to 1 is added according to the value of the ratio. The same applies to the other real shape categories.

In this case, on the basis of the actual shape data in the past predetermined time in the work memory M ₁ , statistical characteristic information of the rolling mill 2, for example, average change trend, variance, correlation, and three-dimensional pattern recognition, are calculated. Based on the statistical characteristic information, that is, this statistical characteristic information may be used as a variable of the confidence level, and the confidence degree for each real shape classification item may be calculated.

For example, it can be seen that at some point in time, the actual shape is extremely different from the actual shape at the time before and after. Specifically, the state of "end extension" is detected only for a moment when the actual shape of "end extension" continues because of abnormality of the raw plate shape before rolling, and the state of the original "end extension" is continued after that. It is the case.

Therefore, by applying an average value from the statistical characteristic information of the real shape data to the past point in time up to the point in time, it is possible to determine the rolling state of the trend without being influenced by the noise element described above.

Next, the case where the change trend of the actual shape data within the past predetermined time is employed as the statistical characteristic information will be described in detail below. The calculation of this tendency is performed according to the processing procedure of the flowchart shown in FIG. 4 (a), for example. In the rolling operation of the aluminum foil 53, the actual shape data of the aluminum foil 53 obtained by the inspection roll 4 on the side of the roll mill 2 is collected by the rolling data collection unit 7 at predetermined time intervals. (S40).

Next, in the rolling status analysis section 8, the real data is compared with the real shape classification item, and is specified as any real shape classification item, and the degree is represented by the number of levels shown by natural numbers 0 to 5. . That is, the rolling state of the aluminum foil 53 is judged (S41).

For example, the elongation difference (α ₂ ) between the site and the tip with the highest elongation in the range of real shape data input from the four elements 4e sequentially from the end where the aluminum foil 53 is pressed, and the real shape data In the case where the ratio (β ₂ / α ₂ ) between the difference between the maximum value and the minimum value (β ₂ ) of the elongation in total exceeds a predetermined set value, the actual shape at this time includes the actual shape classification item "end stretching". Judgment is made, and the grade of "end drawing" is represented by the number of levels represented by the natural numbers 0-5 according to the value of the said ratio.

Furthermore, at step S42, the shape score H _t from -5 to +5 points is determined with respect to the actual shape data at time T, and the figure which stores the shape score H _t within a predetermined time is shown. It is stored in the memory that is not used.

This embodies an empirical handling method for the operator 5 to specify real shape data to any real shape classification item, and for example, to be easily understood by the operator 5 when displayed on the terminal 6. The confidence of the real classification category to which the data is suitable, and the confidence level of the real classification category in which the real classification category and the elongation and elongation state are contradicted are converted into two coordinates from the natural number of the government centered on zero.

For example, if the actual shape classification item is "end drawing" and the degree is level 5, the shape score is set to +5 point, and the level 1 is set to +1 point. In addition, when it is level 5 in "extension extension", it is set as -5 and when it is level 1, it is set as -1 point. If it does not correspond to "end extension" and "end extension", it is set as 0 points.

And if it is necessary to change and adjust a target shape by the input from the terminal 6 of the roll mill 2 side with respect to the request of the operator 5, or by the input from the rolling situation analysis part 8 (S43). In step S44, the change trend of the actual shape within a predetermined time up to the present time is calculated.

For example, the shape change tendency at time T is calculated from the past one shape score H _{t-1 + 1} ,..., H _t . Here, an example calculated from the past 10 shape scores H _t-9 ,..., H _t is shown. First, from the maximum and minimum values among the shape scores H _t-9 ,..., H _t , the shape score aberrations H _d shown in the following equation are obtained.

H _d = max (H _t-9 ,…, H _t )-min (H _t-9 ,…, H _t )

At this time, when this shape aberration H _d is 2 or less, it is judged that the actual shape of the aluminum foil 53 is a stable state, and the tendency of shape change is not recognized.

In the case where H _d is 3 or more, it is first determined whether the real shape is periodically changed.

As if in, and shown in the food _- for example a shaped score the number of times a change in the direction to increase the contour points between the (H _t-9, ..., H _t) as H _+, the number of changes to the decreasing direction to H together

3, 또한│H _{+ -} H - │

3, also

H_d

43

H _d

4

In other words, when the absolute value of the difference between H ₊ and H ₋ is 3 or less and the shape score difference H _d is 3 or more and 4 or less, it is determined that the real shape is in the tendency of periodic change.

And when _Hd is two or more and it is other than common sense, it is judged whether there exists any tendency as a shape conversion tendency is recognized. For example, the shape score H _t and the present development score H _t-9 of the past 10 items are compared at this time.

That is, as shown in Table 3, H _t is calculated when the tip is larger than H _t-9 and shows the elongation tendency. On the contrary, when H _t is smaller, the tip is elongated and there is an elongation tendency.

In addition, if the same H H _t-t and ₉ there is a determination that is compartmentalized phase stability, is treated by bypassing the number of levels in the chamber-like classifiers, which will be described later correction step (S45).

So, for example, when the shape change tendency is "end stretch tendency" at time T, the actual shape classification item at that time is specified as "end stretch", and when the level days is 3, FIG. 4 (b) As shown in Fig. 2, a corrective action is performed in which the number of levels is increased by two with respect to the number of levels of the actual shape classification item at that time (S45). This is for future shape control, considering the shape determination and shape change tendency at the present time, the level 3 is inadequate in the above-described E, so in reality, the degree of "end drawing" is approximately level (5). This is because it seems to be comparable to) (a state where the end stretching is rapidly progressing).

Therefore, in step S47 in FIG. 4A, the target shape data is changed and adjusted so as to obtain a desired actual shape based on the latest real shape classification item specified as described above and the number of levels after correction. This is output to the shape control part 3 (S48).

On the other hand, in the stem S44, when it is determined that the actual shape is in the cyclic change trend, the control gain at the time of changing and adjusting the target shape data is decreased by 30% (S46). Thereby, the tendency of the periodical change of a real shape is suppressed.

On the other hand, if dispersion is applied as the statistical characteristic information, the rolling state can be judged from the degree of dispersion of the real-shape data regarding the shape of the tip of the aluminum foil 53 due to a time change, for example. In other words, when the degree of dispersion is large, it indicates that the rolling situation is unstable. In this case, therefore, the control action indicated in (ii) (ii) to stabilize the rolling state is executed by the shape control section 3.

At this time

Iii) Change the control gain of coolant quantity to low.

Ii) Since the rolling rolls 52 are not sufficiently heated, the amount of coolant is lowered as a whole, or the elongation of the target shape is raised as a whole.

As an example in which the characteristic information is a correlation, a correlation between the level of quarter stretching and the level of end extension is shown in FIG. 5 (a), where the relationship between the two can be said to be in positive correlation. It shows that the elongation and tip elongation are in a prone state at the same time.

In this case, it is gained from the empirical knowledge that the shape adjustment method of reducing the amount of coolant at the end is not effective in order to eliminate the end extension.

On the other hand, when the relationship between the two is uncorrelated as shown in Fig. 5 (b), since the knowledge of the shape adjustment method found in the case of the previous correlation is obtained, the actual shape is determined by the shape adjustment method. Image adjustment is performed.

In addition, the statistical characteristic information is obtained by recognizing a three-dimensional pattern regarding real shape data.

The three-dimensional pattern (P) is shown in FIG. In the figure, the arrow M indicated by the dashed-dotted line shows the temporal positional change of the element 4e in which the elongation detected the maximum value at each time t ₀ to t ₆ . As shown in the drawing, the element 4e having the largest elongation is meandering in the width of several elements 4e in the thin width direction.

In such a state, the coolant 58 should be sprayed equally to the rolling rolls 52 corresponding to the width of the several elements 4e including the site as well as the site that is currently drawn to the maximum. . This means that even when the coolant 58 is sprayed only on the portion that is currently stretched to the maximum, the stretched portion moves only to the adjacent portion.

In this three-dimensional pattern (P), it can be determined not only by the statistical characteristic information described above, but also by the above-described numerical calculation algorithm. However, the recognition method by the neural network which performs pattern recognition of the real shape data of the aluminum foil 53 is effective.

As shown in Fig. 7, the neural network 20 is conceptually provided as a plurality of neurons 15 for calculating and outputting input data by the threshold inverse processing. Arranged, each interlayer is connected via a connection 16.

The neural network 20 uses the pattern data of the actual shape data of the aluminum foil 53 as the input data and the actual classification items and the degree thereof as the output data. Learning is done by changing the connection overlap.

Therefore, when new real shape data is inputted into the neural network 20 of the said learning agent, the said real shape data becomes specific at the same time to the said real shape classification item of what said said. When new real shape data is input to such a neural network 20, this real shape data is specified to the real shape classification item of any of the said at the same time as its degree. Such a neural network 20 may be applied to the rolling situation analysis section 8.

Here, as an operation method for judging the real shape state of the aluminum foil 53, an example of numerical calculation by a specific method determined for each real shape classification item is mainly shown, but the judgment by the neural network 20, or Needless to say, the same computational effect can be obtained by the judgment based on the Lulu base (extra) which has such judgment knowledge.

In addition, the real data which is input to the rolling situation analysis part 8 in the said work memory M1 and is given to calculation is a value of one kind of data at one time point in the said predetermined time, or several kinds of data at one time point. Regardless of the value of the data, the value of one type of data at any time, or the value of several types of data at any time, the data necessary for the calculation may be appropriately used.

In this way the two rolling situation analysis unit (8), the yarn-shaped classifier for the current thread shape data of the aluminum foil 53, and its degree of confidence determined, is written into the working memory (M _2).

③ Generation of control target

The control target data (shape change target (figure 8) and its importance attribute), which are the keys for setting or changing the target shape appropriately, are controlled by the operator 5 in the control target generation unit 9. It is inputted from the terminal 6 on the side of the rolling mill 2, or automatically generated based on the rolling situation data such as the actual shape classification item in the working memory M ₂ and its confidence level.

In this way, the determination of which real shape classification item the real shape data reflecting the current control state of the roll mill 2 belongs to, in other words, the shape change target derived from here and the judgment of its importance are inputted from the terminal. It is achieved by giving importance to any one of using the control target generation unit 9 which performs automatic generation as the subject (attribute) by using the operator as a subject (attribute).

In this automatic generation, the "operation policy"("to extend the roll of the tip in a small path") or "automatically generated" is inconsistent with that generated by the operator (5). The ruling said "prioritize input information" or the ruling "priority of certainty has priority" is applied with reference to the control target setting knowledge base (D ₂ ).

Therefore, even when the automatic generation due to the operator's input is inconsistent, contention between the rules having their shape change targets can be avoided, and the contradictions between the ashers accompanying the same can be eliminated.

For example, if the detected real shape data includes "end stretch" as in the above example, and the confidence degree at that time is 0.8, "end stretched" among the five shape change targets to solve the "end stretch". Selection is made, and the importance corresponding to the confidence level 0.8 falls short of the selected shape change target. The shape change target and its importance are stored in the work memory M ₃ .

The situation in which the importance of the shape change target is calculated in the control target generation section 9 from the confidence level of the actual shape classification item specified as described above is described in detail below.

The importance is a goal of whether or not the target shape data needs to be changed, and is displayed on a graph (FIG. 9) showing the relationship between the degree of confidence of the real shape classification item indicating the degree of elongation of the actual shape data. In the figure, an example of "end extension" is shown among the actual shape categories.

An importance calculation function (f (x)) that defines the degree (importance) of the need to change the target shape is defined for each of the real shape categories, such as tip extension and tip extension. For example, the importance calculation function f ₁ () for the end height

It is defined as

That is the end elongation degree (certainty) that the first limit value (L ₁₎ and are compared, FIG. 9 a first threshold (L ₁₎ is exceeded, its thread shape classification item targeting the shape as shown in Selection is made for modification, and the importance corresponding to the confidence shown in the figure is calculated.

On the other hand, if the confidence level is equal to or less than the first limit value L ₁ , zero importance is given, and its real shape classification item is not contributed to the change of the target shape. Whether the first limit value L ₁ is set individually for each real shape classification item, and each confidence level and the first limit value L ₁ are compared and donated to change the target shape. Judgment is made for every item.

The necessity of changing the target shape is determined for each real shape classification item as described above, and in some cases, it may be determined whether the average of the combination of the importance of each item exceeds the limit. Subsequently, a method of judging on the basis of the combined importance or the average of the necessity of changing the target shape from the selected real shape classification items for changing the target shape will be described in detail.

The real shape classification items are S ₁ , S ₂ ,... If S _i , the synthesis importance calculation function g () obtained by synthesizing the importance calculation function f _si () for each item defined for them is defined as follows.

g (f _si (X ₁ ),… f _si (X ₁ ))

The function g () is

g ≡ (∑f _si (X ₁ )) / i

Or g ≡∑f _si (X ₁ )

f_si(X₁)

1,only,

f _si (X ₁ )

One,

f _si (X ₁ ) is expressed in the form of a sum mean or a sum, such as increasing forging for X ₁ .

Here, with respect to (L ₁ ) in f (L ₁ ), the increase in forging means that f (L ₁ ) <f (L ₂ ) when L ₁ <L ₂ . In this way, the synthetic importance indicating the degree of necessity for starting the inference processing is calculated in the importance of each item in the action candidate inference unit 11, and the synthetic importance is determined by the predetermined second limit value L ₃ (not shown). When the time exceeds the inference process, the target shape data is appropriately changed.

Illustrating the specific example below,

0.4 importance for tip height

The importance level for `` quarter extension '' is 0.6

In addition, "end extension", "intermediate extension",

Medium importance for 「Medium Height」

If the composite importance is

0.4 +0.6 + 0 = 1.0

Becomes If the second limit L _{3 at} this time is 0.9, the larger one of the synthesis importance is inferred.

As a method for determining the importance level that is used as the trigger, one based on the above-mentioned sum total may be employed.

In the present embodiment, as described above, even when the degree of confidence for each item exceeds the limit given to each item, for example, even when the degree of "end extension" exceeds the limit value L ₂ shown in FIG. Inference processing is moved.

When the synthesity importance exceeds the second limit L ₃ , the alarm signal may be output to other alarm devices at the same time to drive the alarm device.

Next, the inference processing of the target shape will be described.

④ Action

④- (1) Ruul Reasoning

The operating condition data including the current target shape data and the current real shape data, processed as described above, the extracted rolling shape data including the extracted real shape classification item and its confidence level, and the shape change target and its importance. The control target data including? Is transmitted from the work memories M ₁ , M ₂ , and M ₃ to the action candidate inference unit 11, respectively. The action candidate reasoning 11 compares each transmitted data with the conditional condition of the loop stored in the action reasoning knowledge base (D ₃ ). Select target shape change data (Fig. 8, hereinafter referred to as action).

As mentioned above, the reasoning process for drawing a conclusion (target shape to be employed) corresponding to the conditions such as operating condition data, rolling situation data, and control target data cannot be relied on by the knowledge (know-how) of the experienced person.

Such reasoning logic is automated in the present invention. For this automatic reasoning, one loop is stored and stored in the action reasoning knowledge base D ₃ .

This reasoning is expressed in the form of "concluding section" if [conditional], and in logical form as shown below.

If [Control Target Data] or [Operation Condition Data, Rolling Condition Data], [Specification of Target Shape Adjustment Parameter and Its Change Degree (Action and Its Accuracy)]

Here, the target shape adjustment parameter is an element that determines target shape data (a target elongation distribution in its path) as shown in Table-5.

As the target shape adjustment parameters constituting the conclusion of each loop, not all parameters shown in Table-4 should be described. In most cases only some of the target shape adjustment parameters necessary to satisfy the conditional part are described with the degree of change.

For example, as shown in FIG. 10, when the actual shape of the aluminum foil 53 is specified by the quarter stretch in the example 1 of the ruwl, and there is no zero below the largest part of the stretch near the quarter section at that time. A loop that designates an action that "stretches near the quarter portion at the zero position is brought under the largest portion" is stored in the action reasoning knowledge base D ₃ .

④- (2) Resolving Contradiction and Hassle

On the other hand, as shown in the example 4 of the example shown in Table-4, there are cases where the tick condition is added to the ruins. For example, when the tip height and the quarter elongation occur at the same time in the real shape, the loop example (2) and the example (3) may be selected. These are

TABLE 4

The action candidate inference unit 11 cannot achieve their contradictions at the same time, respectively, and an error occurs. For example, it is possible to solve these problems by installing additional conditions in the condition of the example of the example of the example of the example of the example of the example of the example of the example of the example of the example of the example of the example of the example of the example of the example of the example of the example of the example of the example of the example of the example. If it is greater than 1 limit but the importance of extending the quota is less than 0.4, the target value of the end level (elongation at the end of the width direction) is raised.

As described above, when real data is specified in a plurality of real shape classification items and there is a contradiction between the actions derived from each real shape classification item, as a means of eliminating these contradictions and avoiding contention between them with the actions. There is a use of a severity (property) that indicates the importance of a control rule set in advance for each of the actual shape classification items. The aluminum foil 53 is classified by thickness, thickness, material, post-process, and delivery destination. In these aluminum foils 53, there is a thing that the actual classification item thereof should never be, for example, "intermediate stretching". In such a case, the actual shape classification item of "intermediate drawing" is given a very large degree of importance in comparison with other real shape classification items, and when an inconsistency occurs between the actions, an action for improving the intermediate drawing is preferentially given. It is supposed to choose.

Furthermore, the contradiction between the above actions can be resolved in view of which real shape classification item becomes a bigger problem when two or more kinds of real shape classification items as described above are specified in the real shape data.

In other words, only the action derived from the real-type classification item having high confidence level may be adopted based on the degree of confidence (attribute) indicating how the actual shape data belongs to each real-type classification item.

It is also possible to resolve the contradiction between the above actions based on the combination of the respective attributes. For example, only actions derived from any real shape classification item are employed by comparing these values based on the sum or multiply of the magnitude and the confidence level for each real shape classification item.

As a method of solving such contradictions, the shape change target of the one with the higher degree of importance can be given priority.

On the other hand, two or more types of shape change targets may be simultaneously selected for a certain real shape data, and the contents of the actions corresponding to the respective shape change targets may be the same. For example, the real shape classification item (FIG. 3 (b)) of the current real shape data is simultaneously specified in the "extension" and "intermediate extension" sections, and the degree of the action as the action derived from each real shape classification item is obtained. This is the case when "raise the end level" (Fig. 8) with the same contents but the same contents is selected at the same time.

It is stored in the action reasoning knowledge base D3 to set the degree of the action in advance in accordance with the degree of confidence of the actual shape classification item specified as the loop applied in such a case.

Therefore, when actions of the same content having different levels as described above are selected simultaneously, these actions are not executed at the same time, and only the ones having the large degree of properties (attributes) are executed, thereby avoiding the trouble of executing the actions. . In contrast, only a small degree of the action may be executed, or an action corresponding to the average value of the degree of each action may be selected or generated.

In addition, when the state of execution of the previous aluminum foil 53 is slightly improved, that is, when the confidence level of the actual shape classification item to be improved has decreased, for example, from the previous 1.0 to 0.8 of the present time, the action performed at this time. Is considered valid. However, if the degree of the action, that is, the value of the target shape adjustment parameter (Table-5) is larger, for example, more preferable results are obtained. Therefore, in order to obtain the desired result, for example, the degree (attribute) of the action is selected, the algorithm for determining the value of the degree is changed, and the degree of the action is drastically when the same actual state appears after the next time. Is changed to

That is, the attribute value is changed based on the evaluation result of the action by the action effect evaluation unit 10.

④- (3) learning of expressionless action

As shown in Fig. 8, several types of action candidates with priority added to the shape change target of 1 are prepared. The highest priority action is realized when any shape change target is selected.

The priority is not fixed but checked per inference. For example, when inference is performed in the action candidate inference unit 11, which action is adopted for which shape change target is employed is stored in the work memory M ₅ , and the action effect evaluation unit 10 at the next inference. It is judged by comparing the importance of this time last time whether or not the last shape change goal was achieved (evaluation of effect).

As a result, if it is determined that the current target shape data changed based on the last shape change target and its importance is valid in the action effect evaluation unit 10, that is, if the importance is lower than the previous time, the adopted action was valid. remember that the priority action on reasoning knowledge base D ₃ is advanced. On the contrary, when it is judged to be invalid, the loop which deduced the action that was applied last time and the selection of its action is stored in the working memory M ₄ .

For example, the change target shown in FIG. 8 is determined by the rolling status data from the rolling status analysis section 8 or the input data from the operator 5 to determine the last time "to extend the edge" at importance 0.6. If so, among the actions accompanying it, the highest priority "raise the end level (priority 1)" is selected, and the target shape data which raises the end level (elongation) according to the importance 0.6 is the shape control unit 3. The target shape data, the shape change target, its importance (0.6), the action and its priority (1), which are outputted to and simultaneously applied, are stored in the action reasoning knowledge base D ₃ . And when the shape change target "to extend the edge" is selected in the prior importance 0.6 or more, the actual state of the end shape in question is often not improved. The action was invalid. On the contrary, if "I want to extend the end" is selected with less than 0.6 importance, the previous action is considered valid.

Therefore ruul deduced the invalid action and its actions are stored in the working memory M _4. And when the same shape change target is selected at the next inference and the invalid action is selected, the no-action action is not applied and it is determined that it is appropriate, and then a high priority action is applied to provide the proper change of the target shape data of this time. Done. As a result, if it is determined that the next action applied to achieve the shape change target is valid, the priority of the action is advanced, and the priority of the invalid action is sequentially lowered. The reasoning priority in the reasoning knowledge base D ₃ is stored.

In this way, even if the target shape obtained by the aluminum foil rolling target shape promoting apparatus 1 does not show an effect on the actual shape and continues the actual shape of the problematic aluminum foil 53, In this case, a different one is selected from the previous action, even if the previous reasoning and the shape change target were the same. As a result, invalid loops are not repeatedly applied, and the actual shape is appropriately changed.

However, as a result of applying the average action suitable for the degree of a plurality of actions having the same contents as described above and having different degrees of execution, in order to solve the above-mentioned inconvenience, there is a case where the action thereof becomes invalid. In such a case, the application of the action before averaging may have become effective, so the ruling that does not perform with a significant drop in priority for the invalid action as described above or the ruling that the original action is applied is stored in the base (D _3). Thereby, matching between the function of eliminating the inconvenience and the function of learning the invalid action is performed.

In the action candidate inference unit 11, it is registered in the working memory M ₅ every time if it is determined that the action to be taken as a candidate is valid in detail according to the check tree shown in FIG. In addition, the action reasoning knowledge base D ₃ includes a main area for storing the loop set and a loop set set for each of the shape change targets by a plurality of loops conditionally having the same shape change target as described above. A retreat area for storing the loops restrained from the set of loops is secured. In addition, after the reasoning process is completed, the ruins are sequentially removed from the main area to the evacuation area.

In the check tree displayed in the drawing, first, whether a matching loop of the shape change target and the shape change target selected at this time exists in the set of belonging loops of the loop, that is, the loop corresponding to the loop set remains in the main region. Checked in (C ₁ ). If it is a case (C ₁ ), proceed to the check of the supplementary conditions in the case (C ₃ ), such as the operating conditions in the conditional part of its ruling.

If the case C ₃ does not correspond, i.e., if all of the predicates, including the ancillary condition, do not match, then its loop is removed from the set of loops of the main region to the evacuation region. If the incident condition is satisfied, whether the effect of the ruul in which the conditional part is satisfied in the case C ₄ is applied in the past has been recognized, that is, whether it is registered in the current working memory M ₄ . Is checked. And if the case (C ₄ ) it is removed from the loop set (main area). If it does not correspond to the case C ₄ , the existence of the contradictory actions is checked in the case C ₅ . As described above, the shape change target is not limited to one, but a plurality of formation change targets may be selected. In this case, contradiction between the actions accompanying each shape change target was king.

Therefore, if the result of the check is valid and there is an action that contradicts the currently checked action in the action already registered in the work memory M ₅ , the importance of the currently selected shape change target in the loop with one action If (property) is large, the role of the loop is registered in the work memory M ₅ , and the contradictory action already registered is removed to the evacuation area.

If the result of the check is valid and there is an action that contradicts the currently checked action in the action already registered in the work memory M ₅ , the importance of the currently selected shape change target in the loop having one action ( Property), the action of its loop is registered in the work memory M ₅ , and the contradictory action already registered is removed to the evacuation area.

On the contrary, if the importance of the shape target with the already registered action is large, the registered action is left in the work memory M ₅ so that the ruwl with the currently checked action is removed from the set of the ruins (main area). Is removed. As a check for contradiction, the importance is checked as an attribute as described above, but it is naturally possible to check based on the above-mentioned other attributes (materiality, diffusivity, judgment subject, degree of action) or a combination thereof. In addition, troublesomeness is eliminated by the same technique. If it does not correspond to case C ₅ , that is, if the respective actions do not contradict, it is checked in case C ₆ whether the shape change targets of these actions are the same. If case C ₆ is applicable, it is checked whether there is a difference in the priority of each action (case C ₇ ), and if there is a difference in priority, only the convenience action having a higher priority is the working memory (M _5). The one-left action with a lower priority is deleted from the work memory M ₅ . If there is no difference in priority, both actions are left together in the work memory M ₅ .

As described above, if it does not correspond to the case C ₁ , it is determined that there are no rules matching the shape change target selected this time, and the case C ₂ is checked.

If the case (C ₂ ) corresponds to the case, that is, all of the ruins in the set of ruins are checked, but if the action candidates are judged to be inadequate according to the checks of the cases (C ₃ to C ₇ ), the work memory (M ₄ ) of the past While resetting (creating) invalid information and selecting the appropriate action candidate under the current priority, the priority is reset, and each set of loops consisting of a loop with reset action candidates of this priority is recovered. After doing so, the inference of the action check is once again performed from the beginning.

On the other hand, if it does not correspond to case (C ₂ ), as a result of verifying all the loops in each loop set, the action candidate determined to be valid will be selected and stored in the working memory (M ₅ ). To exit.

In this way, after the validity is checked one by one, the ruins in the set of ruins having a common shape change target are sequentially removed from the set (main area) and stored in the evacuation area. The ruling check is executed until the ruling set of the main region related to the shape change target is empty. That is, the case that the state of the loop set is empty does not correspond to the case C ₁ .

If two or more shape change targets are selected, the action is executed as described above (C ₁ to C ₇ ) for the shape change target sets having different validity checks.

In this way, the check routine checks the contradiction, priority and validity between the action candidates already registered in the working memory M ₅ and the action candidates to be newly registered and applies them to the change of the target shape data. The validity of the action being attempted and the consistency of the loop are maintained.

⑤ Create target shape

Subsequently, the target shape change data of the action registered in the work memory M ₅ and the degree thereof, that is, "raise the end level. The degree thereof is 0.8" are transmitted to the target shape generating unit 12.

The target shape generation unit 12 changes the values of the target shape adjustment parameters shown in Table-5 and FIGS. 12A to 12D based on the target shape change data.

TABLE 5

For example, when the transfer action "raises the end level. Its degree is 0.8", the value of the target shape adjustment parameter a ₁ regarding the elongation rate of the end is changed and set according to the degree of the action. The target shape data changes. In addition, the shape control part 3 controls the coolant 58 of the roll mill 2 based on the input target shape data after the change.

⑥ Evaluation of applied action

The operation condition data including the newly obtained actual shape data is input to the rolling data collection unit 7 and the same processing as the previous time is repeated. In other words, the shape change target for this actual shape and its importance are calculated in the control target generation section 9 and compared with the previous one in the action candidate inference section 11, respectively.

An embodiment of repeating the above reasoning is shown in FIG. For example, when the shape change target E ₁ calculated from the actual shape data included in a certain operating condition data is "to extend the quarter" and its importance is 0.8, the first reasoning (E ₂ ) is executed. When the action candidate at that time (A1) widened the zero width (A2), the zero point was moved outward. The first and also applies a high action (A1), the target shape data before and after the change by that was like a bar that is displayed in the speculative result (E _3).

However, when evaluating the effect of action (A ₁ ), the importance of E ₁ is not less than 0.8, and its effect was not recognized. Therefore, the loop from which the action A1 was derived was also invalid and stored in the work memory M ₄ .

Subsequently, the second reasoning E ₄ is performed by applying the high priority action A2. In this case, if the ruwl including the shape change target E ₁ is continuously applied and the rolling process is continued, and the operating conditions do not change as compared with the first inference, the action A1 and action A2 are obtained as candidates. Lose. As a result of verifying the history of the action effect by the second inference (E ₄ ), the action (A1) is stored in the work memory (M ₄ ) because the action (A1) was invalid last time. Case C4 in FIG. ₄ is followed by action A2 of high priority to change the target shape data.

If it is determined that the quarter extension is improved as a result, the priority of action A2 is upgraded to that of action A1 and stored in the action reasoning knowledge base D ₃ .

In this way, the inference that the change in the actual shape when the target shape data is given to the actual shape of the aluminum foil 53 and the change (action) of the corresponding target shape data are performed by the action candidate inference unit 11 are performed. When changing the target shape data, the actual shape data obtained as a result thereof, the operating condition data used to set the target shape data, the rolling situation data, or the control target data, each experience value is an inference rule. It is stored in (11). Then, the action candidate inference unit 11 calculates the appropriate target shape data based on the respective experience values so as to realize an ideal realized image (operation policy) at that time point, and the target shape data generation unit calculates the target shape data. It automatically creates via (12) and outputs it to the shape control part 3. As described above, the aluminum foil rolling target shape adjusting device 1 is activated by the key from the rolling mill side terminal 6 by the operator 5 (step 21).

Subsequently, the operator 5 inputs the shape change target information with importance according to the input menu (FIG. 15) displayed on the screen of the terminal 6 (step 22).

Accordingly, the aluminum foil rolling target shape adjusting device 1 analyzes the rolling data transmitted from the shape control section 3 (step 23), prepares a suitable target shape by inference (step 24), and then targets before and after correction. The shape (Fig. 16) is displayed on the screen, and the corrected target shape data is output to the roll mill 2 through the shape control unit 3 (step 25). Then, it is evaluated whether or not the corrected target shape is valid for the actual shape in question (step 26), and enters the input standby state by the operator 5.

At this time, the shape change target is also automatically generated in the target shape adjusting device 1, but if the operator 5 is inconsistent with the input, the importance is high, the operator 5 is input, or The shape change target may be generated on the basis of a loop that gives priority to any of the shape determination results from the actual shape classification item. The aluminum foil rolling target shape adjusting device 1 always monitors the actual shape of the aluminum foil 53 and inferred by the operator 5 when it is determined that the target shape data needs to be changed. The change processing of the target shape data can also be started. For example, set the condition of the comparison with the limit a in the condition of the real classification. Examples of its ruins are given below.

If [actual feature and its confidence level] and [operation conditions], [specify the shape change target and its degree] more specifically, if the end extension confidence level <a, and the pass is 2 passes Second, the importance of trying to extend the end is expressed as 1.0. In other words, when the actual shape changes and the end-stretching degree is below the limit a, the ruling is applied and the reasoning starts.

Further, the aluminum foil rolling target shape adjusting device 1 has a function to improve the obtained real shape data even when it is asymmetric in the thin width direction.

As shown in Figs. 17 and 18, when the actual shape data is judged to be asymmetrical (step 31), the target shape which has been symmetrical to the past with respect to the region with low elongation is temporarily set by the operator 5 temporarily. (Step 32).

This is to deflect the injection quantity distribution of the coolant 58 injected into the rolling rolls 52 in the thin width direction temporarily and to uniformize the lead distribution in the rolling rolls 52, and to start or restart the operation (specially during roll rising). Valid after a roll change. In step 32, the processing is repeated until the actual data becomes symmetrical.

As described above, according to the present embodiment apparatus, even if the operating conditions of the roll mill 2 are delicately changed, the target shape data for controlling the roll mill 2 is automatically and appropriately set according to the change. You can change it.

This makes it possible to stably produce the aluminum foil 53 in any target shape specified in the operation policy without depending on operation experience or skill.

Further, the rolling speed can be increased by the shape of the control phase based on the shape stabilization of the aluminum foil 53 as described above.

Moreover, although the aluminum foil rolling target shape adjustment apparatus 1 employ | adopted the element 4e in which the piezoelectric element was embedded as a sensor which detects the tension | stretching real shape data of elongation and extension | stretching at the time of rolling, the element 4e and external appearance It is also possible to substitute a plurality of air bearing elements having approximately one as the sensor as the sensor, and to detect the actual shape data based on the change in the air pressure thereof.

Although aluminum foil 53 was used as a control object in this embodiment, it is not limited to this, Copper and other metals may be sufficient, and the thickness is irrespective.

That is, FIG. 19 shows a continuous annealing facility 70 such as a steel mill, which is another example of a process to which a target shape adjusting device for adjusting the shape of a belt or plate metal while driving is applied. In this continuous annealing facility 70, a strip-shaped coil member 74 runs between the rolls 73 disposed in the heating furnace 71, and is annealed to a predetermined temperature to adjust its quality and to move in the direction of the arrow. Are taken out.

If the coil member 74 in the heating furnace 71 is not flat, the meandering phenomenon may occur or wrinkles may occur in the stretched portion.

Therefore, the continuous annealing facility 70 detects the shape of the coil member 74 while traveling, e.g., the elongation distribution in the width direction, and adjusts the traveling conditions such as the speed and tension of the coil 74 based on the detected value. It is supposed to be adjusted.

Also in the continuous annealing facility 70, the elongation distribution (surface shape) of the coil member 74 in the extended width direction is expressed as shown in Figs. 20 (a) and 20 (b), for example. In these cases, the surface-shaped coil member 74 shown in FIG. 20 (a) has a larger maximum elongation than that shown in FIG. 20 (b), but has a change in the sliding width direction as a whole. Whereas the wrinkles are unlikely to occur in the presence, the ones shown in FIG. 20 (b) are likely to cause wrinkles even though the maximum stretched wrinkles are small. Moreover, the coil member 74 of the surface shape shown in FIG. 21 (a) is the state in which the center part of the width direction is extending | stretched, and what is shown in FIG. 21 (b) is the state which the edge part is extending | stretching. At this time, the tension of the coil material 74 in the so-called intermediately stretched state in the center portion is more likely to be wrinkled than that in the end stretched state. Therefore, it goes without saying that the application of the target shape adjusting device as described above can achieve an effect equivalent to that achieved in the rolling of the aluminum foil.

The invention can be embodied or embodied in other embodiments without departing from its spirit or its essential features.

Therefore, the above-described embodiment is a preferred embodiment, but is not limited to the above embodiment. In addition, it means that all the changes which occur within the meaning of the scope of invention and the scope of a claim to which a claim is appended are included in a claim.

Claims (6)

In the target shape adjusting device which gives target shape data to the shape control part which controls predetermined shape in the width direction of the strip | belt-shaped or plate-shaped metal material while running, and adjusts the said shape, The shape of a metal material is displayed. A sensor for detecting real shape data, and empirical value storage means for empirically changing the target shape data for the real shape and storing the change of the real shape when the real shape data is given to the shape control unit together with the change of the target shape data; And a first calculation means for creating target shape data based on the actual shape data from the sensor and the experience value from the experience value storage means, and outputting the target shape data to the shape control part. . A target shape adjusting device which applies a target shape data to a shape control unit which controls a shape in the width direction of a belt-shaped or plate-shaped metal material of a traveling circumference and adjusts the shape, and the actual shape of the metal material. A sensor for detecting data, data collection means for collecting and storing the actual shape data, situation analysis means for judging and storing the current processing situation based on the actual shape data from the data collection means, and this situation analysis On the basis of the processing status determination by means and the processing status determination input from the outside, the processing status determination following this situation interpretation means, or the processing status determination input from the outside, the target shape data for the target shape means and the experience shape Is changed to the shape control section to change the actual shape when Change knowledge storage means for storing as knowledge of pairs and pairs, change knowledge evaluation means for evaluating and storing validity of past used target shape change knowledge, and evaluation results of the target shape change knowledge and reference results of the control target Knowledge selection means for selecting, by reasoning, the best and the best target shape change knowledge in the change knowledge storage means, and the target of creating the current target shape data based on the best and the best target shape change knowledge and outputting it to the shape control unit. An apparatus for adjusting a target shape of a metal material, comprising a shape generating means. A shape detecting apparatus for detecting a shape of a band or plate in the width direction while running, the shape detecting device comprising: a sensor for detecting real shape data representing a shape while the metal material is running, and the metal material being driven from the actual shape data from the sensor And a second calculating means for calculating the actual shape classification item indicating the shape state of the shape and the degree thereof. In the metal material rolling shape detection apparatus which detects a shape in the width direction of the strip | belt-shaped metal material extended by the roll rolling mill, The sensor which detects real shape data which shows a shape in the running after rolling of the said metal material, The said sensor And a third calculation means for calculating a real shape classification item indicating the shape state of the metal material after rolling in the real shape data from the same and the degree thereof. In the metal material rolling shape adjusting apparatus which gives a target shape data to the shape fish which controls the shape in the width direction of the strip | belt-shaped metal material extended by the roll rolling machine, and adjusts the said shape, It shows the shape of the rolled metal A sensor for detecting real shape data, first sampling means for periodically collecting and storing the real shape data, shape change tendency calculating means for calculating a change trend of the shape within a predetermined time from the real shape data, and Based on the change trend and the latest real shape information, the latest target shape data, control gain, etc. are determined, and based on this, the 1st shape adjusting means which adjusts the said shape is provided, The metal material rolling characterized by the above-mentioned. Shape adjuster. In the metal material rolling shape adjustment apparatus which adjusts the shape by giving target shape data to the shape control part which controls the shape in the width direction of the strip | belt-shaped metal material extended by the roll rolling machine, the shape of the rolled metal Calculating statistical characteristic information of the shape within a predetermined time from a sensor for detecting real shape data to be displayed, second sampling means for periodically collecting and storing the real shape data, and real shape data stored in the sampling means. A second shape adjusting means for determining the latest target shape data, control gain, etc. based on the statistical characteristic information calculating means and the statistical characteristic information and the latest real shape information, and adjusting the shape based on this; It is provided with a metal material rolling shape adjusting apparatus.

Images