JP2735344B2 - System drive - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention controls a predetermined plant or issues an alarm to the plant or drives an alarm system (hereinafter referred to as a control system or the like) in a timely manner. It relates to a system drive.

(Prior art)

FIG. 19 shows a roll mill 2 for rolling aluminum foil, which is an example of the background of the present invention. In aluminum foil rolling, about 700 to 1700 mm in width and several μm to several hundred μ in thickness wound on the entrance coil 50
m of the raw aluminum foil 51 is rolled by a pair of rolling rolls 52 at a speed of about 300 to 1200 m / min, and the thickness thereof is about 1/2 to 1
Reduced to / 3. Then, the rolled aluminum foil 53 is conveyed in the direction of arrow K by a constant tension generated by the rotation of the drive shaft of the output side coil 64 (FIG. 1), and wound around the output side coil 64.

For example, when a raw aluminum foil 51 having a thickness of several hundred μm is finally rolled into an aluminum foil 53 having a thickness of several μm, the rolling process is repeated several times, and the number of rolling is referred to as the number of passes. Can be

In the metal rolling in micron units as described above, as shown in FIG. 20, although the aluminum foil 53 has the same thickness, a portion that “extends” in the foil width direction (arrow L). And "stretched" parts are remarkably present. That is, the extension portion 54 has a peak portion 56 and a valley portion 57 formed along the transport direction of the aluminum foil 53 (arrow K), and the tension portion 55 has a generally flat shape. Therefore, the aluminum foil shown in the figure
Reference numeral 53 denotes a state in which the central portion in the foil width direction (arrow L) is extended and its end is stretched.

The distribution of the degree of elongation and the degree of tension in the foil width direction (arrow L) is hereinafter referred to as the surface shape or actual shape of the aluminum foil 53. The surface shape has a great influence on the quality of the foil product, and in some cases, a large tension is applied to the tension portion 55, which causes the foil to break. Further, the stretched portion 54 causes wrinkles. And, as for the aluminum foil 53 as the final product, it is natural that a flat shape in which the elongation and tension occur uniformly is desired, but it is not necessarily a flat actual shape for each pass, but an intermediate shape. The shape of the aluminum foil 53 in the path is various.

The surface shape of the aluminum foil 53 as described above is
Can be controlled by changing the shape of As shown in FIGS. 19 to 21, the rolling roll 52 bulges called a thermal crown due to heat generation during rolling and its heat conduction characteristics. The example shown in FIG.
Is bulging. Such a bulge, that is, a thermal crown, depends on the appearance location and the degree of bulge.
Change the surface shape of 53. That is, the aluminum foil rolled at a portion where the degree of expansion of the thermal crown of the rolling roll 52 is large.
In the case of 53, the rolling portion is in an elongated state. Therefore, the surface shape of the aluminum foil 53 depends on the temperature or the spray amount of the coolant 58 (FIG. 1) sprayed toward the rolling roll 52 to cool the rolling roll 52 in the width direction of the aluminum foil 53 (arrow L ( No.
20) can be controlled.

Such shape control of the aluminum foil 53 is performed by the roll mill 2.
This is performed by the shape control unit 3 provided next to. That is, the shape control unit 3, the rolling delivery side is provided rotatably on the roll 52, the foil width direction (arrow L) consisting of 36 to split the element 4 _e to the inspection roll 4, aluminum foil 53 The actual shape data of the elongation and tension of is input. Each element 4 _e has
Each first piezoelectric element (not shown) is embedded acts as a sensor for detecting a pressing force applied to the outer peripheral surface of the element 4 _e.

Then, the aluminum foil 53 pressed on the element 4 _e and pulled in the transport direction (arrow K) by a constant tension is
Contact pressure against the element 4 _e when the elongation portion 54 passes over the element 4 _e is small, when the site 55 tension the contrary has passed is detected greatly.

Therefore, the actual shape of the aluminum foil 53, as shown in FIG. 21, expressed as a width direction of the distribution of elongation obtained by converting the pressure contact force data detected from the respective elements 4 _e (actual shape data). In the case shown in the figure, the quota portion a of the rolling roll 52 is excessively bulged, so that cooling of that portion is promoted,
A target shape is set so that heat is stored in the central portion of 52 and both ends thereof.

The shape control unit 3 compares the actual shape data with the previously input target shape data, and injects the actual shape data toward the portion of the rolling roll corresponding to the element 4 _e having a higher elongation. Increase the amount of coolant 58 that is used. The coolant 58 is provided on the entry side of the rolling roll 52, and is sprayed from a spray pipe 59 which is split in the foil width direction (arrow L) and sprayed.

Thereby, the thermal crown of the rolling roll 52 is relieved,
The part of the aluminum foil 53 corresponding to the quarter part a is deformed toward the tension state. If the actual shape data has a lower elongation, the reverse operation is performed. Incidentally, the target shape data, as to promote cooling of the quota portions a of the rolling roller 52, elongation resulting from element 4 _e corresponding to the site is often set to zero.

As described above, in foil rolling, rolling (passing) is repeated a plurality of times to obtain a final product. The plan of what pass and what thickness or surface shape to obtain the final product in what pass is called an operation policy. In this operation policy, in order to finish the above-mentioned final product into a flat shape with uniform stretch and tension, a target shape of stretch and tension in a middle pass is set together with a target of thickness.

In an actual operation, the distribution of elongation and tension in the above-described intermediate path is different. This is because, as described above, operating conditions such as deformation of the rolling rolls due to heat (thermal crown) differ for each pass. For this reason, among the actual operating conditions, the operating policy in which the target shape of elongation and tension in each pass is determined in consideration of the operating conditions expected to some extent.

However, in many cases, the target shape of the stretch in the operation policy as described above does not match the actual target shape of the stretch / tension by the shape control unit 3, that is, the target shape data. For example, a certain rolling mill at a certain pass
When rolling is performed with a target shape based on the operation policy as shown by the broken line in FIG. 22 as an example, a case where good results are obtained by setting the actual target shape data set in the actual shape control unit 3 as shown by a solid line is an example. It is.

As described above, the reason why the target shape in the operation policy does not match the target shape in the actual control is that the coolant amount for the rolling roll 52 is divided and adjusted in accordance with the element 4 _e of the inspection roll 4. In addition, the thermal crown moves due to the heat conduction of the rolling rolls, and the mode of the movement changes from moment to moment depending on the type of material, operating conditions such as the surface shape of the rolling rolls, heat balance, temperature, rolling speed, and foil shape. It is. Many of the effects of such operating conditions and types of materials cannot be expressed by mathematical models. Therefore, in order to approach the operation policy, the target shape data must be adjusted in accordance with various actual shapes appearing in the middle of each pass.

[Problems to be solved by the invention]

However, since the target shape data described above is left to the operator 5 to determine whether or not the target shape needs to be changed and to start the target shape change adjustment process, such as when the target shape must be changed. Automatic processing of the change adjustment was not possible. For this reason, it has not been possible to adjust the target shape appropriately and in a timely manner by following the operation conditions of the aluminum foil rolling that change slightly.

Furthermore, even when the determination of the necessity of the target shape change and the activation of the target shape change adjustment process are automatically performed by the threshold determination for the detected actual shape data, the actual shape data is set to 0 and 1 after the threshold determination. Since it is processed as binary logic, it has been difficult to precisely grasp the degree of abnormality of the actual shape. For example, when an abnormality occurs only in a certain state A related to the actual shape, the process is not started when the actual shape data slightly exceeds a threshold, and when the state A and another state B are simultaneously abnormal. However, if the actual shape data of the state A slightly exceeds the threshold value, it cannot be determined that the process is started.

On the other hand, if only the actual shape data at the present time and the shape characteristics derived by calculation from the actual shape data are used, not all rolling states can be expressed. That is, characteristic information that cannot be obtained unless data is collected periodically and actual shape data at several past times is used, such as the tendency, average, variance, correlation between data, and three-dimensional pattern recognition of collected data. There are many. Such characteristic information is also important in adjusting the surface shape of the aluminum foil 53. However, at the moment,
If the judgment is based only on the so-called latest actual shape data,
This cannot be said to be sufficient for driving the shape control unit 3.

Also, in the rolling process of aluminum or the like, a shape control device for performing appropriate shape control according to the plant operating state is known as “shape control by fuzzy theory” (Hitachi Review, Hitachi, Ltd., August 25, 1989). Issued daily, No. 71
Vol. 8, No. 8, pp. 103-108). This shape control device is intended for shape control using a coolant of an aluminum rolling mill, a shape detector for detecting shape information of a material to be rolled, storage means for storing the shape information, and a size deviation from the shape information. The shape change over time,
Means for calculating the certainty factor of the spatial relationship with the neighborhood of the shape deviation and the neighborhood of the shape deviation, and from each of the certainty factors, the membership function value of the rate of change of the degree of poor shape in the direction correctable by the coolant using fuzzy rules is calculated. Means for calculating, and a means for calculating a degree of poor shape in a direction that can be corrected by coolant by taking a weighted average for each membership function value, and determining an ON / OFF pattern of a coolant valve using the degree. Things. However, in this shape control device, the certainty factor is small, and the value of the weighted average finally obtained based on the certainty factor is a small value that does not require a change in the coolant valve ON / OFF pattern. Even if it is obvious that the weighted average will always be calculated, the load on the computer will be large and the processing will take a long time, which is a major problem in a control system that requires real-time processing. there were.

Accordingly, it is an object of the present invention to provide a system driving device capable of driving a control system or the like appropriately and in a timely manner according to the operation state of a plant.

[Means for solving the problem]

In order to achieve the above object, the main means adopted by the present invention is, in its gist, a system driving device for driving a control system or the like for a predetermined plant. , A storage unit for storing the plant data, a confidence calculation unit for calculating the confidence for each plant state from the plant data, and a first confidence level for each plant state. Importance calculating means for comparing the threshold with a threshold, selecting a plant state exceeding the first threshold, and calculating the importance of a control target related to the plant state so as to be a monotonic change function with respect to the certainty factor; From the respective degrees of importance, a composite importance is calculated so as to be a monotonic change function with respect to the degree of importance, and when the composite importance exceeds a predetermined second threshold value, It is a system driving device according to a point where it and a control means for driving the control system or the like. The present invention is typically applicable to a material shape adjusting device in metal rolling. In addition, the timing of driving the system can be adjusted using the statistical data, and the determination of the plant state is performed more advantageously by performing the determination using the pattern recognition.

[Action]

According to the present invention, first, from plant data obtained from a predetermined plant, certainty factors for each of one or more types of plant states classified in advance are calculated by the certainty factor calculation means. Next, the importance calculating means compares the certainty factor with a first threshold value set for each plant state, selects a plant state related to the certainty factor exceeding the first threshold value, and The importance of the control target related to the plant state, which becomes a monotonous change function, is calculated.
Therefore, if the degree of certainty of the plant state is equal to or less than the first threshold value, the calculation of the importance regarding the plant state is not performed. Further, in the control means, from the importance for each of the plant states, a composite importance in which the respective importance is comprehensively calculated by a monotonic change function using the respective importance as a variable is obtained, and the composite importance is a predetermined value. When the value exceeds the second threshold, a control system or the like for the plant is automatically driven. In addition, when the importance regarding all the plant states has not been obtained, the process of obtaining the composite importance is not performed. When the first threshold value is equal to the predetermined second value and the composite importance is equal to the predetermined second value.
This is because if the value is set to a value small enough to predict that the value does not exceed the threshold, the composite importance does not exceed the second threshold in such a case. Thus, the first
By restricting the processing of calculating the importance and the composite importance based on the threshold, useless calculation is minimized, and the processing can be sped up.

Further, if the plant data is calculated from the plant data, for example, the characteristic information of the plant state within a predetermined time in the past, and based on the calculated information, the above-described confidence calculation unit and subsequent steps are executed. The control system and the like can be driven appropriately and timely by judging the trendy plant state without being affected by the noise data.

〔Example〕

Subsequently, embodiments embodying the present invention will be described with reference to the accompanying drawings to provide an understanding of the present invention.

FIG. 1 is a schematic view showing a system arrangement of an aluminum foil rolling target shape adjusting apparatus according to one embodiment of the present invention.
FIG. 3 is a configuration diagram showing a processing flow of the aluminum foil rolling target shape adjusting device, and FIG. 3 (a) is an explanatory diagram showing main parts of actual shape data represented by an elongation distribution in a foil width direction.
FIG. 4B is an explanatory diagram showing the actual shape classification items of the aluminum foil classified by pattern, FIG. 4A is a flowchart showing a processing procedure for judging a change tendency of the actual shape data, and FIG. FIGS. 5A and 5B are explanatory views showing an action of correcting the number of levels of the actual shape classification items in consideration of the shape change tendency. FIGS. 5A and 5B show the correlation between the number of levels of the two actual shape classification items. FIG. 6, FIG. 6 is a three-dimensional graph showing a three-dimensional pattern of the actual shape data of the aluminum foil over time, FIG. 7 is a schematic diagram conceptually showing a neural network, and FIG. FIG. 9 is an explanatory diagram showing an example of the relationship between a change target and an action candidate corresponding to the change target. FIG. 9 is a graph showing the relationship between the importance of the shape change target and the certainty factor when the actual shape classification item is tight. Figure shows action candidate inference Explanatory view showing an example of changing the target shape rules used and using the same reasoning in, 11
FIG. 12 is an explanatory diagram showing a processing procedure of a routine for checking the validity of an action to be applied by a check tree, and FIG. 12A is an explanatory diagram showing a target shape adjustment parameter used for changing a target shape. Figure, Figure (b)
State diagram illustrating a change in circumstances a ₃ of the parameters, the figure (c) shows a state diagram illustrating the state central portion of the target shape to be adjusted by a ₄ of the parameters in the order pattern, the (d) of FIG Is a state diagram showing a situation where the central portion is an inverted pattern, FIG. 13 is a schematic explanatory diagram showing an inference execution example for changing the target shape, FIG. 14 is a flow sheet showing a processing flow of target shape adjustment, 15 is a display diagram showing an input menu displayed on the screen of the rolling mill-side terminal, FIG. 16 is a display diagram showing an example of a target shape displayed on the screen, and FIG. 17 is an asymmetrically obtained actual shape. FIG. 18 is a flow chart showing a method of correcting a shape, and FIG. 18 is a schematic explanatory view showing a situation of correcting the asymmetric actual shape.

In the following description, elements common to those of the roll rolling mill 2 shown in FIGS. 19 to 21 will be denoted by the same reference numerals, and description thereof will be omitted.

Further, the following embodiments are merely examples embodying the present invention, and do not limit the technical scope of the present invention.

The aluminum foil rolling target shape adjusting device 1 according to the present embodiment includes:
It is applied to process online control as an expert system that systematizes the control know-how of the 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 with reference to FIG.

As shown in the figure, the aluminum foil rolling target shape adjusting device 1 includes a rolling data collection unit 7, a rolling state analysis unit 8, a control target generation unit 9, an action candidate inference unit 11, a target shape generation unit 12, and an action effect. Rolling condition analysis knowledge base mainly composed of evaluation unit 10 and storing operational knowledge
D ₁ , control goal setting knowledge base D ₂ and action inference knowledge base D ₃ (including knowledge of maintaining rule consistency and action validity) and working memory for temporarily storing various data
M ₁ , M ₂ , M ₃ , M ₄ , M ₅ . The outline of the processing contents in each of the above units will be described below.

In the aluminum foil rolling target shape adjusting device 1, first, an inference process is started by a key input from the terminal device 6 on the aluminum foil rolling mill 2 side, and operating condition data from the aluminum foil rolling mill 2 is transmitted through the shape control unit 3. Is entered.

Rolling data collection unit rolling data collecting unit 7 receives the operation condition data including the actual shape data from the shape control unit 3 writes into the working memory M _1.

Rolling state analyzing unit The rolling state analyzing unit 8 analyzes the actual shape data and determines the rolling state of the aluminum foil 53. That is, the actual shape data, to determine in advance several are classified into the feature pattern how each adapted to the actual shape pattern stored in the rolled condition analysis knowledge base D _1.

 At the same time, the current target shape data is analyzed.

Control target generation unit The control target generation unit 9 determines the direction in which the actual shape of the aluminum foil 53 is changed based on the analysis result of the actual shape data by the rolling state analysis unit 8 and the input from the terminal 6 by the operator 5. Set the control target.

Action Candidate Inference Unit- (1) Rule Inference The action candidate inference unit 11 is an IF-THEN-format action in which the above-mentioned control target and operating conditions and the like are used as a condition part and an action and the like for realizing the above-mentioned control target are used as a conclusion part. the rule inference according to the rules stored in the inference knowledge base D _3, working memory an action is determined to be valid
It is written in the M _5. At the time of this writing, the following processing is executed.

-(2) Resolution of contradiction / redundancy When contradictory action candidates are listed, the action of the control target that is more important is applied.

-(3) Learning of invalid action In the above rule, when a plurality of action candidates exist for a certain control target, priorities are assigned to the action candidates, and only the action having the highest priority is applied. I do. Actions executed to achieve a certain control target and having no effect are not repeatedly applied even if the same control target is set next time.

Target shape generation unit The target shape generation unit 12 generates new target shape data based on the applied action and outputs the data to the shape control unit 3. Based on the target shape data, the shape control unit 3
Controls the aluminum foil rolling mill 2.

Action Effect Evaluation Unit The action effect evaluation unit 10 evaluates whether the control of the aluminum foil rolling mill 2 based on the applied action was effective based on the result of the data analysis and the inquiry to the operator 5. In action evaluated as invalid, working memory M ₄ to be stored, is referred to when the next action candidate Inference action candidate inference section 11.

 Hereinafter, this embodiment will be described in detail.

As shown in FIGS. 1 and 2, the aluminum foil rolling target shape adjusting device 1 controls the shape control unit 3 which controls the injection amount or temperature of the coolant 58 so as to adjust the actual shape of the aluminum foil 53. Target shape data serving as a guide for the control is output, and operation condition data is input from the shape control unit 3.

As a method of adjusting the actual shape of the aluminum foil 53, there is also a method of controlling the push-up force of the push-up roll 60 that urges the upper side of the lower rolling roll 52 toward the upper rolling roll 52, In the present embodiment, only the control of the coolant 58 will be described below.

Collection of Operating Condition Data In the aluminum foil rolling target shape adjusting device 1, the inspection roll 4 is provided with an extension portion 54 of the aluminum foil 53 at the time of rolling.
As a set of elements 4 _e having a sensor for detecting actual shape data indicating the tension portion 55 (FIG. 20), the shape control unit 3 is provided downstream of the rolling roll 52 in the transport direction (arrow K). The actual shape data is transferred to the rolling data collection unit 7
(FIG. 2). The rolling data collection unit 7 writes the operational condition data transferred from the shape control unit 3 at every predetermined time interval (Table 1) in the working memory M ₁ At the same time, the rolling state analysis unit 8 is activated.

Analysis of actual shape data From the actual shape data detected by element 4 _e (sensor), the elongation state and tension state (each actual shape classification item) of aluminum foil 53 during rolling, and the rolling state for calculating their degree analysis knowledge base D _1, as shown in FIG. 3 (b), for example, the actual shape classification items such as "end-clad" - "mumps elongation" is detected from the element 4 _e, the actual shape data indicating the actual shape , And a specific program for each item for specifying the actual shape classification item.

Here, the method of specifying the actual shape classification items will be described first. Actual shape data detected as tension distribution from the element 4 _e, as shown in FIG. 3 (a), obtained in the form of elongation distribution in the foil width. The actual shape data shown in the figure includes an end portion, a quarter portion, a center portion A, and a center portion B from the outside, and the entire center portion is constituted by the center portion A and the center portion B. In this case, the tension portion 55 is located at the center portion B, and the extension portions 54 are located at the quarter portions on both sides.

The actual shape classification items stored in the rolled condition analysis knowledge base D _1, as shown in FIG. 3 (b), are classified into five main types, as described below.

(1) “Tension”: When the elongation at the end shows a lower value toward the end, it is considered as an end, and whether the value at the end is the minimum value of the whole.

Judgment is made based on the degree of elongation difference between the edge and the quota.

(2) "End elongation": Contrary to the case of end tension, a case where the value of the end is significantly larger than that of other parts. As described above, an actual shape having a somewhat extended end is often preferable, but an excessively extended end is regarded as a problem shape.

(3) “Quarter elongation”: Judgment is made based on how large the elongation percentage value of the part that is most elongated near the part corresponding to the zero point set in the target shape is larger than that of the end part. As described above, the actual shape in the quarter portion is the easiest to stretch, and almost this type appears in the actual actual shape.

(4) "Medium tension": tension in the center (low elongation)
Is compared with the one at the end.

(5) "Medium elongation": The difference between the elongation percentage value of the most elongated part in the center part and that of the most elongated part as a whole (in most cases, the end part or the quarter part) is small. Or if it is negative, it is determined that the medium growth. The middle elongation is classified into two types: general middle elongation with the central part stretched, the area between the quota part and the central part is extended, and the central part is stretched by the mumps.

Other unique actual shape classification items include the following “non-target” and “zero point inappropriate”.

“Asymmetric”: Normally, the target shape is symmetrical in the foil width direction, and the actual shape is generally the symmetrical shape, but a shape in which the symmetry is broken is called an asymmetric shape. As a criterion, the difference between the maximum elongation percentage and the minimum elongation percentage at both ends is large.

One end is identified as being stretched, and the other as edge extension.

Are considered asymmetric if any of Of course, both cases may be satisfied.

“Inappropriate zero point” means a case where the zero point portion set in the target shape does not match the portion showing the maximum elongation percentage in the actual shape. Normally, heat easily accumulates in the quarter part a of the rolling roll 52, and the real-shaped quarter part corresponding to the quarter part a is most easily stretched. Therefore, when setting the target shape, a zero point is set at the most extended portion of the quarter portion of the actual shape. And
If these parts are displaced, it is necessary to make adjustments so that they match.

From the detected actual shape data, it is determined which of the actual shape classification items the current shape state corresponds to in accordance with the method described in the section “specification method” in FIG. 3B. Such techniques are stored as programs in the rolling condition analysis knowledge base D ₁ as described above.

As noted above, one or more solid shape classification items cause actual shape data input from the work memory M ₁ is in the rolled status analyzing unit 8 is calculated by the specific program.

Normally, it is not always the case that only one actual shape classification item for which a certain actual shape is determined to be in that state is selected. Since the actual shape data is obtained as a result of operating conditions that are intricately entangled, it is often determined that there are a plurality of actual shape classification items. In such a case, there are actual shape classification items having a strong causal relationship with the actual shape data and weak classification items. The strength of such a causal relationship, that is, the degree of the shape state is referred to as a certainty factor.

The rolling state analysis unit 8 converts the actual shape data into 1 or 2 at a certain degree of certainty derived by an appropriate function.
Refine the above solid shape classification items, and stores the said actual shape category and confidence to the working memory M _2. For example, within the range of the actual shape data aluminum foil 53 is inputted from the four elements 4 _e from the placed end elongation difference alpha ₂ as a whole the actual shape data of the most high growth sites and end of the index When the ratio β ₂ / α ₂ of the difference β ₂ between the maximum value and the minimum value of the elongation percentage exceeds a predetermined set value, the actual shape at this time includes the actual shape classification item “end elongation”. And a certainty factor from 0 to 1 is added according to the value of the ratio. The same applies to other actual shape classification items.

In this case, on the basis of the actual shape data within a predetermined past time period from the working memory M _1, the statistical characteristic information of rolling mill 2, for example, the average, the change trend, dispersion, correlation, and calculating the three-dimensional pattern recognition Based on the statistical property information, that is, using the statistical property information as a variable of the certainty factor, the certainty factor for each actual shape classification item may be calculated.

For example, the actual shape at a certain point in time may be extremely different from the actual shape before and after that for some reason. Specifically, due to an abnormality in the shape of the raw material plate before rolling, when the actual shape of the "end tension" continues, the state of "edge extension" is detected for a moment, and then the original "edge tension"
This is the case where the state continues.

Therefore, if the average value is applied among the statistical characteristic information of the actual shape data at the past several time points up to the certain time point,
The rolling state of the trend can be determined without being affected by the above-mentioned noise factor.

Next, a case where a change tendency of the actual shape data within a predetermined time in the past is adopted as the statistical characteristic information will be described in detail below. The calculation of this tendency is performed, for example, according to the processing procedure of the flowchart shown in FIG. During the rolling operation of the aluminum foil 53, the actual shape data of the aluminum foil 53 obtained from the inspection roll 4 on the roll mill 2 is collected by the rolling data collection unit 7 at predetermined time intervals (S4).
0).

Next, the actual shape data is compared with the actual shape classification item in the rolling condition analysis unit 8 and specified as a certain actual shape classification item, and the degree is represented by the number of levels indicated by natural numbers 0 to 5. That is, the rolling state of the aluminum foil 53 is determined (S41).

For example, 4 from the end where the aluminum foil 53 is pressed.
Within the range of the actual shape data input from the four elements 4 _e , the difference in extension rate α ₂ between the portion and the end having the highest elongation rate and the difference β ₂ between the maximum value and the minimum value of the elongation rate in the entire actual shape data
If the ratio β ₂ / α ₂ exceeds the predetermined set value, it is determined that the actual shape at this time includes the actual shape classification item “end elongation”, and according to the value of the ratio, The degree of “edge elongation” is represented by the number of levels indicated by natural numbers 0 to 5.

Further, in step S42, with respect to the actual shape data at time T, is determined the shape points H _t to +5 points -5 points, stored in the storage unit (not shown) for storing the shape points H _t within a predetermined time period You.

This embodies an empirical handling method when the operator 5 specifies actual shape data to an arbitrary actual shape classification item, and is, for example, so that the operator 5 can easily understand when displayed on the terminal 6. The reliability of the actual shape classification item to which the actual shape data fits and the confidence of the actual shape classification item whose real shape classification item and elongation / tension state contradict each other are defined as both coordinates consisting of positive and negative natural numbers centered on 0. It is converted.

For example, if the actual shape classification item is “edge elongation” and the degree is “level 5”, the number of shape points is +5, and if the level is “1”, the shape point is +1. In addition, level 5 with "edge tension"
In this case, the score is -5 points, and in the case of level 1, the score is -1 point.
If it does not correspond to "edge extension" or "edge tension", it is scored 0 points.

If it is necessary to change and adjust the target shape by the input from the terminal 6 on the roll mill 2 side or by the input from the rolling condition analysis unit 8 at the request of the operator 5 (S43), in step S44, The change tendency of the actual shape within a predetermined time until is calculated.

For example, the shape change trend at the time T in the past i pieces of shape points H _{t-i +1,} ..., is calculated from the H _t. Here is the past
Ten shaped score H _t-9, ..., an example which is calculated from the H _t. First, the shape number H _t-9, ..., the maximum and minimum shape score gap shown in the following equation from the value H _d of the H _t is determined.

_{H d = max (H t-} 9, ..., H t) -min (H t-9, ..., H t) at that time, if the shape score gap H _d is 2 or less, the actual shape of the aluminum foil 53 Is in a stable state, and no shape change tendency is recognized.

If _Hd is 3 or more, it is first determined whether the actual shape is in a periodically changing state. For example, the shape points _Ht-9 , ..., _Ht
If that, as shown in the following equation, _- the number of shape points is changed in the direction of increased between the H _+, the number of changes to the decreasing direction _{H | H + -H - | ≦} 3, and 3 ≦ H _d ≦ 4 That is, when the absolute value of the difference between H ₊ and H ₋ is 3 or less and the shape point difference H _d is 3 or more and 4 or less, it is determined that the actual shape has a periodic change tendency. .

When _Hd is 2 or more and other than the above equation, a shape change tendency is recognized, and it is determined what the tendency is. For example, the current number of shape points H _t
Is compared with the shape point number _{Ht-9 at the} tenth point in the past.

That is, as shown in Table 2, _Ht is higher than _Ht-9 . If the value also indicates a large value, it indicates that the end has a tendency to elongate, while if the value indicates a small value, it is determined that the end has a tendency to be taut. Further, when _Ht and _Ht-9 are equal, it is determined that the actual shape is stable, and the process is performed bypassing the level number correction step (S45) of the actual shape classification item described later.

Therefore, for example, if the shape change tendency at time T is “edge extension tendency” and the actual shape classification item at that time is specified as “edge extension” and the number of levels is 3,
As shown in FIG. 4 (b), a correction action is performed to increase the number of levels by 2 with respect to the number of levels of the actual shape classification item at that time (S45). This is because level 3 in the above example is inappropriate if the shape judgment at the present time and the shape change tendency are taken into consideration in performing shape control in the future, Is considered to be almost equivalent to level 5 (a state in which edge elongation is rapidly progressing).

Therefore, in step S47 in FIG. 4 (a), the target shape data is obtained based on the latest actual shape classification item specified as described above and the corrected number of levels so that a desired actual shape is obtained. The change is adjusted and output to the shape control unit 3 (S48).

On the other hand, when it is determined in step S44 that the actual shape has a periodic change tendency, the control gain for changing and adjusting the target shape data is given by being reduced by 30% (S46). Thereby, the periodic change tendency of the actual shape is suppressed.

On the other hand, when variance is applied as the statistical characteristic information,
For example, the rolling state can be determined from the degree of dispersion of the actual shape data on the shape of the end of the aluminum foil 53 due to a temporal change. That is, when the degree of the dispersion is large, it indicates that the rolling state is unstable. Therefore, in such a case, the following control actions (i) and (ii) for stabilizing the rolling state are performed by the shape control unit 3.

At this time, (i) the control gain of the coolant amount is changed and set lower.

(Ii) Since it is conceivable that the rolling roll 52 is not sufficiently heated, the amount of coolant is decreased overall or the elongation percentage of the target shape is increased overall.

FIG. 5A shows the correlation between the quarter elongation level and the edge tension level as an example in which the characteristic information is a correlation. At this time, the relationship between the two can be said to be positively correlated, indicating that the quarter elongation and the edge tension are likely to occur simultaneously.

In such a case, it has been obtained as empirical knowledge that a shape adjustment method such as reducing the amount of coolant at the end portion to eliminate the end tension is not effective. On the other hand, if the relationship between the two is uncorrelated, as shown in FIG. 5 (b), the knowledge has been obtained that the shape adjustment method adopted in the case of the positive correlation is effective. The actual shape is adjusted by the shape adjustment method.

Further, it is also possible to recognize the three-dimensional pattern relating to the actual shape data as the statistical characteristic information. The three-dimensional pattern P is shown in FIG. In the figure, an arrow M indicated by a two-dot chain line, elongation at each time t ₀ ~t ₆ indicates the temporal position changes of the element 4 _e which detects the maximum value.
As illustrated, the element 4 _e of the elongation maximum is meandering in width of several foil width direction amino element 4 _e. In such a state, not only the part that is currently maximally extended,
Equally intensively the rolling roll 52 which corresponds to the width of said several elements 4 _e including the site must inject coolant 58. This is because, even if the coolant 58 is sprayed only to the portion that is currently maximally extended, the extended portion only moves to the adjacent portion. Statistical characteristic information represented by such a three-dimensional pattern P, such as “the part of the actual shape in the stretched state is meandering with time” can be determined by the above-described numerical calculation algorithm. A recognition method using a neural network for performing pattern recognition of the actual shape data of the aluminum foil 53 is effective.

In the neural network 20, as shown in FIG. 7, a plurality of neurons 15 for calculating and outputting input data by threshold processing are conceptually arranged as an input layer, an intermediate layer, and an output layer. Connected via 16. Then, the neural network 20 uses the pattern data of the actual shape data of the aluminum foil 53 as input data, uses the actual shape classification items and their degrees as output data, and the correspondence between the two changes the connection weight of the connection unit 16. It is learned by doing. Therefore, when new real shape data is input to the learned neural network 20, the real shape data is specified in any of the real shape classification items together with the degree. Such a neural network 20 may be applied to the rolling situation analysis unit 8.

That is, the rolling state analyzing unit 8 is a certainty factor calculating unit and a characteristic information calculating unit.

Here, as a calculation method for determining the actual shape state of the aluminum foil 53, a numerical calculation example by a specific method determined for each actual shape classification item has been mainly described, but the determination by the neural network 20 or such a method is used. It goes without saying that a similar calculation effect can be obtained by a determination based on a rule base (not shown) storing the determination knowledge.

The actual shape data to be subjected to the input operation to the rolling status analyzing unit 8 from the working memory M _1, the value of the type of data at a time in the predetermined time, or several data values at one time Alternatively, it may be one of the values of one type of data at several points in time or the values of several types of data at several points in time, and data necessary for the calculation may be appropriately used.

Thus, the rolling status analyzing unit 8, the current actual shape category and confidence with respect to the actual shape data of the aluminum foil 53 is determined, and written into the working memory M _2.

That is, the rolling status analyzing unit 8 is a plant state classifying means, a rolling situation analysis knowledge base D ₁ is the pattern storage means.

Generation of Control Target The control target data (shape change target (FIG. 8) and its importance), which is a key in appropriately setting or changing the target shape, is rolled by the operator 5 in the control target generation unit 9. or is input from the rolling mill 2 side of the terminal 6, or on the basis of the working memory M actual shape classification items and the rolling status data of the confidence or the like in ₂ is automatically generated. In this automatic generation, "the operation policy (such as" rolling by extending the end greatly in a predetermined pass ")" or "when the input by the operator 5 and the automatically generated one conflict, the rule such "give priority to input information of the operator is applied with reference to the control target setting knowledge base D _2.

For example, if the detected actual shape data includes “edge extension” as in the above example and the confidence at that time is 0.8, five shape change targets are set in order to eliminate “edge extension”. Is selected, and the importance corresponding to the certainty factor (0.8) is given to the selected shape change target. Then, the shape change target and its importance are stored in the working memory M _3.

The situation where the importance of the shape change target is calculated in the control target generation unit 9 from the certainty factor of the actual shape classification item specified as described above will be described in detail below.

The importance is a measure of whether the target shape data needs to be changed, and is shown in a graph (FIG. 9) showing the relationship between the actual shape data and, for example, the certainty factor of the actual shape classification item indicating the degree of edge tension. Is shown. In the drawing, an example of “edge tension” in the actual shape classification items is shown.

For each of the actual shape classification items such as edge tension and edge elongation,
An importance calculation function (f (x)) that gives the degree (importance) of the necessity of changing the target shape is defined. For example, the importance calculation function f ₁ () for “edge tension” is Is defined as

That is, the end tension degree (confidence) is compared with the first threshold value L _1, as shown in FIG. 9, when exceeding the first threshold value L _1, for the real shape classification item of target shape change The degree of importance corresponding to the selected degree of certainty is calculated.
On the other hand, if the confidence level is a first threshold value L ₁ or less, the degree of importance 0 is given, the actual shape category never be combined upon change in the target shape. Wherein for each solid shape classification item, the first threshold value L ₁ is respectively set individually, each of the confidence and said first threshold value L ₁ is the comparison operation, whether to Kyosu the change of the target shape items It is determined every time.

Whether or not it is necessary to change the target shape is determined for each actual shape classification item as described above, and sometimes it is determined whether or not the average of the combined importance of each item exceeds a threshold value. is there.

Subsequently, a method of determining whether or not the target shape needs to be changed based on the combined importance or the average thereof from the actual shape classification item selected to change the target shape will be described in detail.

If the actual shape classification items are S ₁ , S ₂ ,..., S _i , the importance calculation function f for each item defined for them
_{The combination} importance calculation function g () obtained by combining _si () is defined as follows.

g (f _s1 (x ₁ ), ..., f _si (x _i )) The function g () is g≡ (Σf _si (x _i )) / i, or g≡Σf _si (x _i ) where 0 ≦ f _si (x _i ) ≦ 1, f _si (x _i ) is x _i is expressed in the form of a sum average or a sum, such as a strong monotone increase.

Here, when f (L _i ) is (L ₁ ), the strong monotonic increase means that when L ₁ <L ₂ , f (L ₁ ) <f (L ₂ ).

As described above, the combined importance indicating the degree of necessity of starting the inference processing is calculated from the importance of each item in the action candidate inference unit 11, and the combined importance is determined by the predetermined second threshold L.
When the number exceeds ₃ (not shown), the inference processing is started, and the target shape data is appropriately changed.

The following is a specific example. When the importance for "edge tension" is 0.4, the importance for "quarter elongation" is 0.6, and the importance for other ("edge elongation", "medium elongation", "medium tension") is 0. In the case of the sum, the combined importance is 0.4 + 0.6 + 0 = 1.0. If the second threshold value L ₃ is 0.9 in this case, since the direction of the synthetic importance is high, the inference process is started. As a method of determining the importance as the trigger, the above-described method based on the sum average may be employed.

As described above, since the composite importance is a strong monotone increasing function (an example of a monotonic change function) for each importance, if the value of the importance, that is, the confidence is small, the value of the composite importance is Can be almost certainly predicted not to exceed the second threshold. Thus, if the confidence factor is the first threshold value L ₁ or less, is not calculated importance of the actual shape classification item, further, when the importance of all the real shape categories unsolicited, i.e., If the value of the synthetic importance is clear that does not exceed the second threshold value L _2, since the synthetic importance processing for obtaining is not performed, unnecessary calculations are minimized and speed of processing It becomes possible.

In this embodiment, as described above, when the confidence of each item exceeds the threshold value given to each, for example, the degree of "end tension" is also inferred when exceeding the threshold value L ₂ shown in FIG. 9 Processing is started.

Further, when the synthetic importance level exceeds the second threshold value L _3, simultaneously outputs a start signal to an alarm device, not shown, may be driving the alarm device.

 Next, the target shape inference processing will be described.

Application of Action-(1) Rule Inference Operating condition data including current target shape data and current actual shape data, rolling condition data including extracted actual shape classification items and their certainty factors, processed as described above. , And the control target data including the shape change target and its importance are transferred from the work memories M ₁ , M ₂ , M ₃ to the action candidate inference unit 11, respectively. Action candidate inference unit 11
Indicates the data transferred and the action inference knowledge base
Collating the condition part of the rule stored in the D _3, the result of the verification, extracts a rule which satisfies the condition part are all true, the target shape change data (FIG. 8 in the conclusion part of the rule , Hereafter called action). As described above, the inference processing for deriving a conclusion (a target shape to be adopted) corresponding to the conditions such as the operation condition data, the rolling state data, and the control target data depends on the knowledge (know-how) of an experienced person, as described above. I can't get it. In the present invention, such inference processing is automated. Rules for such automated reasoning is integrated in the action inference knowledgebase D _3, are stored. Such a rule is expressed in the form of "if [condition part], then [conclusion part]", and is expressed by a logical product as shown below.

If [control target data] and [operating condition data, rolling status data], [designation of target shape adjustment parameter and its change degree (action and its degree)] Here, the target shape adjustment parameter is a table. As shown in -4, it is an element that determines target shape data (a target elongation rate distribution in the path). Not all the parameters shown in Table 4 are described as the target shape adjustment parameters that constitute the conclusion of each rule. In many cases, only a part of the target shape adjustment parameters necessary to satisfy the condition part is described together with the degree of change.

For example, as shown in FIG. 10, in rule example 1,
If the actual shape of the aluminum foil 53 is specified as the quarter extension, and there is no zero below the portion where the extension near the quarter is the largest, then the position of the zero point is set below the portion where the extension near the quarter is the largest. stored in the action inference knowledge base D ₃ is a rule that specifies the action, such as come "have to.

-(2) Resolution of Inconsistency / Redundancy On the other hand, as seen in Rule Example 4 shown in Table 3, there are cases where incidental conditions are added to rules. For example, in the case where the edge tension and the quarter elongation occur simultaneously in the actual shape, rule example 2 and rule example 3 may be selected. They are Since both are established at the same time, the action candidate inference unit 11 cannot resolve the inconsistency, and an error occurs. Therefore, this problem can be solved by providing an additional condition in the condition part of the rule example 2 and setting it to rule example 4. That is, in rule example 4, if the importance of extending the edge and the importance of extending the quota are both greater than the first threshold, but the importance of extending the quota is less than 0.4, The target value of the end level (elongation rate at the end in the width direction) is increased. As a method for eliminating such inconsistency or redundancy, the shape change target having the higher importance can be prioritized.

On the other hand, there are cases where two or more types of shape change targets are simultaneously selected for certain actual shape data, and the contents of actions corresponding to the respective shape change targets are the same. For example, the actual shape classification items (FIG. 3 (b)) of the current actual shape data are simultaneously specified as "edge tension" and "medium extension", and the degree of action derived from each actual shape classification item is determined. The difference is that "increase the level of the end" (FIG. 8) having the same content is simultaneously selected. As a rule that applies to this case, it is used to set the degree of action in response to the confidence of the actual shape classification item identified is stored in the action inference knowledgebase D ₃ in advance.

Therefore, when actions having the same contents having different degrees as described above are selected at the same time, these actions are not executed at the same time. Redundancy is avoided. Here, conversely, only an action with a small degree of action may be executed, or an action corresponding to the average value of the actions may be selected or generated.

-(3) Learning of invalid action Further, as shown in FIG. 8, several types of action candidates with priority added to one shape change target are prepared. Then, when a certain shape change target is selected, the action having the highest priority is executed.

The priority is not fixed and is checked for each inference. For example, if the inference in action candidate inference unit 11 is executed, with respect to which the shape change target, which action it has adopted is stored in the working memory M _5, at action effect evaluation section 10 at the next inference, the last reshaping Whether or not the target has been achieved (evaluation of the effect) is determined by comparing the previous and current importance levels.

As a result, if the current target shape data changed based on the previous shape change target and its importance is determined to be valid in the action effect evaluation unit 10, that is, the importance is set to a value lower than the previous value. If so, the adopted action is considered valid and the action inference knowledge base D ₃
Is raised. Conversely, if it is determined to be invalid, and rules which infers selected last applied in action and the action was not effective it is stored in the working memory M _4.

For example, the shape change target shown in FIG. 8 indicates that the last time “I want to extend the end” is of importance 0 based on the rolling status data from the rolling status analysis unit 8 or the input data from the operator 5.
If it is determined in step 6, the highest priority "increase the level of the edge (priority 1)" among the actions associated with it is selected, and the level of the edge (elongation rate) according to the importance level 0.6. Is output to the shape control unit 3, and the simultaneously applied target shape data, shape change target, its importance (0.6), action, and its priority (1) are stored in the action inference knowledge base D _3. Is stored. Then, when the shape change target “I want to extend the end” is selected with the importance of 0.6 or more at the time of determining the shape change target of the present time, the edge tension state of the actual shape in question is improved. In many cases, the last applied action was invalid. Conversely, if "I want to extend the end" is selected this time with an importance of less than 0.6, it is determined that the previous action was effective.

Therefore, the actions that have been invalidated, the rules deduced the action is written in the working memory M _4. Then, even if the same shape change target is selected at the next inference and the invalid action is selected, the invalid action is not applied, and the next highest priority action determined to be appropriate is applied. Then, it is subjected to the change of the appropriate target shape data this time. 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 raised and the priority of the invalid action is lowered. Changed priorities as described above is written in terms of the priority of the action inference knowledgebase D _3.

In this way, even in the case where the target shape obtained by the aluminum foil rolling target shape adjusting device 1 has no effect on the actual shape and the actual shape of the problematic aluminum foil 53 continues. In this inference, a different action from the previous action is selected even if the shape change target is the same as that in the previous inference. Thereby, the actual shape is appropriately changed without the invalid rule being repeatedly applied.

Whether action mentioned as candidates is valid in the action candidate inference unit 11, 11 after being checked in detail with check tree shown in FIG., The work each time when it is determined that the appropriate memory M ₅ Registered in. still,
The aforementioned actions inference knowledgebase D _3, rule set comprising a plurality of rules having a condition part as described above same shape change target is set for each of the shape change target, the main area and the for storing the rule set A save area for storing rules removed from the rule set is secured. The rules for which the inference processing has been completed are sequentially removed from the main area to the save area.

In the check tree shown in the figure, first, it is determined whether a rule that matches the currently selected shape change target and the shape change target exists in a rule set to which the rule belongs, that is, a rule corresponding to the above rule set is included in the main area. there are any remaining is checked in case C _1. If Case C ₁ is true,
In Case C ₃ proceeds to check the strings attached according to an operating condition such as the condition part of the rule.

If not corresponding to the case C _3, that is, when all the condition part including incidental condition does not match, the rule is removed to the save area from the rule set of the main area. If satisfied the incidental condition, in case C _4, whether the action of the rule in which the condition part is satisfied all of its effect was not observed when there to have been applied in the past, i.e., the current working memory M ₄ It is checked whether it is registered. Then, if applicable to the case C _4, to remove the rule from the rule set (main area). Case C ₄
If not applicable, the presence of actions conflicting with each other is checked in case C _5. As mentioned above,
The number of shape change targets set is not limited to one, and a plurality of shape change targets may be selected. In this case, it is often the case that actions associated with each shape change target contradict each other.

Therefore, the check result, among the action registered already working memory M ₅ is to be valid, if any action that is inconsistent with the action that is currently checked, among the rules with an action , it is registered in the action working memory M ₅ of the rule if is larger severity shape change target that is currently selected, actions that already the conflicting registered are removed to save area.

Conversely, already left the work registered actions memory M ₅ If is larger importance shaped target with an action of the registered, rules the rule with the action that is checked It is removed from the set (main area) to the save area. If Case C ₅ is not applicable,
That is, when the above each action is consistent, whether the shape change goal of these actions are the same is checked in case C _6. If applicable to the case C _6, check whether there is a difference in the priorities of the respective action (Case C _7), if there is a difference in priority, only the action of the larger priority working memory M ₅ Left in
The smaller action of priority is erased from the working memory M _5. When there is no difference in priority, both actions are left together in the working memory M _5.

As described above, if not applicable to the case C _1, the rules that match the currently selected shape change target is judged not in the rule set proceeds to check the case C _2.

If applicable to the case C _2, i.e. rules of each rule set in has been checked all, if it is determined that the action candidates by checking the case C ₃ -C ₇ becomes all incompatible,
Resets (clears) the historical invalid information in the working memory M _4, Under the current priority since it is not possible to elect an appropriate action candidates, and resets the priority is reset in the priority degree After recovering each rule set consisting of rules with action candidates
Restart the action check inference from the beginning.

On the other hand, if applicable to the case C _2, a result of verification of all the rules in the set each rule, since supposed action candidates considered appropriate is stored in the work memory M ₅ is selected, to the action check The inference ends.

As described above, the rules in the rule set composed of rules having a certain shape change target in common are checked from the rule set (main area) sequentially after the validity of the action is checked. Stored in the evacuation area. Then, the rule check is executed until the rule set of the main area related to the shape change target becomes empty.
That is, if the state in which the rule set is empty does not correspond to the case C _1.

If two or more shape change targets are selected, the validity check of the above-mentioned action is executed for another set of shape change targets in the same manner as above (C _{1 to} C ₇ ). .

Thus, by the check routine, working inconsistency between action candidate memory M ₅ already about to be newly registered as action candidates registered, the priority, the efficacy results and the like are checked, target shape The validity of actions to be applied to data changes and the integrity of rules are maintained.

Following generation of the target shape, the action and the degree registered in the working memory M _5, namely "raising the end level, the degree 0.8"
Is transferred to the target shape generation unit 12.

The target shape generation unit 12 changes the values of the target shape adjustment parameters shown in Table 4 and FIGS. 12 (a) to 12 (d) based on the target shape change data. You. For example, when the above-mentioned action is “raise the end level, the degree is 0.8”, the value of the target shape adjustment parameter a ₁ related to the elongation rate of the end is determined according to the degree of the action. The change is set, and the target shape data changes. Further, the shape control unit 3 controls the coolant 58 of the roll mill 2 based on the input target shape data after the change.

Evaluation of applied action Then, operation condition data including the actual shape data newly obtained this time is input to the rolling data collection unit 7, and the same processing as the previous time is repeated. That is, the shape change target for the current actual shape and its importance are calculated by the control target generation unit 9 and compared with the previous one in the action candidate inference unit 11, respectively.

A specific example of repeating the above inference is shown in FIG. For example, some operating conditions reshaped target E _1, which is calculated from the actual shape data included in the data, when the importance level is "want strung quota" is 0.8, the first inference E ₂ It is executed, and the action candidate at that time is (A1) If the width of the zero is increased, (A2) the zero is moved outward. Here, the applied high-priority action A1, it before and after the target shape data are modified by were as shown in the inference result E _3.

However, when evaluating the effect of action A1, E ₁
Was not less than 0.8, and its effect was not observed. Therefore, the rule to derive the above action A1 is stored in the work memory M ₄ as was ineffective.

Then, apply the next highest priority action A2 to 2
Times eyes of inference E ₄ is executed. In this case, the rules including the shape change target E ₁ is continued subsequently applied by rolling process, if operating conditions are not the same as during the first inference, the action A1 and action A2 is obtained as a candidate. As a result of verifying the history of actions effect by inference E ₄ of the second, since the action A1 is stored in the work memory M ₄ as was ineffective last, but this time not applied (in Fig. 11 case C _4), the next highest priority action A2 is applied to the target shape data is changed.

As a result, if it is determined that the quota elongation is improved, the priority of the action A2 can be promoted and stored in the action inference knowledgebase D ₃ than that of the action A1.

In this way, the change of the actual shape (shape change target) and the change of the corresponding target shape data (action) when given target shape data with respect to the actual shape of the aluminum foil 53 are performed by the action candidate inference unit 11. Target shape data changed at each inference performed, and actual shape data obtained as a result, operating condition data used for setting the target shape data, rolling state data, or empirical values such as control target data. but it is stored in the action inference knowledgebase D ₃ as inference rules. Then, the action candidate inference unit 11 calculates appropriate target shape data based on the respective empirical values so as to realize an ideal actual shape (operating policy) at that time, and sets the target shape data as a target. It is automatically created via the shape generator 12 and output to the shape controller 3.

The aluminum foil rolling target shape adjusting device 1 as described above,
As shown in FIG. 14, the operation is started by the operator 5 tapping the key from the rolling mill terminal 6 (step 21). Subsequently, the operator 5 inputs the shape change target information together with the 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 transferred from the shape controller 3 (step 23), creates an appropriate target shape by inference (step 24), and displays the target shape before and after the correction (FIG. 16) on the screen. At the same time, the corrected target shape data is output to the roll mill 2 via the shape control unit 3 (step 25). Then, it was evaluated whether the corrected target shape was valid for the problematic actual shape (step 2).
6) Waiting for input by the operator 5.

At this time, the target shape adjusting device 1 also automatically generates a shape change target. If the target is inconsistent with the one input by the operator 5, the target having the higher importance is input by the operator 5. Alternatively, the shape change target may be generated based on a rule that gives priority to any one of the above and the one based on the shape determination result from the actual shape classification item.

The aluminum foil rolling target shape adjusting device 1 constantly monitors the actual shape of the aluminum foil 53, and when it is determined that the target shape data needs to be changed, the inference is started by the operator 5, Alternatively, the target shape data changing process can be started. For example, a comparison condition between the certainty factor of the actual shape classification item and a certain threshold value α is set in the condition part of the rule. A specific example of the rule will be described below.

If [the actual shape classification item and its certainty] and [operating condition], then [designation of the shape change target and its degree] More specifically, if the endurance certainty <α, If the path is the second pass, we want to extend the end, and the importance is expressed as 1.0. That is, when the actual shape changes and the degree of end extension becomes equal to or less than the threshold α, the above rule is applied,
Inference will begin.

Further, the aluminum foil rolling target shape adjusting device 1 has a function of improving even when the obtained actual shape data is asymmetric in the foil width direction. As shown in FIGS. 17 and 18, when it is determined that the actual shape data is asymmetric (step 31), the target shape, which was previously symmetric with respect to the portion having a low elongation, is determined by the operator. 5 is temporarily set high (step 32). This is a roll 52
This is for temporarily deflecting the distribution of the amount of coolant 58 injected into the roll width direction to uniformize the heat distribution in the rolling roll 52, particularly at the start of operation (during the temperature rise of the roll) or resumption. It is effective at the time (after the roll change). The process in step 32 is repeated until the actual shape data becomes symmetric.

As described above, the aluminum foil rolling target shape adjusting apparatus 1 according to the present embodiment sets the appropriate and timely target shape in accordance with the change even when the operating conditions of the roll rolling mill 2 slightly change. Is automatically started. Thereby, the aluminum foil 53 having a predetermined actual shape can be stably produced.

The aluminum foil rolling target shape adjusting device 1 employs an element 4 _e in which a piezoelectric element is embedded as a sensor for detecting actual shape data of elongation and tension at the time of rolling.
A plurality of air-bearing element to the element 4 _e and appearance substantially one substitute as the sensor, it is also possible to detect the actual shape data based on the change in the air pressure.

Further, the control symmetry is not limited to the aluminum foil 53, but may be copper or other metal, and the thickness is not limited.

〔The invention's effect〕

As described above, according to the present invention, in a system driving device that drives a control system or the like for a predetermined plant, a sensor group that detects plant data related to one or more types of plant states, and a storage that stores the plant data Means, a certainty factor calculating means for calculating the certainty factor for each plant state from the plant data, and comparing the certainty factor for each plant state with a first threshold value for each state.
Importance calculating means for selecting a plant state exceeding a threshold value and calculating the importance of a control target related to the plant state so as to be a monotonic change function with respect to the certainty factor; A composite importance is calculated to be a monotonic change function with respect to the importance, and the composite importance is calculated as a predetermined second value.
Since it is a system drive device characterized by comprising a control means for driving the control system or the like when the threshold value is exceeded, even if the operating state of the plant is slightly changed, according to the change, A control system or the like can be automatically driven appropriately and in a timely manner. Thereby, the operation state of the plant can be stably maintained.

In addition, since the calculation processing of the importance and the composite importance is restricted based on the first threshold, useless calculation is minimized, and the processing speed can be increased.

[Brief description of the drawings]

FIG. 1 is a schematic diagram showing a system arrangement of an aluminum foil rolling target shape adjusting device according to one embodiment of the present invention, FIG. 2 is a configuration diagram showing a processing flow of the aluminum foil rolling target shape adjusting device,
FIG. 3 (a) is an explanatory diagram showing the main parts of the actual shape data represented by the elongation distribution in the foil width direction, and FIG. 3 (b) is an explanation showing the actual shape classification items of the aluminum foil classified by pattern. FIG. 4 (a) is a flowchart showing a processing procedure for judging a change tendency of the actual shape data, and FIG. 4 (b) is an action for correcting the number of levels of the actual shape classification items in consideration of the shape change tendency. FIG. 5 (a) and FIG. 5 (b) are graphs showing the correlation between the number of levels of the two actual shape classification items, and FIG. 6 is a three-dimensional graph of the actual shape data of the aluminum foil over time. FIG. 7 is a schematic diagram conceptually showing a neural network, FIG. 8 is an explanatory diagram showing an example of a relationship between a shape change target for an actual shape classification item and an action candidate corresponding thereto, FIG. Figure is important for shape change target Is a graph showing the relationship between the confidence factor when the actual shape classification item is edged, and FIG. 10 shows an example of a rule used for inference by the action candidate inference unit and an example in which a target shape is changed using the rule. FIG. 11 is an explanatory view showing a processing procedure of a routine for checking the validity of an action to be applied by a check tree, and FIG. 12 (a) is a target shape adjustment used for changing a target shape. explanatory view showing parameters, FIG. (b) is a state diagram showing the state change of a ₃ of the parameter, and FIG. (c) the central portion of the target shape to be adjusted by a ₄ of the parameter is in the order pattern FIG. 14 (d) is a state diagram showing the situation, FIG. 13 (d) is a state diagram showing the situation where the central part has an inverted pattern, FIG. 13 is a schematic explanatory view showing an inference execution example for changing the target shape, and FIG. Shows the process flow of target shape adjustment Flow sheet, FIG. 15 is a display diagram showing an input menu displayed on the screen of the rolling mill-side terminal, FIG. 16 is a display diagram showing an example of the target shape displayed on the screen, and FIG. Flow sheet showing a method of correcting the actual shape that was given, FIG. 18 is a schematic explanatory diagram showing a situation of correcting the asymmetric real shape,
FIG. 19 is a schematic perspective view showing a roll rolling mill as an example of the background of the present invention, FIG. 20 is an external view showing a surface shape of an aluminum foil after rolling, and FIG. 21 is a sectional view of a rolling roll and aluminum foil. FIG. 22 is an explanatory diagram showing a correlation between the actual shape of the aluminum foil and a target shape for controlling the actual shape. FIG. 22 is a graph showing a target shape in operation of an aluminum foil and a target shape set for control at the same time. It is. [Reference Numerals] 1 ...... aluminum foil rolling target shape adjuster 2 ...... rolling mill 7 ...... rolling data acquisition unit (sampling means) M ₁ ...... working memory (sampling means) 8 ...... rolling status analyzing unit ( certainty calculating means) (plant state classifying means) (characteristic information calculation means) 9 ...... control target generator (importance calculating means) 11 ...... action candidate inference unit (control means) 53 ...... aluminum foil D ₁ ...... Rolling situation analysis knowledge base (pattern storage means) L ₁ ...... First threshold

──────────────────────────────────────────────────の Continued on the front page (51) Int.Cl. ⁶ Identification code FIG05B 9/02 B21B 37/00 116S 23/02 302 BBH (56) References JP-A-2-244203 (JP, A) JP Hei 1-243102 (JP, A) JP-A-63-163505 (JP, A) Toshio Sakai, et al., "Shape Control by Fuzzy Theory", Hitachi Review, Hitachi, Ltd., August 1989 Issued on the 25th, Volume 71, Issue 8, P.E. 103-108

Claims (3)

(57) [Claims] 1. A system driving apparatus for driving a control system or the like for a predetermined plant, a group of sensors for detecting plant data relating to one or more types of plant states, a storage unit for storing the plant data, A certainty factor calculating means for calculating the certainty factor for each plant state from the plant data; comparing the certainty factor for each plant state with a first threshold value for each state; selecting a plant state exceeding the first threshold value;
Importance calculating means for calculating the importance of the control target related to the plant state, which becomes a monotonic change function with respect to the certainty factor; A system driving apparatus comprising: a calculating means for calculating such a composite importance, and a control means for driving the control system or the like when the composite importance exceeds a second predetermined threshold value. 2. A system driving apparatus for driving a control system or the like for a predetermined plant, wherein: a sensor group for detecting plant data relating to one or more types of plant states; a storage unit for storing the plant data; Plant state classification means for classifying feature patterns of a plurality of plant states in advance from plant data; pattern storage means for storing the feature patterns of the plant state; and calculation of certainty factor for each feature pattern of each plant state from the plant data A confidence calculating means for comparing the certainty of each plant state feature pattern with a first threshold value for each pattern, selecting a plant state feature pattern that exceeds the first threshold value, Important to calculate the importance of the control target related to the feature pattern so as to be a monotonic change function Calculating means for calculating, from each of the importance levels, a composite importance level which becomes a monotonic change function with respect to the importance level, and driving the control system or the like when the composite importance level exceeds a predetermined second threshold value And a control means for causing the system to be driven. 3. A system driving device for driving a control system or the like for a predetermined plant, wherein a group of sensors for detecting plant data relating to one or more types of plant states, and periodically collecting and storing the plant data. Sampling means; characteristic information calculating means for calculating characteristic information of the plant state within a predetermined time from the plant data; and confidence calculating means for calculating certainty for each plant state according to the characteristic information of the plant state And comparing the certainty factor for each plant state with a first threshold value for each state, selecting a plant state that exceeds the first threshold value,
Importance calculating means for calculating the importance of the control target related to the plant state, which becomes a monotonic change function with respect to the certainty factor; A system driving apparatus comprising: a calculating means for calculating such a composite importance, and a control means for driving the control system or the like when the composite importance exceeds a second predetermined threshold value.