JP2003211209A - Abnormality diagnostics for rolling mill - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To contrive speeding-up and improvement in accuracy in the estimation of abnormal factors for a coil part where the abnormality (defective quality) occurs. <P>SOLUTION: The diagnostics includes a data collecting part 102 for collecting values observed at the time of operation, a quality discriminating part 103 for deciding normal/defective condition of a plate quality on the collected data, an abnormal content discriminating part 104 which divides the entire coil length by areas in accordance with the state of plate speed and which specifies abnormal parts by detecting the presence/absence of abnormality for each area, a primary diagnostic part 105 which specifies abnormality signals by comparing with diagnostic threshold values set from a threshold value constructing part 107 for each feature amount by the observed signal values and statistical values, and a secondary diagnostic part 106 which estimates factors of each abnormal part by selecting the diagnostic logic of the abnormal factors assumed for each area. A threshold value adjusting part 108, in the case where normal/defective condition in the discrimination result by the quality discriminating part 103 is different from the diagnostic result by the diagnostic part 105 or 106, corrects the diagnostic threshold value by the latter so that the diagnostic accuracy is enhanced. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for diagnosing equipment failure and operation failure for a rolling mill constituted by multi-stage stands in a steel plant.

[0002]

2. Description of the Related Art In a rolling line such as a cold rolling mill in a steel plant, various failures such as equipment failure, material failure, and control failure, which are considered as factors for the occurrence of abnormalities such as plate thickness fluctuation and shape fluctuation, are considered Quickly extract the factors,
It is necessary to improve the yield. For this reason, conventionally, as described in JP-A-4-9214, after detecting a board defect, various operating data and equipment data are used to identify the cause of occurrence by a diagnostic logic such as an expert system. It was

[0003]

In the above-mentioned prior art, the location of abnormality in the entire coil length and the type of abnormality (local / whole area) are specified, and the abnormality factor estimation processing is performed for all areas. Therefore, the processing speed tends to be slow. In addition, since there is no consideration that the abnormal factors differ depending on the site, and all possible abnormal factors are searched for in the entire region, it is easy to include noise as compared with the diagnosis that limits the site, There was a problem that the diagnostic accuracy also decreased. Furthermore, when the characteristics of the target change due to aging, etc., the diagnostic threshold value used to derive the diagnostic result does not become an appropriate value that reflects the current situation, but there is no means to properly correct this. There was also the problem of a decrease in diagnostic accuracy due to a poor threshold.

An object of the present invention is to overcome the above problems of the prior art and to provide a high speed and highly accurate abnormality diagnosis method and apparatus for a rolling mill.

[0005]

Means for Solving the Problems The present invention for achieving the above-mentioned object is to judge the quality of a plate using operation data collected during rolling, and when the plate is judged to be defective, In a rolling mill abnormality diagnosis method that estimates anomaly factors, a region is divided according to the characteristics of the plate speed with respect to the total coil length, and changes in plate thickness and / or shape are detected for each region, and the regions with fluctuations are detected as abnormal. It is characterized in that it is specified as a part. The area is an FGC (near welding) part, a low speed part,
It is divided into an acceleration part, a high speed part and a deceleration part.

Further, for the abnormal portion, a state quantity observed at the time of operation or a statistical state quantity obtained by subjecting the state quantity to statistical processing and a diagnostic threshold value set for each state quantity or statistical state quantity. It is characterized in that a diagnosis for identifying an abnormal state quantity is performed by comparing The state quantity or statistical state quantity used in the "threshold calculation" for deriving the conclusion of presence or absence of abnormality by comparing with the threshold value will be referred to as a feature quantity below.

Further, it is diagnosed whether the abnormal portion is correct or not for each abnormal region. These abnormal factors include individual factors such as variations in the base material plate thickness, unevenness of the base material hardness, slip between the plate and roll, roll eccentricity, and sensor failure. It is characterized in that logic is selected and diagnostic processing is performed.

The diagnostic apparatus to which the method of the present invention can be applied is a quality judgment for judging the quality of a plate by using a data collecting section for collecting values observed during operation and the data collected by the data collecting section. And a diagnosis unit that estimates a factor of quality abnormality when the plate is determined to be defective by the quality determination unit, divides the region according to the state of plate speed in the entire coil length, and determines the abnormality of each region. It has an abnormality content determination unit that detects the presence or absence and identifies an abnormal site for each area, and the diagnosis unit targets the abnormal site identified by the abnormality content determination unit and makes a diagnosis using data related to the site. It is characterized by executing.

The diagnosis unit describes a diagnosis process associated with each part specified by the abnormality content determination unit, and selectively executes the diagnosis process corresponding to the part specified by the abnormality content determination.

Further, the abnormality diagnosing device has a threshold value adjusting unit, and when the result of the quality deciding unit is good and the result of the diagnosing unit outputs an abnormality, the abnormality deriving unit does not detect the abnormality. When the diagnostic threshold value is corrected to a direction in which an abnormality determination is unlikely to occur and the determination result of the quality determination unit is defective and the diagnostic result of the diagnostic unit outputs a normal result, the diagnostic threshold value and the magnitude relationship with it The diagnostic threshold value whose absolute difference from the threshold value for determining normality / abnormality is less than a certain value is corrected so that abnormality determination is likely to occur.

[0011]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will now be described in detail with reference to the drawings. FIG. 1 shows the configuration of an abnormality diagnosing device for a rolling mill according to an embodiment of the present invention. Abnormality diagnosis device 101
Is a data collection unit 102, a quality determination unit 103, an abnormality content determination unit 10
4, primary diagnosis unit 105, secondary diagnosis unit 106, threshold construction unit 10
7, a threshold adjustment unit 108 and a guidance unit 109.

The data collection unit 102 is, for example, a plant 100 which is a tandem cold rolling mill composed of five stands.
The operation data is collected from and is sent to the quality judgment unit 103.
The quality determination unit 103 determines the quality of this coil each time one coil is rolled, and determines the quality of the coil.

The abnormality content determination unit 104 divides the coil, which has been determined to be defective, into regions according to the plate speed state,
A region where an abnormality has occurred is extracted as an abnormal portion. The primary diagnosis unit 105 performs a statistical calculation such as an average value or a variance value on the directly detected values such as tension and load, targeting the abnormal portion extracted by the abnormality content determination unit 104. Further, those that are far apart from those of the normal coil are detected as abnormal, and the result is transmitted to the guidance unit 109. The secondary diagnosis unit 106 executes a process of specifically estimating a possible abnormality factor such as a base material plate thickness variation or roll eccentricity, targeting the abnormal portion extracted by the abnormality content determination unit 104, and the result To the guidance unit 109.

The threshold construction unit 107 constructs a diagnostic threshold to be used in the diagnostic processing of the primary diagnosis 105 and the secondary diagnosis 106, and a value for deriving a conclusion of presence / absence of abnormality by comparing with the diagnostic threshold. (This value is called a feature amount) is accumulated, and the threshold value is calculated using the average value and the variance value thereof.

The threshold adjustment unit 108 keeps the consistency between the determination result by the quality determination unit 103 and the diagnostic result by the secondary diagnosis unit 106, that is, when the determination result by the quality determination unit 103 is good, The diagnosis result of the secondary diagnosis unit 106 is “no abnormality”, and if the determination result is defective, the diagnosis threshold value is adjusted so that some variation factor is diagnosed as the cause of abnormality.

The guidance unit 109 converts the output results of the primary diagnostic unit 105 and the secondary diagnostic unit 106 into a file as a diagnostic log,
This is stored and is output to the display device 110 to notify the operator of the diagnosis result.

Next, the quality judgment unit 103, the abnormality content judgment unit 104, the primary diagnosis unit 105, which are the main parts of the diagnostic device of this embodiment,
Secondary diagnosis unit 106, threshold construction unit 107, and threshold adjustment unit
Details of the processing of 108 will be described.

The quality judgment unit 103 uses the operation data for one coil thickened by the data collection unit 102 to execute the processing according to the processing flow shown in FIG. In S3-1,
With respect to the evaluation standard of the plate prepared in advance, a value (quality 1 to 3) for determining this is calculated. In S3-2, Fig. 2
Scores are obtained for each item (evaluation standard) using the evaluation table 200 shown in FIG. In S3-3, the points of each item are totaled, and the final evaluation of the corresponding coil is A using the comprehensive evaluation table 201.
Determined by ~ D. A is the best, B and C are downgraded, D
Becomes a defective coil.

The abnormality content determination unit 104 basically focuses on the plate speed and divides the entire coil length into partial regions according to the state of the plate speed. That is, the data is divided into the low speed portion, the acceleration portion, the high speed portion, the deceleration portion, and the FGC (Flying Gauge Changing; near the welding portion) portion for management. When there are a plurality of areas, each divided area is managed as a separate divided area.

With respect to each of the divided areas, the variation of the plate and the shape is detected, and the area having the variation (abnormality) is specified as the abnormal portion. Also, due to this, the abnormal part is located at the front end, the rear end,
It is also possible to determine an abnormal state of whether it has occurred in the middle of the stomach, locally, or in the entire area.

FIG. 4 shows a processing flow of the abnormality content determination unit 104, and a specific processing flow will be described. In S4-1, the FGC portion is extracted from the welding point of the target coil to a fixed length in the front-rear direction or until feedback control (AGC; Automatic Gauge Control) is started. In S4-2, the moving differential value v _md of the exit side plate speed v of the final stand is calculated.

For this v _md , a plurality of state judgment thresholds (acceleration state judgment threshold: th _{ac c} , deceleration state judgment threshold: th _dcc , constant speed state judgment threshold: th
_cnst ) is used to extract the constant velocity area, acceleration area, and deceleration area. In S4-4, high / low judgment threshold (t
h _LH ) is used to divide the constant velocity region into a high speed part and a low speed part. In S4-5, the variation of the final stand delivery side plate thickness and the variation of the shape are detected for each of the extracted regions, and the region having the variation is specified as the variation region.

FIG. 5 shows the structure of the primary diagnosis unit. The primary diagnosis unit 105 includes a characteristic amount calculation unit 500, a characteristic amount table 501, a primary diagnosis execution unit 502, and a threshold value table 503.

The primary diagnosis unit 105 uses the feature quantity table 5 of FIG.
As shown in 01, the operator compares the statistical values (average, variance, maximum moving variance, etc.) of the values normally observed by the operator, such as plate thickness, tension, and load, to those Determine normal / abnormal. In the primary diagnosis unit 105, these statistical values correspond to feature quantities. The threshold value table 503 stores the threshold value V _th of each characteristic amount obtained by accumulating the characteristic amounts of a plurality of normal coils. here,
Assuming the average value μ of the characteristic amount, the standard deviation σ of the characteristic amount, and the coefficient α for obtaining the threshold value, V _th is calculated by the equation (1).

[0025]

[Equation 1]

FIG. 7 shows a processing flow of the primary diagnosis section 105. In S7-1, the feature amount (name) corresponding to the variable region extracted by the abnormality content determination unit 104 is extracted from the feature amount table 501. In S7-2, the value of the extracted feature quantity (name) is calculated. For example, when the FGC section is selected as the variable region, the feature amount with FGC in the feature amount table 501 is selected, and the actual values such as tension and load are used to calculate the average, variance,
Calculate the maximum value of moving variance. In S7-3, the threshold value V _th corresponding to the calculated feature amount is set in the threshold value table.
Extract from 503. In S7-4, a normal / abnormal determination is made for each feature amount calculated in S7-2, using the threshold value _Vth obtained in S7-3. That is, a certain feature amount is a threshold value V _th
If it exceeds, the feature amount becomes abnormal, and if it does not exceed, it becomes normal. Hereinafter, the process of comparing the feature amount with the diagnostic threshold value and concluding normality / abnormality by exceeding / not exceeding the threshold value will be referred to as "threshold value calculation".

FIG. 8 shows the configuration of the secondary diagnosis section. The secondary diagnosis unit 106 includes a region-variation factor table 800 and a factor diagnosis group 80.
1, a factor diagnosis selection unit 802, a threshold value table 803, and a detailed information calculation unit 804. The secondary diagnosis unit 106 prepares a factor diagnosis logic 805 that estimates the possibility of the factor for each assumed variation factor, and selectively executes this to investigate the variation factor.

The area-variation factor table 800 records the areas and the corresponding relationships between the factor diagnoses occurring in each area. The factor diagnosis group 801 is composed of the factor diagnosis logic 805 prepared by the number of assumed variable factors. The threshold value table 803 stores threshold values used by each factor diagnosis logic 805 to execute diagnosis, and the factor diagnosis logic 805 refers to the threshold value table 803 to perform diagnosis.

FIG. 9 shows a processing flow of the secondary diagnosis section 106. In S9-1, the factor diagnosis selection unit 802 extracts the factor diagnosis logic corresponding to the variation region specified by the abnormality content determination unit 104 by referring to the region-variation factor table 800.

FIG. 10 shows an example of the area-variation factor table 800. The table 800 stores the variation factors for each area. For example, there are two fluctuation factors in the low speed region, namely, the base material plate thickness and hardness unevenness, and there are five fluctuation factors in the high speed region: base material plate thickness, hardness unevenness, roll eccentricity, slip and defective detector.

In S9-2, the selected factor diagnosis logic is selected from the factor diagnosis group 801 and activated. In S9-3, the activated factor diagnosis logic 805 refers to the threshold value table 803 to execute the factor diagnosis.

FIG. 11 shows an example of the threshold table 803. In the threshold value table 803, the threshold value obtained from the equation (1) is set for each area and factor. For example, F
The preform thickness variation, which is one factor of GC unit, threshold VH_ the feature quantity (moving variance value of the base plate thickness) V _Hmax
th is a feature value (maximum value of frequency spectrum of base metal plate thickness)
The threshold value P_th is stored for each Pmax.

In S9-4, when the result diagnosed by the factor diagnosis logic 805 is abnormal, the detailed information calculation unit 804 adds detailed information such as an abnormal portion and an abnormal actuator and transmits it to the guidance unit 109.

As described above, in the primary diagnosis unit 105, the operator is informed of abnormalities in operation data such as tension and load that he or she is usually accustomed to, and in addition to this, the secondary diagnosis unit 106 specifies the assumed fluctuation factor name. By making such a notification, it is possible to perform a multi-faceted diagnosis such as the verification of the secondary diagnosis result using the primary diagnosis result, and it is possible to provide the operator with significant information that can improve the diagnostic accuracy.

Next, a specific example of the factor diagnosis logic 805 will be described. In this example, the base material plate thickness variation diagnosis logic when the base material plate thickness variation is a factor, the hardness unevenness abnormality diagnosis logic when the base material hardness unevenness is a factor, and the slip between the roll and the plate is a factor. Slip diagnosis logic when it becomes, roll eccentricity diagnosis logic when roll eccentricity is a factor,
A total of five detector failure diagnosis logics that are caused by either the plate thickness gauge or the plate speed gauge will be described.

FIG. 12 shows a processing schematic diagram of the base material plate thickness variation diagnosis logic, and FIG. 13 shows a processing flow. The variation of the base metal plate thickness is a variation factor over almost the entire region. This logic will be explained according to the processing flow. In S13-1,
The moving dispersion value V _h of the final stand exit side plate thickness h is given by the equation (2),
It is calculated using the equation (3).

[0037]

[Equation 2]

Further, the maximum value V _Hmax of V _h is calculated.

[0039] In S13-2, the search of the base material plate moving variance V _Hmax thick portions corresponding, the maximum value of which is calculated in the same manner as S13-1. In S13-3, frequency conversion of the plate thickness on the delivery side of the final stand is performed, and n points of frequencies at which the power spectrum exceeds the threshold value gth1 are extracted. In S13-4, the maximum value of the frequency spectrum of the base material plate thickness existing in the vicinity of the frequency width w extracted in S13-3 is determined, and this is set as P _max .

[0040] In S13-5, the magnitude of V _Hmax and P _max, and a threshold operation using a threshold stored in the threshold value table 803, normal that each result to an OR operation / Determine the anomaly. That is, if either one is abnormal, it is determined to be abnormal.

By the above processing, by observing the behavior when the base material variation is the cause from both the time domain and the frequency domain, it is possible to estimate the base material variation factor with high accuracy.

FIG. 14 shows a schematic diagram of the hardness unevenness abnormality diagnosis logic, and FIG. 15 shows a processing flow. This logic uses the ratio of the calculated load to the measured load (hereinafter Zp) to determine the hardness unevenness.
When there is no unevenness in hardness, Zp is constant. Therefore, it is diagnosed that the region where Zp fluctuates is the region where unevenness in hardness occurs.

In S15-1, the moving variance value V _h of the final stand outlet side plate thickness is calculated by the equation (2), and the maximum value V _h is calculated.
_{Find Hmax} . In S15-2, Zp of the site corresponding to the V _Hmax
The moving variance value of is _calculated, and the maximum value V _{zpmax thereof} is _calculated . S1
In 5-3, the threshold value of V _zpmax is calculated, and the conclusion of normality / abnormality is derived. According to this logic, it is possible to estimate the hardness unevenness defect with high accuracy by using Zp that can noticeably observe the hardness behavior.

FIG. 16 shows a schematic diagram of the slip diagnosis logic, and FIG. 17 shows a processing flow. In this logic, the measured advance rate fa is used to perform a slip diagnosis by extracting a phenomenon in which fa, which is a value in the vicinity of 1, is rapidly changed in the negative direction.

[0045] In S17-1, similarly to the base material thickness variation detect logic, the final stand delivery calculates the moving variance of the side thickness, obtains this maximum value V _Hmax. In S17-2, it is confirmed whether or not fa is such that fa <0 in the range of the width w of the region corresponding to V _Hmax . If it exists, the process moves to S17-3, and if not, the process ends. In S17-3,
The moving average of fa _mv is calculated by equation (4), and the absolute difference between the maximum value and the minimum value of fa _mv within the width w of the part corresponding to V _Hmax (P
/ P) is calculated.

[0046]

[Equation 3]

At S17-4, a threshold value is calculated using the P / P calculated at S17-3 to derive a conclusion. According to this logic, the fa that corresponds to the slip phenomenon on a one-to-one basis is used, and the varying fa is smoothed by the moving average, so that the qualitative behavior of this can be accurately detected, resulting in highly accurate slip. Anomalies can be estimated.

FIG. 18 shows a schematic diagram of the roll eccentricity diagnosis logic, and FIG. 19 shows a processing flow. In S19-1, the final stand output side plate thickness is frequency converted. In S19-2, the maximum power spectrum (Pmax (i) i; stand No.) in the vicinity of the rotation frequency w is extracted for each roll. In S19-3, a threshold value is calculated for Pmax, and a roll exceeding the threshold value is specified as an eccentric roll. The threshold value at this time is set by the threshold value construction unit 107, and is calculated using Pmax in the normal state as a feature amount as described later. Threshold construction section
By using the threshold value appropriately set by 107, it is possible to detect the eccentric roll with high accuracy.

FIG. 20 shows a processing flow of the detector failure diagnosis logic. This diagnostic logic focuses on the property that the mass flow value (= plate thickness x plate speed x plate width) is constant at each stand in a constant speed state in order to detect a defect in the plate thickness meter or plate speed meter, Observe the mass flow value of each stand in the high speed part of the plate speed, and derive the conclusion that the detector (plate thickness gauge or plate speed meter) of the stand whose mass flow value is far apart from other stands is abnormal. To do.

First, in S20-1, the movement dispersion value of the final stand delivery side plate thickness is calculated in the same manner as the base material plate thickness variation diagnosis logic, and the maximum value V _Hmax thereof is obtained. For S20-2, V
_Using the plate speed and the outlet plate thickness near the time w corresponding to _Hmax , the average mass flow value xi of each stand xi (I; stand N
o. ) Is calculated. Here, a constant value is used as the plate width.

At S20-3, the average value E and the standard deviation σ of the mass flow values of the stands other than the stand i are calculated using the equations (5) and (6).

[0052]

[Equation 4]

At S20-4, it is determined whether the mass flow value of the stand i satisfies the expression (7).

[0054]

[Equation 5]

If the result of S20-4 is Yes in S20-5, it is determined that the mass flow value of the i-stand is normal, that is, the detector is normal. If No, it is determined that the i-stand detector is abnormal.

At S20-6, it is determined whether or not all the stands have ended. If all the stands have ended, the process ends. If not, the process moves to S20-3, and the same process is executed at another stand.

According to this logic, the stand having the abnormality detector can be detected by paying attention to the disturbance of the mass flow value, and the degree of separation of the i stand mass flow value from other stands can be calculated by the equation (5), ( By using the equation (6), it is possible to estimate the detector failure with high accuracy.

FIG. 21 shows the structure of the threshold construction unit. The threshold construction unit 107 includes a coil A determination detection unit 2100, a feature amount calculation unit 2101, a threshold construction data updating unit 2102, a threshold construction data table 2103, and a threshold table updating unit 2104. It Data table for threshold construction
Reference numeral 2103 denotes an average value μ of feature quantities used in threshold calculation in the primary diagnosis unit 105 and the secondary diagnosis unit 106, a standard deviation σ, and the number of coils N used for calculating these values for each threshold value. Is stored.

FIG. 22 shows a processing flow of the threshold construction unit 107. Startup processing is performed for each coil. In S22-1, coil A
The determination detection unit 1900 confirms the determination result of the quality determination unit 103. In S22-2, it is checked whether or not the determination result is the A determination, and if it is the A determination, the process proceeds to S22-3. If not, the process ends.

In S22-3, the feature quantity calculation unit 2101 calculates the feature quantity used for the threshold calculation in the primary diagnosis unit 105 and the secondary diagnosis unit 106. In S22-4, the threshold construction data updating unit 2102 uses the feature amount calculated in S22-2, the average value Ave (N) of each feature amount stored in the threshold construction data table 2103, Standard deviation Stddev (N) (8)
Equation (9) is updated. Here, Data (N) is the feature amount calculated in S22-2.

[0061]

[Equation 6]

In S22-5, whether there is an operator's instruction,
Alternatively, when the number of coils N is N _max or more, the threshold value table updating unit 2104 causes the threshold value construction data table 21 to be displayed.
Using the value of 03, the threshold value V _th is calculated by the equation (10), and by overwriting this value, the threshold table 503 of the primary diagnostic unit 105 and the threshold table of the secondary diagnostic unit 106 80
Update 3 respectively.

[0063]

[Equation 7]

Here, for β, an appropriate value around 3 is selected. According to the threshold construction unit 107, since the average value μ and the standard deviation σ are sequentially updated by the equations (8) and (9), it is not necessary to store the value of the feature quantity for the number of coils, and it is stored. It is possible to construct a threshold that efficiently uses the area.

FIG. 23 shows the configuration of the threshold correction unit 108.
The threshold value correction unit 108 is a coil determination detection unit 2300, a diagnosis execution unit 2301, a correction threshold value calculation unit 2302, a non-reaction detection unit 2303,
Correction target diagnostic factor extraction unit 2304, correction threshold value calculation unit 2305
And a threshold table updating unit 2306.

FIG. 24 shows a processing flow of the threshold correction unit 108, and the details of the processing will be described according to the flow. S24-1
Then, in the coil determination detection unit 2300, the quality determination unit 103
Check the judgment result of. In S24-2, it is checked whether the determination result is A or D. If the determination is A, the process proceeds to S24-3. If the result is D, the process proceeds to S24-5, and if neither, the process ends.

In S24-3, the diagnostic execution unit 2301 is the secondary diagnostic unit.
The factor diagnosis logic 805 that started 106 and executed diagnosis for all areas and judged the result of A as abnormal.
To extract. In S24-4, the threshold value of the factor diagnosis logic 805 extracted in S24-3 is corrected.
Is corrected in the direction that makes it difficult to react (direction in which it is difficult to make an abnormality judgment).

On the other hand, in the processing S24-5 for the D judgment, it is verified whether or not the diagnosis result of the secondary diagnosis section 106 is normal. If it is normal, the process proceeds to S24-6, and if it is not normal, the process ends. In S24-6, the correction target diagnosis factor extraction unit 2304 causes the factor diagnosis logic that satisfies the expression (11).
Extract 805.

[0069]

[Equation 8]

Here, γ is the judgment coefficient, the feature amount is the value used for the threshold value calculation, and Stddev is the standard deviation used when the threshold value is constructed.

In S24-7, the threshold value of the factor diagnostic lock 805 extracted in S24-6 is set to the corrected threshold value calculation unit 2305.
Correct the direction in which is easy to react (direction in which it is easy to make an abnormality judgment). In S24-8, the threshold value table updating unit 2306
Writes the corrected threshold to the threshold table 803 and updates it.

In the above flow, the correction procedure for the threshold value of the secondary diagnosis unit 106 is shown.
By handling the feature amount described in the feature amount table 501 instead of 805, the threshold value adjustment of the primary diagnosis unit 105 can be performed similarly.

[0073]

According to the rolling mill abnormality diagnosis apparatus of the present invention,
An abnormal part can be specified for each region classified according to the plate speed state, and a factor diagnosis adapted to the region can be selectively executed for the part, so that high-speed and highly accurate diagnosis is possible.

Further, since the diagnostic threshold value is appropriately adjusted so as to maintain the consistency between the quality judgment result of the plate (coil) and the diagnostic result, the appropriate threshold value is maintained even when the characteristic of the target is changed, and is constantly high. Accurate diagnosis can be performed.

[Brief description of drawings]

FIG. 1 is an overall configuration diagram of a rolling mill abnormality diagnosis apparatus according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram of a quality judgment evaluation table.

FIG. 3 is a processing flowchart of the quality determination unit 103.

FIG. 4 is a processing flowchart of an abnormality content determination unit 104.

FIG. 5 is a configuration diagram of a primary diagnosis unit 105.

FIG. 6 is an explanatory diagram of a feature amount table 501.

FIG. 7 is a processing flow chart of the primary diagnosis unit 105.

FIG. 8 is a configuration diagram of a secondary diagnosis unit 106.

FIG. 9 is a processing flowchart of the secondary diagnosis unit 106.

FIG. 10 is an explanatory diagram of a region-variation factor table 800.

FIG. 11 is an explanatory diagram of a threshold table 803.

FIG. 12 is a schematic diagram of a base material plate thickness variation diagnosis logic.

FIG. 13 is a processing flowchart of a base material plate thickness variation diagnosis lock.

FIG. 14 is a schematic diagram of a hardness unevenness abnormality diagnosis logic.

FIG. 15 is a processing flowchart of a hardness unevenness abnormality diagnosis logic.

FIG. 16 is a schematic diagram of a slip diagnosis logic.

FIG. 17 is a processing flow chart of slip diagnosis logic.

FIG. 18 is a schematic diagram of a roll eccentricity diagnosis logic.

FIG. 19 is a processing flowchart of a roll eccentricity diagnosis logic.

FIG. 20 is a processing flowchart of a detector failure diagnosis logic.

FIG. 21 is a configuration diagram of a threshold construction unit 107.

FIG. 22 is a processing flowchart of the threshold construction unit 107.

FIG. 23 is a configuration diagram of a threshold correction unit 108.

FIG. 24 is a processing flow chart of the threshold correction unit 108.

[Explanation of symbols]

100 ... Plant (tandem cold rolling mill), 101 ... Abnormality diagnosis device, 102 ... Data collection unit, 103 ... Quality judgment unit, 104 ... Abnormality content judgment unit, 105 ... Primary diagnosis unit, 106 ... Secondary diagnosis unit, 10
7 ... Threshold building unit, 108 ... Threshold adjusting unit, 109 ... Guidance unit, 110 ... Display device, 200 ... Evaluation table, 201 ...
Overall evaluation table, 500 ... Feature amount calculation unit, 501 ... Feature amount table, 502 ... Primary diagnosis execution unit, 503 ... Threshold table
(For primary), 800 ... Area-variation factor table, 801 ... Factor diagnosis group, 802 ... Factor diagnosis selection unit, 803 ... Threshold table
(For secondary), 804 ... Detailed information calculation unit, 805 ... Factor diagnosis logic, 2100 ... Coil A determination unit, 2101 ... Feature amount calculation unit, 2102
… Threshold data updating unit, 2103… Threshold data table, 2104… Threshold table updating unit, 2300
... coil determination detection unit, 2301 ... diagnosis execution unit (secondary diagnosis unit activation unit), 2302 ... correction threshold calculation unit, 2303 ... non-reaction detection unit, 2304 ... correction target diagnostic factor extraction unit, 2305 ... correction threshold Value calculator, 2306 ... Threshold table update unit.

   ─────────────────────────────────────────────────── ─── Continued front page    (72) Inventor Yoichi Sugita             7-1-1, Omika-cho, Hitachi-shi, Ibaraki Prefecture             Inside the Hitachi Research Laboratory, Hitachi Ltd. (72) Inventor Masahiro Kayama             5-2-1 Omika-cho, Hitachi City, Ibaraki Prefecture             Ceremony company Hitachi Ltd. Omika factory (72) Inventor Kenichi Yoshioka             5-2-1 Omika-cho, Hitachi City, Ibaraki Prefecture             Inside the Hitachi Information Control System (72) Inventor Naoki Hirota             5-2-1 Omika-cho, Hitachi City, Ibaraki Prefecture             Inside the Hitachi Information Control System (72) Inventor Kim Yong Nam             South Korea Jeollanam-do Gwangho-dong, Gwangyang             Address 700 Pohang Sogo Steel Co., Ltd. (72) Inventor Kakho Chul             South Korea Jeollanam-do Gwangho-dong, Gwangyang             Address 700 Pohang Sogo Steel Co., Ltd. (72) Inventor issue             South Korea Jeollanam-do Gwangho-dong, Gwangyang             Address 700 Pohang Sogo Steel Co., Ltd. (72) Inventor Shin Sung Cap             South Korea, Jeollanam-do             Address 700 Pohang Sogo Steel Co., Ltd. F-term (reference) 4E024 GG10                 5H223 AA20 BB05 CC01 EE06

Claims (5)

[Claims] 1. An abnormality diagnosis for estimating quality of a plate by using collected operation data during rolling of a rolling mill having a plurality of stands and estimating an abnormal factor when the plate is determined to be defective. In the method, the variation of the base material sheet thickness corresponding to the variation part of the final stand delivery side sheet thickness is observed, and when the base material sheet thickness is different from the normal degree of variation, and / or the final stand delivery side sheet thickness and the Abnormality of the rolling mill characterized by comparing the frequency components of the base material plate thickness, and if they differ from the normal fluctuation degree, it is estimated that the fluctuation of the base material plate thickness is an abnormal factor of poor quality. Diagnostic method. 2. An abnormality diagnosis for estimating quality of a plate by using collected operation data during rolling of a rolling mill having a plurality of stands and estimating an abnormal factor when the plate is determined to be defective. In the method, the zp value, which is the ratio of the calculated load and the actual load, is calculated. A method for diagnosing abnormalities in a rolling mill, which is characterized by presuming that it is an abnormal factor of poor quality. 3. An abnormality diagnosis for estimating quality of a plate by using collected operation data during rolling of a rolling mill having a plurality of stands, and estimating a cause of abnormality when the plate is determined to be defective. In the method, when the moving average fa of the measured advanced ratio is calculated, and the behavior of the fa corresponding to the variation part of the final stand delivery side plate thickness changes in the negative direction and the fa becomes negative, A method for diagnosing a rolling mill abnormality, characterized in that slip between a plate and a roll is estimated to be an abnormal factor of poor quality. 4. An abnormality diagnosis for estimating quality of a plate by using collected operation data during rolling of a rolling mill having a plurality of stands, and estimating an abnormality factor when the plate is determined to be defective. In the method, the frequency component f of the final stand delivery side plate thickness is obtained, the maximum value fmax of the frequency component f in the vicinity of the rotation frequency of each roll is obtained, and the maximum value fmax of the frequency component f of the plate having a normal size An abnormality diagnosing method for a rolling mill, comprising estimating that the eccentricity of the roll is an abnormality factor of poor quality when the size (threshold value) is exceeded. 5. An abnormality diagnosis for estimating quality of a plate by using collected operation data during rolling of a rolling mill having a plurality of stands and estimating an abnormality factor when the plate is determined to be defective. In the method, the mass flow value given by the product of the plate thickness detected by the plate thickness gauge, the plate speed detected by the plate speed gauge, and the plate width is calculated for each stand, and the relative mass flow value of the other stands is calculated. When the mass flow value indicating a different value is detected, it is estimated that the strip thickness gauge and / or the strip speed meter of the stand having the detected mass flow value is estimated to be defective. Abnormality diagnosis method.