A big data driven method for abnormal thickness control performance backtracking in rolling process

Abstract

The present invention provides a big data driven rolling process thickness control performance abnormality backtracking method, including: selecting benchmark data from historical production data of the same number of passes and plate types according to the similarity of reduction distribution; analyzing the passes with degraded performance using the angle contribution value method based on the selected benchmark data to find the abnormal loop candidate set; performing transfer entropy analysis between loops according to the time series data of multiple candidate abnormal loops, and calculating the transfer entropy values ​​between each abnormal loop; establishing a cause-and-effect relationship diagram according to the calculated transfer entropy values, and locating the root cause of the abnormality according to the cause-and-effect relationship diagram. The technical solution of the present invention can not only be applied to the thick plate finishing process, but also can be extended to other complex industrial processes.

Description

Technical Field

[0001] The present invention relates to the technical field of control performance evaluation in a steel thick plate rolling process, and in particular to a big data driven rolling process thickness control performance abnormality backtracking method. Background Art

[0002] The finishing rolling process is the process that has the greatest impact on the thickness of thick plates in the entire process of thick plate rolling. This process is a control process with a large number of loops, including rolling force, bending roll force, roll gap, roll speed, etc. When an abnormality occurs, the abnormal oscillation often causes abnormalities in various variables in the loop, resulting in multiple abnormal alarms, which makes it impossible for on-site personnel to handle it immediately. Due to the coupling of multiple loops in the process, the nonlinearity introduced by the servo valve failure and other factors causes oscillations in multiple variables and loops of the control system, making it difficult to analyze the source of the abnormality. Therefore, when product quality is abnormal or equipment control accuracy decreases during the production process, it is difficult to effectively determine which specific equipment is in poor condition or the control loop performance is degraded. Therefore, in order to find the root cause of the abnormality, it is necessary to establish an abnormal root cause tracing model for the finishing mill.

[0003] However, the existing abnormality backtracking methods in China often require experts to check the circuits and devices one by one based on fault experience, which requires downtime processing, has a great impact on the on-site production line, and often takes several hours or days, resulting in product quality degradation or production interruption. Therefore, the present invention aims to provide a data-driven rolling process thickness control abnormality backtracking method based on angle contribution value and transfer entropy, realizes root cause tracing through abnormality causal graph, and assists on-site personnel to quickly locate abnormalities.

[0004] The current domestic patent technologies are as follows: The patent "A rolling process fault diagnosis method based on unbalanced data" trains the DBN classification model through historical data and labels, and can classify the rolling process data into balanced data and unbalanced data, thereby obtaining unbalanced (i.e. fault) data. The patent "An IPCA rolling process online fault diagnosis method with variable control limits" uses the incremental principal component analysis method to calculate the control limits based on historical normal data, and uses the control limits to perform fault diagnosis on real-time online data. The above technologies have the following shortcomings:

[0005] (1) Both require manual screening of modeling data from historical data in advance;

[0006] (2) The cause of the fault cannot be diagnosed, that is, the root cause cannot be traced. Summary of the invention

[0007] According to the technical problems raised above, the present invention provides a big data driven rolling process thickness control performance abnormality backtracking method. The present invention first proposes a benchmark data selection method based on the steel plate thickness reduction distribution for the steel rolling process; then the angle contribution value is combined with the transfer entropy method to trace the root cause of the abnormal loop.

[0008] The technical means adopted by the present invention are as follows:

[0009] A big data driven rolling process thickness control performance abnormality backtracking method comprises the following steps:

[0010] S1. Select benchmark data;

[0011] S2. Based on the selected benchmark data, the angle contribution value method is used to analyze the passes with degraded performance and find the abnormal circuit candidate set;

[0012] S3. According to the actual operation value of the abnormal circuit, the transfer entropy analysis between the circuits is performed to calculate the transfer entropy value between each abnormal circuit;

[0013] S4. A causal relationship diagram is established based on the calculated transfer entropy value, and the root cause of the abnormality is located based on the causal relationship diagram.

[0014] Furthermore, in step S1, the selected benchmark data is specifically:

[0015] According to the similarity of reduction distribution, the benchmark data is selected from the production data with the same number of passes and plate type in the historical data.

[0016] Furthermore, according to the similarity of the reduction distribution, the benchmark data is selected from the production data of the same number of passes and plate type in the historical data, specifically:

[0017] S11, let the thickness reduction amount of the data to be diagnosed be:

[0018] D II =[d II,1 ,…,d II,o ]

[0019] Among them, o is the number of channels of process data to be diagnosed, d II,i is the thickness reduction of the i-th pass of the data to be diagnosed;

[0020] S12. Select data from the historical database whose thickness results meet the requirements, whose plate shape quality is excellent, whose number of passes is consistent with the steel type and the data to be diagnosed, as the data to be selected;

[0021] S13, defining the thickness reduction similarity between the selected data and the diagnosed data as follows:

[0022]

[0023] Among them, d i It represents the reduction amount of the ith pass of any thick plate in the selected data;

[0024] S14. Select the historical thick plate data with the smallest similarity S value of thick plate reduction as the benchmark data, which is used to evaluate the control performance of each loop of the process data to be diagnosed.

[0025] Furthermore, the specific implementation process of step S2 is as follows:

[0026] S21. Based on the selected benchmark data, there is the following covariance matrix:

[0027] cov(y Ⅱ )P=cov(y Ⅰ )PΛ

[0028] Among them, y Ⅱ represents normal data, y Ⅰ Indicates monitoring data;

[0029] S22. Definition The performance-degraded subspace is composed of the column vectors in P corresponding to the values ​​of Λ in the diagonal matrix that are greater than 1. The angle-based contribution value calculation formula is as follows:

[0030]

[0031] Among them, ||·|| represents the 2-norm of the vector, e k =[0...0 k-1 1 0...0] T is a unit vector with the kth row being 1 and the rest being 0, l represents a subspace with worse performance The dimension of The number of columns, express The kth row vector of , the above equation explains that the angle-based contribution index is completely determined by the load; compared to the subspace with poor performance If the angle-based contribution index cosθ k >ε r , the corresponding loop / variable can be identified as the contribution to the worse subspace; where ε r It is a manually set threshold parameter, usually 0.707.

[0032] Furthermore, the specific implementation process of step S3 is as follows:

[0033] S31. Consider two continuous random variables x and y, each with N samples, that is, x i ∈[x1,x2,…,xN ],y i ∈[y1,y2,…,y N ], let y i+h represents the value of variable y at time i+h, that is, h steps into the future from time i, where h is called the prediction horizon and p(·) represents the joint probability density function, which is calculated as follows:

[0034] p(x,y)=p(x)p(y)

[0035] Among them, x and y represent two random independent signals, through the embedding vector, It can capture the temporal dynamics of x and y;

[0036] S32. According to the Bayesian principle, the transition probability is defined as:

[0037]

[0038] where p(·|·) represents the past value x i and i When the future value x is known i+h A probability with a definite value;

[0039] S33, Order and Respectively represent the embedding vectors using the historical values ​​of y and x; k represents the embedding dimension of y, and l represents the embedding dimension of x; τ represents the time interval allowed for the embedding vector in time scaling, and h = τ ≤ 4 is set as the estimation method. τ is generally selected as 1, and the actual size can be adjusted according to the sampling rate;

[0040] S34. Based on step S33, the calculation formula of the transfer entropy from x to y is as follows:

[0041]

[0042] Among them, in the selection of k and l parameters, it is necessary to calculate t for k = 1,...10, l = 1,...10 respectively. x→y , select t x→y The k and l at the maximum value are used as parameter values, and t at this time x→y For the desired result.

[0043] Furthermore, the specific implementation process of step S4 is as follows:

[0044] S41, if the transfer entropy value t calculated in step S3 x→y When ≥ε, it is considered that x is the cause of y. Generally, ε is taken as 0.08. The adjacency matrix of each pass is obtained according to the transfer entropy value of each variable.

[0045] S42, establishing a cause-effect relationship graph according to the adjacency matrix;

[0046] S43. Find the root cause of the anomaly based on the flow of the cause-effect diagram.

[0047] Compared with the prior art, the present invention has the following advantages:

[0048] 1. The big data driven rolling process thickness control performance abnormality tracing method provided by the present invention first proposes a benchmark data selection method based on the distribution of steel plate thickness reduction for the steel rolling process; then the angle contribution value is combined with the transfer entropy method to determine whether the overall performance of multiple control loops has declined, and the root cause of the abnormal loop is traced. The method of the present invention is not only applied to the thick plate finishing rolling process, but also can trace the root cause of the abnormality.

[0049] 2. The big data-driven rolling process thickness control performance abnormality backtracking method provided by the present invention does not require pre-training or calculation of historical data.

[0050] 3. The big data-driven rolling process thickness control performance abnormality backtracking method provided by the present invention can solve the problem that traditional traceability methods require expert experience and reduce downtime costs.

[0051] 4. The big data-driven rolling process thickness control performance abnormality backtracking method provided by the present invention can solve the problem that the root cause of the abnormal loop cannot be determined.

[0052] Based on the above reasons, the present invention can be widely promoted in the fields of control performance evaluation during the rolling process of thick steel plates. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0054] Figure 1 The figure is a flow chart of the method of the present invention.

[0055] Figure 2 Contribution diagram of each loop provided in an embodiment of the present invention.

[0056] Figure 3 A cause-and-effect diagram of an abnormal loop provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0057] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0058] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The following description of at least one exemplary embodiment is actually only illustrative and is by no means intended to limit the present invention and its application or use. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0059] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0060] Unless otherwise specifically stated, the relative arrangement of the parts and steps described in these embodiments, the numerical expressions and numerical values ​​do not limit the scope of the present invention. At the same time, it should be clear that, for ease of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship. The technology, methods and equipment known to ordinary technicians in the relevant field may not be discussed in detail, but in appropriate cases, the technology, methods and equipment should be regarded as part of the authorization specification. In all examples shown and discussed here, any specific value should be interpreted as merely exemplary, rather than as a limitation. Therefore, other examples of exemplary embodiments may have different values. It should be noted that similar numbers and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further discussed in subsequent drawings.

[0061] In the description of the present invention, it is necessary to understand that the directions or positional relationships indicated by directional words such as "front, back, up, down, left, right", "lateral, vertical, perpendicular, horizontal" and "top, bottom" are usually based on the directions or positional relationships shown in the drawings. They are only for the convenience of describing the present invention and simplifying the description. Unless otherwise specified, these directional words do not indicate or imply that the device or element referred to must have a specific direction or be constructed and operated in a specific direction. Therefore, they cannot be understood as limiting the scope of protection of the present invention: the directional words "inside and outside" refer to the inside and outside relative to the contours of each component itself.

[0062] For ease of description, spatially relative terms such as "above", "above", "on the upper surface of", "above", etc. may be used here to describe the spatial positional relationship between a device or feature and other devices or features as shown in the figure. It should be understood that spatially relative terms are intended to include different orientations of the device in use or operation in addition to the orientation described in the figure. For example, if the device in the accompanying drawings is inverted, the device described as "above other devices or structures" or "above other devices or structures" will be positioned as "below other devices or structures" or "below their position devices or structures". Thus, the exemplary term "above" can include both "above" and "below". The device can also be positioned in other different ways (rotated 90 degrees or in other orientations), and the spatially relative descriptions used here are interpreted accordingly.

[0063] In addition, it should be noted that the use of terms such as "first" and "second" to limit components is only for the convenience of distinguishing the corresponding components. If not otherwise stated, the above terms have no special meaning and therefore cannot be understood as limiting the scope of protection of the present invention.

[0064] The present invention provides a big data driven rolling process thickness control performance abnormality backtracking method, comprising the following steps:

[0065] S1. Select benchmark data;

[0066] In specific implementation, as a preferred embodiment of the present invention, in step S1, the reference data is selected as follows:

[0067] According to the similarity of reduction distribution, the benchmark data is selected from the production data with the same number of passes and plate type in the historical data.

[0068] S11, let the thickness reduction amount of the data to be diagnosed be:

[0069] D II =[d II,1 ,…,d II,o ]

[0070] Among them, o is the number of channels of process data to be diagnosed, d II,i is the thickness reduction of the i-th pass of the data to be diagnosed;

[0071] S12. Select data from the historical database whose thickness results meet the requirements, whose plate shape quality is excellent, whose number of passes is consistent with the steel type and the data to be diagnosed, as the data to be selected;

[0072] S13, defining the thickness reduction similarity between the selected data and the diagnosed data as follows:

[0073]

[0074] Among them, d i It represents the reduction amount of the ith pass of any thick plate in the selected data;

[0075] S14. Select the historical thick plate data with the smallest similarity S value of thick plate reduction as the benchmark data, which is used to evaluate the control performance of each loop of the process data to be diagnosed.

[0076] S2. Based on the selected benchmark data, the angle contribution method is used to analyze the passes with degraded performance and find abnormal circuits;

[0077] In specific implementation, as a preferred embodiment of the present invention, the specific implementation process of step S2 is as follows:

[0078] S21. Based on the selected benchmark data, there is the following covariance matrix:

[0079] cov(y Ⅱ )P=cov(y Ⅰ )PΛ

[0080] Among them, y Ⅱ represents normal data, y Ⅰ Indicates monitoring data;

[0081] S22. Definition The performance-degraded subspace is composed of the column vectors in P corresponding to the values ​​of Λ in the diagonal matrix that are greater than 1. The angle-based contribution value calculation formula is as follows:

[0082]

[0083] Among them, ||·|| represents the 2-norm of the vector, e k =[0...0 k-1 1 0...0] T is a unit vector with the kth row being 1 and the rest being 0, l represents a subspace with worse performance The dimension of The number of columns, express The kth row vector of , the above equation explains that the angle-based contribution index is completely determined by the load; compared to the subspace with poor performance If the angle-based contribution index cosθ k >ε r , the corresponding loop / variable can be identified as the contribution to the worse subspace; where ε r It is a manually set threshold parameter, usually 0.707.

[0084] S3. According to the actual operation value of the abnormal circuit, the transfer entropy analysis between the circuits is performed to calculate the transfer entropy value between each abnormal circuit;

[0085] In specific implementation, as a preferred embodiment of the present invention, the specific implementation process of step S3 is as follows:

[0086] S31. Consider two continuous random variables x and y, each with N samples, that is, x i ∈[x1,x2,…,x N ],y i ∈[y1,y2,…,y N ], let y i+h represents the value of variable y at time i+h, that is, h steps into the future from time i, where h is called the prediction horizon and p(·) represents the joint probability density function, which is calculated as follows:

[0087] p(x,y)=p(x)p(y)

[0088] Among them, x and y represent two random independent signals, through the embedding vector, It can capture the temporal dynamics of x and y;

[0089] S32. According to the Bayesian principle, the transition probability is defined as:

[0090]

[0091] where p(·|·) represents the past value x i and i When the future value x is known i+h A probability with a definite value;

[0092] S33, Order and Respectively represent the embedding vectors using the historical values ​​of y and x; k represents the embedding dimension of y, and l represents the embedding dimension of x; τ represents the time interval allowed for the embedding vector in time scaling, and h = τ ≤ 4 is set as the estimation method. τ is generally selected as 1, and the actual size can be adjusted according to the sampling rate;

[0093] S34. Based on step S33, the calculation formula of the transfer entropy from x to y is as follows:

[0094]

[0095] Among them, in the selection of k and l parameters, it is necessary to calculate t for k = 1,...10, l = 1,...10 respectively. x→y , select t x→y The k and l at the maximum value are used as parameter values, and t at this time x→y For the desired result.

[0096] S4. A causal relationship diagram is established based on the calculated transfer entropy value, and the root cause of the abnormality is located based on the causal relationship diagram.

[0097] In specific implementation, as a preferred embodiment of the present invention, the specific implementation process of step S4 is as follows:

[0098] S41, if the transfer entropy value t calculated in step S3 x→y When ≥ε, it is considered that x is the cause of y. Generally, ε is taken as 0.08. The adjacency matrix of each pass is obtained according to the transfer entropy value of each variable.

[0099] S42, establishing a cause-effect relationship graph according to the adjacency matrix;

[0100] S43. Find the root cause of the anomaly based on the flow of the cause-effect diagram.

[0101] Example

[0102] The embodiment of the present invention is the operating data of the finishing mill of a large steel enterprise. The sampling period of the sensors related to the equipment is 0.004 seconds. According to the field experience and the previous mechanism analysis, 9 key circuits are selected as the experimental variables of the method, namely: rolling force, hydraulic cylinder transmission side position, hydraulic cylinder operation side position, No. 1 servo valve opening, roll gap, roll gap offset, bending roll force, upper roll speed and lower roll speed. The selected variables are all related to the thickness control of the steel plate, which is also the focus of the rolling process. The system running the program is Win10 system, the CPU is i7-7700HQ 2.8GHz, and the program's operating environment is matlab2016a and Anocanda3.

[0103] Figure 1 A flowchart of a rolling process abnormality backtracking method based on angle contribution value and transfer entropy provided by an embodiment of the present invention is shown, which specifically includes the following steps:

[0104] Step 1: Select benchmark data.

[0105] In this embodiment, the case selected is that the thickness result of the rolled steel plate is abnormal during a certain rolling process. This case contains 8 passes, and 10 historical normal data (all containing 8 passes, the plate type is B4 and the abnormal steel plate is consistent) are selected as the candidate data. The similarity between the candidate data and the data to be diagnosed is calculated according to the similarity formula:

[0106]

[0107] Among them, the similarities between the 10 selected data and the data to be diagnosed are 0.0183, 0.0113, 0.0103, 0.0086, 0.0045, 0.0049, 0.0100, 0.0052, 0.0063, and 0.0108, respectively. The data with the closest similarity, that is, the thick plate case with a reduction similarity of 0.0045, is selected as the benchmark data.

[0108] Step 2: Select the first pass as the data to be diagnosed, and use the angle contribution value-based method to analyze the pass with degraded performance and find the abnormal circuit. The angle-based contribution value calculation formula is as follows:

[0109]

[0110] Among them, ||·|| represents the 2-norm of the vector, e k =[0...0 k-1 1 0...0] T is a unit vector with the kth row being 1 and the rest being 0, l represents a subspace with worse performance The dimension of The number of columns, express The kth row vector of , the above equation explains that the angle-based contribution index is completely determined by the load; compared to the subspace with poor performance If the angle-based contribution index cosθ k >ε r , the corresponding loop / variable can be identified as the contribution to the worse subspace; where ε r is a manually set threshold parameter, usually 0.707. Figure 2 As shown, the angle contribution diagram result of the process to be diagnosed is calculated; according to the manually set threshold parameters, it can be further seen that the performance of the performance degradation loop in the performance degradation pass, that is, the loop exceeding the dotted line.

[0111] Step 2: According to the actual operation value of the abnormal circuit, conduct transfer entropy analysis between circuits, calculate the transfer entropy value between each abnormal circuit, and establish a cause-effect relationship diagram based on the transfer entropy result. The calculation formula of the transfer entropy value is as follows:

[0112]

[0113] Each pass will generate an adjacency matrix. The value in the adjacency matrix is ​​greater than the set threshold ε, which is 0.1 (selected based on experience). A directed graph is established for the adjacency matrix generated by each pass, that is, the final causal relationship graph, as shown in Figure 3 As shown. From a process perspective, the performance of x2 (hydraulic cylinder transmission side position) is inconsistent with the control performance of x3 (hydraulic cylinder operation side position), which can easily lead to abnormal control systems. Since x4 is the servo valve opening, which directly operates the hydraulic cylinder position, x3 and x4 (servo valve opening) affect each other; x3, x4, and x8 (upper roller speed) all have an impact on x9 (lower roller speed). The above three circuits x3, x4, and x8 can be regarded as the root causes of abnormalities.

[0114] The effectiveness of the method of the present invention is reflected in that, in relatively complex situations, the number of loops to be checked is reduced from 9 at the beginning to 4 through positioning based on angle contribution values, and then reduced to 3 through causal analysis based on transfer entropy. The operating efficiency of the method of the present invention is as follows: When there are 4 abnormal loops, it is necessary to perform causal analysis on (4-1)*4=12 pairs of loops, and diagnose that the abnormal loop is x4 (servo valve opening), which is caused by the performance degradation of the servo valve, which is consistent with the actual on-site investigation results, and effectively traces back to the root cause of the abnormality; the time required is about 20 minutes, while manual investigation usually takes several days or hours, which effectively improves the timeliness of root cause tracing. The positioning of the performance degradation loops of 10 groups of thickness abnormal cases was verified and statistically analyzed, and the final diagnostic accuracy was 86.4%, which reflects the effectiveness of this patent.

[0115] It can be seen from the above embodiments that for the retrospective diagnosis of abnormal performance of multi-loop control of the finishing mill, the method of the present invention first finds the benchmark data through a method based on the similarity of the pressing amount, and then performs an angle-based contribution diagram analysis on the benchmark data to further obtain the performance degradation loop, and finally performs a causal analysis based on the transfer entropy on the loop with performance degradation to obtain a causal relationship diagram and find the root cause of the performance degradation.

[0116] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A big data driven rolling process thickness control performance abnormality backtracking method, characterized in that: The steps include: S1. Select benchmark data; S2. Based on the selected benchmark data, the angle contribution value method is used to analyze the passes with degraded performance and find the abnormal circuit candidate set; the specific implementation process of step S2 is as follows: S21. Based on the selected benchmark data, there is the following covariance matrix: those(y Ⅱ )P=cov(y Ⅰ )PΛ Among them, y Ⅱ represents normal data, y Ⅰ Indicates monitoring data; S22. Definition The performance-degraded subspace is composed of the column vectors in P corresponding to the values ​​of Λ in the diagonal matrix that are greater than 1. The angle-based contribution value calculation formula is as follows: Among them, ||·|| represents the 2-norm of the vector, e k =[0...0 k-1 1 0...0] T is a unit vector with the kth row being 1 and the rest being 0, l represents a subspace with worse performance The dimension of The number of columns, express The kth row vector of , the above equation explains that the angle-based contribution index is completely determined by the load; compared to the subspace with poor performance If the angle-based contribution index cosθ k >ε r , the corresponding loop / variable can be identified as the contribution to the worse subspace; where ε r The threshold parameters are manually set; S3. According to the actual operation value of the abnormal circuit, the transfer entropy analysis between the circuits is performed to calculate the transfer entropy value between each abnormal circuit; S4. A causal relationship diagram is established based on the calculated transfer entropy value, and the root cause of the abnormality is located based on the causal relationship diagram.

2. The big data driven rolling process thickness control abnormality backtracking method according to claim 1 is characterized in that: In step S1, the selected benchmark data is specifically: According to the similarity of reduction distribution, the benchmark data is selected from the historical production data with the same number of passes and plate type.

3. The big data driven rolling process thickness control performance abnormality backtracking method according to claim 2 is characterized in that: According to the similarity of the reduction distribution, the benchmark data is selected from the production data with the same number of passes and plate type in the historical data, specifically: S11, let the thickness reduction amount of the data to be diagnosed be: D II =[d II,1 ,…,d II,o ] Among them, o is the number of channels of process data to be diagnosed, d II,i is the thickness reduction of the i-th pass of the data to be diagnosed; S12. Select data from the historical database whose thickness results meet the requirements, whose plate shape quality is excellent, whose number of passes is consistent with the steel type and the data to be diagnosed, as the data to be selected; S13, defining the thickness reduction similarity between the selected data and the diagnosed data as follows: Among them, d i It represents the reduction amount of the ith pass of any thick plate in the selected data; S14. Select the historical thick plate data with the smallest similarity S value of thick plate reduction as the benchmark data, which is used to evaluate the control performance of each loop of the process data to be diagnosed.

4. The big data driven rolling process thickness control performance abnormality backtracking method according to claim 1 is characterized in that: The specific implementation process of step S3 is as follows: S31. Consider two continuous random variables x and y, each with N samples, that is, x i ∈[x1,x2,…,x N ],y i ∈[y1,y2,…,y N ], let y i+h represents the value of variable y at time i+h, that is, h steps into the future from time i, where h is called the prediction horizon and p(·) represents the joint probability density function, which is calculated as follows: p(x,y)=p(x)p(y) Among them, x and y represent two random independent signals, through the embedding vector, It can capture the temporal dynamics of x and y; S32. According to the Bayesian principle, the transition probability is defined as: where p(·|·) represents the past value x i and i When the future value x is known i+h a probability with a definite value; S33, Order and denote the embedding vectors using the historical values ​​of y and x respectively; k denotes the embedding dimension of y, l denotes the embedding dimension of x; τ denotes the time interval allowing the embedding vector to scale in time, and h = τ ≤ 4 is set as the estimation method; S34. Based on step S33, the calculation formula of the transfer entropy from x to y is as follows: Among them, in the selection of k and l parameters, it is necessary to calculate t for k = 1, ... 10, l = 1, ... 10 respectively. x→y , select t x→y The k and l at the maximum value are used as parameter values, and t at this time x→y For the desired result.

5. The big data driven rolling process thickness control performance abnormality backtracking method according to claim 1 is characterized in that: The specific implementation process of step S4 is as follows: S41, if the transfer entropy value t calculated in step S3 x→y >ε, then x is considered to be the cause of y, and the adjacency matrix of each pass is obtained according to the transfer entropy value of each variable; S42, establishing a cause-effect relationship graph according to the adjacency matrix; S43. Find the root cause of the anomaly based on the flow of the cause-effect diagram.

Images