A new method for diagnosing shape quality of cold continuous rolling based on causal analysis

Abstract

The application discloses a new method for diagnosing shape quality of cold continuous rolling based on causal analysis, which comprises the following steps: reconstructing an attractor manifold according to two features in a same system to obtain a state space; reconstructing an attractor manifold again according to a feature time delay; finding a nearest neighbor point and calculating a nearest neighbor estimation; defining a CME score formula, calculating a CME score, judging whether the CME score converges to a constant or not, and if the CME score converges to a constant, it is indicated that a causal relationship exists between the two features. The application of the diagnosing method to diagnosing shape quality of cold continuous rolling can effectively monitor shape quality abnormality.

Description

Technical Field

[0001] This invention belongs to the field of cold-rolled strip steel, and in particular relates to a new method for diagnosing the shape and quality of cold continuous rolled strips based on causal analysis. Background Technology

[0002] As a crucial pillar of my country's national economy, the cold rolling industry primarily employs multi-stand continuous rolling. As a typical complex industrial process, it is characterized by multiple operating conditions, numerous variables, nonlinearity, and high-dimensional heterogeneous data, significantly increasing the technical challenges of cold rolling system modeling and simulation, quality monitoring, and anomaly diagnosis. Due to the complexity and continuity of the cold rolling process and control system, any anomaly in a single rolling stage will inevitably lead to subsystem malfunctions and even affect the overall product quality of the cold rolling production line. Therefore, anomaly diagnosis based on monitoring results and precise identification of the causes of anomalies can effectively reduce the anomaly rate in the rolling process, improve the quality of rolled products, and ensure the continuity of rolling production. However, current methods based on traditional mechanistic models are insufficient for process monitoring and anomaly diagnosis in the complex cold rolling industry. With continuous breakthroughs in computer data storage and processing technologies, the cost of data storage and analysis has been greatly reduced. Data-driven cold rolling process monitoring and anomaly diagnosis are flourishing, promoting the digital and intelligent transformation of the cold rolling industry and providing technical support for improving rolled product quality and rolling production safety.

[0003] The most important quality indicator in cold-rolled strip steel data is the flatness, which reflects the unevenness of the final product surface. The formula for calculating the flatness is shown below.

[0004]

[0005] Where i represents dividing the strip into i transverse sections, λ(i) represents the difference in elongation of the strip after the i-th section, ΔL represents the difference in length between the i-th strip and the reference strip, and L represents the reference length of the strip. IU value is usually used as the unit of elongation difference. The physical meaning of 1IU is that there is an elongation difference of 10μm per meter of the strip after the residual stress of the coil is eliminated. In actual production process, IU is often used to represent the shape deviation.

[0006] When an anomaly is detected during process monitoring, it is necessary to diagnose the anomaly, analyze its causes, and optimize accordingly to eliminate it. Anomaly diagnosis requires accurately identifying the specific cause of the anomaly within a highly coupled, multivariate production process. The description of this relationship is mainly divided into correlation and causation. Causation, compared to correlation, includes directionality over a time scale and more accurately reflects the internal logic of the actual mechanism model. Therefore, in the diagnosis of anomalies in the cold rolling industry process, causal diagnosis is more meaningful. Summary of the Invention

[0007] The purpose of this invention is to provide a new method for diagnosing the shape and quality of cold-rolled steel sheets based on causal analysis, so as to solve the problems existing in the prior art.

[0008] To achieve the above objectives, this invention provides a novel method for diagnosing the shape and quality of cold-rolled steel sheets based on causal analysis, comprising:

[0009] Step 1: Obtain two features of the cold continuous rolling nonlinear industrial system, obtain the discrete-time series of the two features, reconstruct the state space of the discrete-time series of the two features respectively, and obtain the first attractor manifold and the second attractor manifold.

[0010] Step 2: Obtain the possible time delays between features, and reconstruct an attractor flow based on the time delays to obtain a third attractor manifold. Reconstruct the third attractor manifold and the second attractor manifold according to each possible time delay.

[0011] Step 3: Calculate the nearest neighbor estimate for each point in the reconstructed second attractor manifold;

[0012] Step 4: Construct a cross-mapping evaluation score formula, calculate the cross-mapping score between two features according to the cross-mapping evaluation score formula, and determine whether there is a causal relationship between the two features and the strength of the causal relationship based on the cross-mapping score;

[0013] Step 5: Through the processes described in Steps 1 to 4, perform causal relationship analysis on several characteristics of the cold continuous rolling nonlinear industrial system to achieve intelligent anomaly diagnosis of the cold continuous rolling nonlinear industrial system.

[0014] Optionally, in step one, the time delay vector of the feature of any sequence in the discrete time series of the two features is calculated, and the time delay vector is used as a new embedding vector to complete the reconstruction of the state space and obtain the first attractor manifold and the second attractor manifold.

[0015] Optionally, the process of calculating the nearest neighbor estimate in step three includes: finding the K nearest neighbor points of each point in the reconstructed second attractor manifold and finding the point corresponding to the K nearest neighbor points in the third attractor manifold, and calculating the nearest neighbor estimate corresponding to each point based on the attractor.

[0016] Optionally, the formula for calculating the nearest neighbor estimate for each point in step three is as follows:

[0017]

[0018] Where z(t) j To reconstruct points in the third attractor manifold, M Y It is a second attractor manifold.

[0019] Optionally, in step four, a cross-mapping evaluation score formula is constructed based on the reconstructed attractor manifold and the calculated nearest neighbor estimate.

[0020] Optionally, the process of determining whether a causal relationship exists between the two features and the strength of that causal relationship in step four includes:

[0021] If the cross-mapping score converges to a constant, then a causal relationship exists. If a causal relationship exists, the strength of the causal relationship between the two features is determined based on the magnitude of the cross-mapping score.

[0022] The technical effects of this invention are as follows:

[0023] The method used in this invention is a nonlinear causal analysis method, which eliminates the limitation that Granger causality can only be used for stochastic linear decomposable systems; it can effectively analyze the causal relationship between various features with time delay, and fully distinguish the time delay effect and causality between different features; it can be used in complex industrial process systems with nonlinearity and feedback regulation characteristics to intelligently diagnose anomalies and identify the causes of anomalies; when applied to the diagnosis of cold continuous rolling sheet shape quality, it can effectively monitor sheet shape quality anomalies. Attached Figure Description

[0024] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0025] Figure 1 This is an image of the IU value of cold-rolled steel coil in an embodiment of the present invention;

[0026] Figure 2 This is a CME causal diagram of the back tension and IU value of the 5 racks in this embodiment of the invention;

[0027] Figure 3This is a graph showing the actual values ​​of the back tension and IU value of the 5 frames in this embodiment of the invention;

[0028] Figure 4 This is a CME causal diagram showing the relationship between strip speed and IU value in frames 4-5 of this invention.

[0029] Figure 5 This is a graph showing the actual values ​​of strip speed and IU value for frames 4-5 in this embodiment of the invention;

[0030] Figure 6 This is a flowchart of the method in an embodiment of the present invention. Detailed Implementation

[0031] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0032] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0033] Example 1

[0034] like Figure 1-6 As shown, this embodiment provides a novel method for diagnosing the shape and quality of cold-rolled steel sheets based on causal analysis, including:

[0035] The present invention aims to diagnose and analyze the sheet shape quality that occurs during the cold rolling industrial production process, and then establish a sheet shape quality diagnosis model based on causal analysis to diagnose sheet shape quality and analyze the causes of abnormalities, so as to detect and deal with problems in a timely manner and avoid interfering with the normal production process and product quality.

[0036] To achieve the above objectives, the present invention provides the following technical solution:

[0037] Step 1: Reconstruct the attractor manifold

[0038] According to Takens theory, in nonlinear dynamic systems, the reconstruction of a basic manifold based on time-delay embeddings only requires calculating the time delay vector of one of the time series features and using it as a new embedding vector to construct a new state space.

[0039] The state space of two discrete-time series X(t) and Y(t) of the same system is reconstructed, and the reconstructed attractor flow is shown in equations (1) and (2).

[0040] x(t)=<X(t),X(t-τ),X(t-2τ),L,X(t-(E-1)τ)> (1)

[0041] y(t)=<Y(t),Y(t-τ),Y(t-2τ),L,Y(t-(E-1)τ)> (2)

[0042] In the formula, L represents the length of the time series X(t) and Y(t); E represents the embedding dimension; and τ represents the unit time delay.

[0043] Where t∈[1+(E-1)τ,L], the set of x(t) is the attractor manifold M corresponding to X(t). X = {x(t)}, the attractor manifold corresponding to Y(t) is M. Y ={y(t)}.

[0044] Step 2: Reconstruct the attractor manifold based on the characteristic time delay

[0045] Reconstruct a third attractor manifold M Z ={z(t)}, where z(t)≡x(t-τ i ), τ i The possible time delays between features are Γ={τ1,τ2,L,τ m}

[0046] For each possible time delay, let z(t) = x(t-τ). i And reconstruct the attractor manifold M Z and M Y .

[0047] Step 3: Find the nearest neighbor and calculate the nearest neighbor estimate;

[0048] For M Y For each point, find the K nearest neighbors y(t) j (j=1,2,L,K) and find the corresponding z(t) j )∈M Z (j=1,2,L,K), then the corresponding nearest neighbor estimation is used. The specific calculation formula is shown in (3):

[0049]

[0050] Step 4: Define the CME scoring formula and calculate the CME score.

[0051] The CME score is defined as s(τ), and the specific calculation formula is shown in (4):

[0052]

[0053] In the formula n z The length of —z(t), i.e., the length of the third attractor flow; ∑ p —A diagonal matrix whose values ​​are the diagonal elements of cov(p,p).

[0054] Step 5: Determine the causal relationship between the two features based on the CME score;

[0055] When the CME score converges to a constant, it indicates that there is a causal relationship between the two features. The strength of the causal relationship between the two features is judged by the size of the CME score. Different features may have different convergence constants, but when the features are the same, a larger score indicates a stronger causal relationship.

[0056] The method provided by this invention was applied to the plate shape quality diagnosis of a thin plate cold rolling mill.

[0057] Step 1: Analyze the causes of anomalies based on actual cold continuous rolling production process of steel coils;

[0058] IU value of steel coil Figure 1 The data shows that the IU values ​​of the steel coil are significantly higher in the leading section and near the 300th sample point, which are identified as outliers. Figure 1 As shown.

[0059] The possible causes of the anomalies include: Variable 1: tension at the 5th stand; Variable 3: strip length; Variable 5: thickness at the 1st stand entrance; Variable 6: thickness at the 5th stand exit; and Variable 12: forward slip value at the 5th stand. Comparing these variables with the actual IU values, it is impossible to accurately determine which factors are the direct causes of the IU value anomalies. A CME anomaly diagnosis experiment was conducted using cold-rolled steel coil data, and the results are explained.

[0060] Step 2: Calculate the CME score, draw a cause-and-effect diagram, and diagnose the causes of the anomalies;

[0061] like Figure 2 The figure shows the causality of the tension after the 5th stand of the steel coil with respect to the IU value under different time delays. It can be seen from the figure that the causality of the tension after the 5th stand with respect to the IU value reaches its maximum when the time delay is 10 moments. Therefore, it can be concluded that after the sudden change in the tension after the 5th stand causes the abnormal IU value, its influence is transmitted to the strip shape and causes the strip IU value to become abnormal.

[0062] like Figure 3The figure shows the actual tension and IU value of the 5-rack back tension. It can be seen from the figure that the tension of the 5-rack back tension reaches its maximum at the 315th sampling point, and the actual IU value reaches its maximum at the 325th sampling point. The difference between the two is 10 sampling points, which is consistent with the results obtained by CME, thus verifying the validity of the CME results. The tension of the 5-rack back tension will cause the IU value to be abnormal 10 times after the abnormality.

[0063] like Figure 4 The figure shows the causality of the strip speed between stands 4 and 5 with the IU value under different time delays. It can be seen from the figure that the causality of the strip speed between stands 4 and 5 with the IU value reaches its maximum when the time delay is 10 moments. Therefore, it can be concluded that after the sudden change in the strip speed between stands 4 and 5 that causes the abnormal IU value occurs 17 moments later, its influence is transmitted to the strip shape and causes the strip IU value to become abnormal.

[0064] like Figure 5 The figure shows the actual strip speed and IU value between stands 4 and 5. It can be seen from the figure that the strip speed between stands 4 and 5 starts to change abruptly at the 302nd sampling point, and the actual IU value starts to show an anomaly at the 315th sampling point. The difference between the two is about 13 sampling points, which is similar to the result obtained by CME, thus verifying the validity of the CME result. The strip speed between stands 4 and 5 can cause the corresponding anomaly in the IU value at about 15 time points.

[0065] The above description is merely a preferred embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A novel method for diagnosing the shape and quality of cold-rolled steel sheets based on causal analysis, characterized in that, Includes the following steps: Step 1: Obtain two features of the cold continuous rolling nonlinear industrial system, obtain the discrete-time series of the two features, reconstruct the state space of the discrete-time series of the two features respectively, and obtain the first attractor manifold and the second attractor manifold. In step one, the time delay vector of the feature of any sequence in the discrete time series of the two features is calculated, and the time delay vector is used as a new embedding vector to complete the reconstruction of the state space and obtain the first attractor manifold and the second attractor manifold. Step 2: Obtain the possible time delays between features, and reconstruct an attractor flow based on the time delays to obtain a third attractor manifold. Reconstruct the third attractor manifold and the second attractor manifold according to each possible time delay. Step 3: Calculate the nearest neighbor estimate for each point in the reconstructed second attractor manifold; The process of calculating the nearest neighbor estimate in step three includes: finding each point in the reconstructed second attractor manifold. Find the nearest neighbor in the third attractor manifold and find the nearest neighbor with it. For each of the nearest neighbor points, the nearest neighbor estimate is obtained by calculating the attractor; The formula for calculating the nearest neighbor estimate for each point in step three is as follows: in, To reconstruct the points in the third attractor manifold, It is a second attractor manifold; Step 4: Construct a cross-mapping evaluation score formula, calculate the cross-mapping score between two features according to the cross-mapping evaluation score formula, and determine whether there is a causal relationship between the two features and the strength of the causal relationship based on the cross-mapping score; Define CME score as The specific calculation formula is as follows: ； In the formula for The length of the third attractor, i.e., the length of the flow of the third attractor; For A diagonal matrix whose diagonal elements are values; Estimating for nearest neighbors; Delay per unit of time; The process of determining whether there is a causal relationship between the two features and the strength of the causal relationship in step four includes: If the cross-mapping score converges to a constant, then a causal relationship exists. If a causal relationship exists, the strength of the causal relationship between the two features is determined based on the magnitude of the cross-mapping score. Step 5: Through the processes described in Steps 1 to 4, perform causal relationship analysis on several characteristics of the cold continuous rolling nonlinear industrial system to achieve intelligent anomaly diagnosis of the cold continuous rolling nonlinear industrial system.

2. The novel method for diagnosing the shape and quality of cold-rolled steel sheets based on causal analysis according to claim 1, characterized in that, In step four, a cross-mapping evaluation score formula is constructed based on the reconstructed attractor manifold and the calculated nearest neighbor estimate.

Images