Intelligent deviation rectifying method and system for strip steel cutting

Abstract

The invention relates to the field of data processing, in particular to an intelligent deviation correction method and system for strip steel tailoring, and the method comprises the steps: firstly obtaining and smoothing deviation contour data of strip steel in real time; then, by calculating the local variation energy of the second derivative of the deviation profile, identifying the region with the form having significant change, and further calculating the feature scale of the region; then, dynamically generating a self-adaptive analysis window according to the feature scale of the macroscopic deviation; and finally, utilizing low-order polynomial regression in the window to accurately decouple the complex deviation form into basic parameters such as translation, inclination, C-shaped bending and S-shaped bending, and transmitting the basic parameters to a controller so as to realize rapid and accurate differential deviation correction. According to the method, the macroscopic deviation is identified by calculating the local variation energy and the feature scale, the adaptive analysis window is constructed, and the deviation form is decoupled by using local polynomial regression, so that accurate control is realized.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and particularly to an intelligent deviation correction method and system for strip steel cutting. Background Art

[0002] In fields such as steel and automotive manufacturing, the continuous production and processing of strip steel are core processes. During the transportation of strip steel to the cutting station, due to various factors such as upstream processes, mechanical vibrations, or internal stresses of materials, the strip steel often undergoes lateral position deviations, i.e., the so-called "running deviation" phenomenon. This running deviation has a complex form and may include multiple components such as overall translation, tilt, C-shaped bend, or S-shaped bend. If these deviations cannot be accurately identified and corrected in a timely manner, it will seriously affect the cutting accuracy, resulting in a decline in product quality and material waste. Therefore, it is crucial to perform real-time, accurate online monitoring and intelligent deviation correction control on the running deviation form of strip steel.

[0003] Currently, frequency domain analysis methods such as Fourier transform (FFT) are often used in the industry to process the data collected by sensors. This method can identify the periodic components existing in the deviation by performing spectral analysis on the deviation data within the entire observation window. However, in the specific scenario of strip steel intelligent deviation correction, the traditional FFT method has obvious defects. First, the macroscopic deviations of strip steel, such as large C-shaped bends or S-shaped bends, are usually non-periodic and localized events. As a global analysis tool, FFT is difficult to accurately locate the specific positions where these deviations occur and lacks spatial resolution ability. Second, in order to accurately analyze long waveform deviations, FFT requires a very long sampling window, which will introduce a large computational delay, conflicting with the requirement of the cutting station for control real-time performance. Finally, the components obtained by FFT analysis are in the frequency domain, and their physical meanings are not intuitive enough, making it difficult to directly decouple them into specific form parameters such as translation, tilt, and bend required by the controller, bringing difficulties to subsequent precise deviation correction control. Summary of the Invention

[0004] To address the contradiction between the significant computational delay and the real-time control requirements of the cutting station, this invention proposes an intelligent deviation correction method for strip cutting in a first aspect. The method includes: acquiring the real-time deviation profile of the strip along its length upstream of the cutting station; calculating the local variation energy of the deviation profile, wherein the local variation energy is positively correlated with the variance of the second derivative of the deviation profile within a local window of a set length; locating the energy peaks in the local variation energy curve along the length direction and calculating the characteristic scale of each energy peak; wherein the characteristic scale is positively correlated with the width of the corresponding energy peak at a set threshold height; filtering out macroscopic deviations whose characteristic scales are greater than a preset scale threshold; constructing an adaptive analysis window centered on the peak position and with a width proportional to the characteristic scale for each macroscopic deviation; performing polynomial fitting on the deviation profile within the adaptive analysis window to obtain fitting coefficients; and transmitting the fitting coefficients to a deviation correction controller to execute a deviation correction action.

[0005] Compared to existing technologies that typically employ fixed windows or simple feedback control based solely on deviation values, this invention, by calculating the local variation energy related to the second derivative of the deviation profile, can more sensitively identify complex shape defects such as S-shaped bends that are difficult to detect using traditional methods. Furthermore, by screening macroscopic deviations with large feature scales and constructing an adaptive analysis window with matching dimensions, the focus of analysis and correction can be precisely concentrated on the truly large and complex deviations that need to be addressed, avoiding ineffective or excessive responses to minute, high-frequency noise. This adaptive analysis method ensures the accuracy of subsequent polynomial fitting, thereby generating more precise correction control parameters and significantly improving the ability to identify and correct complex strip shape defects.

[0006] Furthermore, obtaining the real-time deviation profile includes: deploying multiple displacement sensors upstream of the cutting station to collect lateral deviation data points of the strip edge; and performing digital filtering on the lateral deviation data points to obtain a smooth and continuous deviation profile function. .

[0007] Furthermore, the local variation energy The specific calculation method is as follows: ; in This represents the deviation profile function; Indicates the location The second derivative at point; This indicates the length of the local window; Display window The arithmetic mean of the inner second derivatives.

[0008] By defining the local variation energy as the variance of the second derivative of the deviation profile within a local window, this invention provides an effective mathematical method for quantifying the severity of local bending in strip steel. Compared to traditional methods that only use deviation values ​​or first derivatives, the variance of the second derivative is more sensitive to changes in curvature such as C-bends and S-bends.

[0009] Furthermore, the feature scale The specific calculation method is as follows: ; in Indicates the location of the energy peak; Indicates position Local variation energy at the location; Represents the Herveside step function; This represents a threshold coefficient with a value between 0.2 and 0.8.

[0010] By integrating energy curves exceeding a specific energy threshold, this method can quantitatively assess the impact range of each deviation region along the strip length. Unlike existing technologies that simply treat all deviations equally, this characteristic scale calculation allows the system to distinguish between long, gentle bends and short, sharp bends, providing crucial information for identifying macroscopic deviations that require focused attention.

[0011] Furthermore, the range of the adaptive analysis window is: ,in The peak position of the macroscopic deviation, half the window width. The specific calculation method is as follows: ;in This indicates a proportionality coefficient greater than 1; This represents the characteristic scale associated with the peak location.

[0012] By making the half-width of the analysis window proportional to the feature scale of the identified deviation, intelligent and adaptive adjustment of the analysis area is achieved. Compared with the conventional approach of using a fixed-size window for analysis, the adaptive window of this invention ensures that the analysis range exactly and completely covers the entire macroscopic deviation region, greatly improving the accuracy and specificity of subsequent polynomial fitting.

[0013] Furthermore, the proportionality coefficient The value range is 1.2 to 2.0.

[0014] Furthermore, the polynomial fitting is a third-order polynomial regression performed using the least squares method, specifically: ; in This represents the deviation profile after fitting. This represents the peak position of the macroscopic deviation; The fitting coefficient is denoted as .

[0015] Compared to first-order models that can only describe translation and tilt, or second-order models that can only describe C-shaped bends, third-order polynomials can simultaneously and accurately describe various typical shape defects of strip steel, such as translation, tilt, C-shaped bends, and S-shaped bends. This makes the mathematical model of this invention highly compatible with the physical deformation characteristics of strip steel, enabling it to capture the complete form of macroscopic deviations more comprehensively and accurately, thus laying the foundation for generating comprehensive correction instructions.

[0016] Furthermore, in the equation obtained by the third-order polynomial regression, the zero-order, first-order, second-order, and third-order coefficients correspond to the translation, tilt, C-shaped bending degree, and S-shaped bending degree of the strip, respectively.

[0017] Furthermore, after receiving the fitting coefficients, the correction controller drives the correction roller actuator to perform a compensation action to correct the macroscopic deviation.

[0018] In a second aspect, the present invention provides an intelligent correction system for strip cutting, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, an intelligent correction method for strip cutting of the present invention is implemented.

[0019] The technical effects of this invention are as follows: This invention first effectively identifies complex shape deviations such as S-bends by calculating the variance of the second derivative of the deviation profile. Second, it employs an adaptive analysis window matched to the deviation characteristic scale for local data analysis, overcoming the limitations of traditional fixed-window analysis. Finally, it uses a third-order polynomial to accurately fit the selected macroscopic deviations, directly mapping the fitting coefficients to physical deformations such as translation, tilting, C-bends, and S-bends. This method can accurately diagnose and quantify complex deviations, achieving intelligent and high-precision correction of strip steel shape defects. Attached Figure Description

[0020] Figure 1 This is a schematic flowchart illustrating an intelligent correction method for strip cutting according to an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the analysis curve of the local variation energy calculation process in an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the energy peak detection and screening method in an embodiment of the present invention. Figure 4 This is a schematic diagram illustrating the local morphological decoupling fitting results of an embodiment of the present invention; Figure 5 This is a schematic diagram illustrating the structure of an intelligent correction system for strip cutting according to an embodiment of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0023] An example of an intelligent correction method for strip cutting: like Figure 1 As shown, the intelligent correction method for strip cutting according to the present invention includes: S1. Real-time acquisition of multi-point lateral deviation of strip steel, and preprocessing of data smoothing and noise suppression using filtering algorithm.

[0024] In this embodiment, multiple non-contact displacement measurement sensors need to be deployed at equal or non-equal intervals upstream of the cutting station on the strip conveyor line, along the strip running direction. For example, laser displacement sensors or linear arrays can be used. Cameras. These sensors are used to acquire real-time data from one or both edges of the strip at multiple preset monitoring points. Lateral deviation at the location ,in , This represents the total number of sensors. In this way, a set of discrete data points describing the current position and shape of the strip can be obtained. .

[0025] It is understandable that the raw data points may contain measurement noise introduced by factors such as electrical noise, mechanical vibration, or changes in ambient light. To improve the accuracy and robustness of subsequent analysis, preprocessing of this discrete data point set is necessary. Specifically, a digital filtering algorithm can be used to smooth the data. As a preferred approach, this embodiment employs a Gaussian filter or a median filter. The Gaussian filter effectively suppresses Gaussian white noise through weighted averaging, while the median filter effectively removes impulse noise while preserving edge information relatively well. After filtering, a smooth and continuous real-time deviation profile function is obtained within a finite, continuously scrolling observation window. This is used to provide a high-quality data foundation for subsequent analysis.

[0026] S2. Calculate the energy variance of the second derivative of the deviation profile within the sliding window, and construct a local variation energy curve to quantify the degree of drastic spatial change in the strip shape.

[0027] In the steps Obtain a smooth deviation profile function Next, this step aims to identify ROI regions in the deviation profile where the morphology changes significantly, while ignoring parts with relatively flat or uniform morphological changes. To this end, this embodiment introduces a local variation energy index to quantify the local morphological complexity or degree of curvature of the deviation profile at each location point.

[0028] Its construction is based on the deviation profile function. The second derivative It can intuitively represent its local curvature. When the shape of the strip changes, for example from a straight line to a curve, or... At the inflection point of a bend, its curvature changes drastically. Therefore, by calculating the energy of the second derivative within a local sliding window, the degree of anomaly in the strip morphology within that region can be reliably measured. Specifically: ; in Indicates the location Local variation energy at the location; Steps The deviation profile function obtained after preprocessing; Indicates the deviation profile at position The second derivative at that point, which in a physical sense corresponds to the local curvature at that point; It is a preset, relatively small sliding integration window length to ensure computational locality. As a preferred approach, this window length... It can be set according to the sensor spacing and the expected minimum disturbance wavelength, for example, it can be set to arrive The average sensor spacing is times that of the average sensor spacing; Indicates in window The arithmetic mean of the inner second derivatives.

[0029] According to the above formula, when the deviation profile is in position The surrounding area has a complex morphology, for example, there are The inflection point of the bend or At the apex of the bend, its curvature Dramatic changes can occur, resulting in a large variance within that window, which in turn affects the calculated local variation energy. It exhibits a relatively high peak value; conversely, if the strip is nearly straight or inclined at a uniform speed in this region, its curvature is constant or nearly constant, and the variance of the second derivative approaches zero. The value will be very low. Therefore, through analysis The distribution along the length of the strip can effectively locate the key areas where the strip shape undergoes significant changes.

[0030] like Figure 2 As shown in the figure, this illustrates the construction process of the local variation energy index. The upper subplot (smoothed deviation profile) displays the smoothed, continuous deviation profile function obtained after the S1 step preprocessing. ; Middle subplot (second derivative of the deviation profile): shows The second derivative of the strip directly corresponds to the local curvature of the strip at each point in a physical sense. It can be seen that in the straight sections of the strip, the second derivative is close to zero; in the curved sections, there are sharp fluctuations. Lower subplot (local variation energy distribution): This is the final output of step S2. The curve is obtained by calculating the second derivative energy within a sliding window. The peaks on the curve clearly mark regions where the strip morphology changes significantly, such as the two main peaks near 45 meters and 95 meters, providing a basis for subsequent identification of macroscopic deviations.

[0031] S3. Perform peak detection on the local variation energy curve to locate the deviation event, and extract the physical characteristic scale of the deviation by calculating the full width at half maximum (FWHM) of the energy peak.

[0032] In the steps Calculate the local variation energy curve Subsequently, the peak on the curve corresponds to the region of morphological change. However, these changes may originate from macroscopic deviations that significantly affect cutting quality, such as a large... Bending can also originate from localized, high-frequency, harmless disturbances. Therefore, this step requires... The key information that can distinguish between the two is extracted from the curve, namely the characteristic scale corresponding to the macroscopic deviation.

[0033] Its construction is based on the macroscopic deviation of a long waveform, such as a large one spanning tens of meters. The curvature change caused by the bending is gradual, and the corresponding variation energy is distributed over a relatively long spatial range. The curve shows a relatively broad and slowly changing energy peak; while a local, transient perturbation will produce a narrow and sharp energy peak. Therefore, the width of the peak is directly related to the physical scale of the deviation event.

[0034] The specific steps are as follows: First, for Peak detection is performed on the curve to locate the center of all energy peaks. For each detected peak, its characteristic scale is calculated. Specifically: ; in Is with peak point The characteristic scale of the associated deviation event, in physical terms, is the width of the spatial influence range of the deviation event; It is the Herveside step function, and its function value is [value] when its independent variable is greater than zero. Otherwise, it is 0; It is a threshold coefficient used to determine the effective support range of the peak. As a preferred option, The value can be set to At this point, the formula calculates the width of the energy peak at half its height. ), which is a commonly used and robust parameter value for measuring peak width. When When the value is high, such as The calculated scale is more sensitive to the shape of the peak; when When the value is low, such as This makes it more sensitive to the gradual change at the bottom of the peak. Implementers can evaluate and select a setting value that yields better results based on sensitivity and stability.

[0035] The above formula calculates the total length of the region where the energy value exceeds a certain proportion of the peak value through integration. This is applicable when the strip exhibits a large-scale, slowly changing energy... Bending or When bending, its corresponding The summit was very wide, leading to the calculations... A large value indicates that this deviation needs close monitoring. Conversely, a small value indicates a brief, localized jitter. The summit was very narrow. The value will be very small. Preferably, by setting a reasonable threshold, such as a threshold that is several times larger than the width of the strip itself, those [strips] can be effectively filtered out. Larger macroscopic deviations that have a substantial impact on downstream cutting accuracy are filtered out, and local disturbances that do not need to be processed are removed.

[0036] like Figure 3 As shown, it illustrates the process of peak detection and macroscopic deviation screening.

[0037] Upper subplot: On the energy curve obtained from S2, the peak detection algorithm is used to locate the center position of all energy peaks, i.e., the red dots, corresponding to the key areas where the deviation shape changes. Lower subplot: Shows the characteristic scale of each peak being calculated and filtered according to a preset threshold (e.g., ">10.0 meters" as shown in the figure). The peaks marked with red triangles that are ultimately retained are those macroscopic deviations caused by large-scale, long-waveform C-shaped or S-shaped bends that have a substantial impact on the trimming accuracy, while narrow and sharp local disturbances have been filtered out.

[0038] It should be noted that in this example, local disturbances were not filtered out, so the label points in the upper and lower subplots are the same.

[0039] S4. Based on the feature scale extracted in step S3, an adaptive analysis window is dynamically generated, and a third-order polynomial regression is used for local fitting within the window to achieve accurate quantitative decoupling of the composite deviation morphology.

[0040] For the steps Each macroscopic deviation selected is then analyzed using its characteristic scale. This dynamically generates an appropriately sized analysis window, within which the original deviation profile is analyzed. Perform local morphological decoupling to replace global decoupling. analyze.

[0041] The location of the macroscopic deviation has been precisely pinpointed in the steps described above. and its physical scale Subsequent analysis can then be conducted using... Centered on, with Targeted mathematical modeling is performed within a local window of reference length. To this end, this embodiment constructs an adaptive scaling analysis window half-width, with: ; in This represents the half-width of the analysis window customized for the current macroscopic deviation; Indicates a value greater than The scaling factor. Its function is to ensure that the analysis window can completely cover the entire deviation morphology area and its transition zone. As a preferred solution, The range of values ​​can be to .like If the value is too small, it may cause the window edges to truncate part of the deviation shape, affecting the fitting accuracy; if If the value is too large, it will introduce too many irrelevant flat parts, increasing the computational burden and potentially reducing the sensitivity to fitting the core shape. Therefore, in this embodiment, it can be... Set as .

[0042] In the above formula, the feature scale The larger the macroscopic deviation, the larger the analysis window will be automatically obtained. Conversely, this achieves a highly adaptive match between the analytical scale and the content of the bias.

[0043] Finally, in the case of Within the defined local window, the original deviation profile data within that interval A low-order polynomial regression model is applied for fitting. For example, this embodiment uses a third-order polynomial model because it is sufficient to describe translation, tilting, and... bends and The composite form of the bend is as follows: ; By applying the least squares method within this local window, a unique set of fitting coefficients can be obtained. Specifically: : Indicates the center of the deviation core region The overall translation amount at the location; : indicates in The local tilt at a certain point corresponds to the skew angle of the strip. Quantified The degree and direction of the bend. The sign of the value indicates the direction of the bend (convex or concave), and the absolute value indicates the severity of the bend. Quantified The degree and direction of the bend. The presence and magnitude of its value characterize whether and to what extent the strip is twisted in that region.

[0044] like Figure 4 As shown, it illustrates the decoupling process of the deviation event detected at 29.6 meters; Analysis window: The X-axis range of the entire plot represents the feature-scale-based data in S4. Dynamically generated adaptive analysis window; local fitting: the blue curve in the figure represents the original bias data within this window, while the red curve represents the fitted model obtained through third-order polynomial regression. ; Parameter decoupling: The text box in the upper left corner lists the four core morphological parameters obtained by the least squares method: .

[0045] The above decoupled morphological parameters The data is transmitted to the downstream correction controller. Based on these parameters, the controller drives the corresponding actuators, such as correction rollers, to perform precise and differentiated compensation actions. Because the entire analysis process is local and event-driven, and the analysis window size is dynamically adaptive, this invention reduces the system's response delay without sacrificing the accuracy of identifying macroscopic deviations in long waveforms, achieving rapid response and high precision in correction control.

[0046] Compared with existing technologies, the method proposed in this invention constructs an adaptive-scale analysis framework by introducing the concepts of local variation energy and characteristic scale. This framework can dynamically adjust the analysis window to match the physical scale of the deviation event, thereby solving the problems of traditional methods. The method faces a fundamental contradiction between real-time performance and analytical accuracy. Simultaneously, through local multinomial regression, it achieves the ability to handle translation, tilt, and... bends and The precise decoupling and quantification of complex deviations such as bends provide diagnostic information with rich spatial attributes for downstream precise control, significantly improving the performance and efficiency of the entire intelligent deviation correction system.

[0047] An example of an intelligent deviation correction system for strip cutting: On the other hand, the present invention also provides an intelligent correction system for strip cutting. For example... Figure 5 As shown, an intelligent strip cutting correction system includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement an intelligent strip cutting correction method according to the first aspect of the present invention.

[0048] An intelligent correction system for strip cutting also includes other components well known to those skilled in the art, such as communication interfaces. The settings and functions of these components are known in the art and will not be described in detail here.

[0049] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented using computer-readable / executable instructions stored or otherwise maintained by such a computer-readable medium.

Claims

1. A smart correction method for strip steel cutting, characterized in that, The method includes: Obtain the real-time deviation profile of the strip steel along its length upstream of the cutting station; Calculate the local variation energy of the deviation profile, which is positively correlated with the variance of the second derivative of the deviation profile within a local window of a set length. Locate the energy peaks in the local variation energy curves along the length direction, and calculate the characteristic scale of each energy peak; the characteristic scale is positively correlated with the width of the corresponding energy peak at a set threshold height; Macroscopic deviations with feature scales greater than a preset scale threshold are selected; for each macroscopic deviation, an adaptive analysis window is constructed with its peak position as the center and its width proportional to the feature scale; within the adaptive analysis window, the deviation profile is fitted with a polynomial to obtain the fitting coefficients; The fitting coefficients are transmitted to the correction controller to perform correction actions.

2. The intelligent correction method for strip steel cutting according to claim 1, characterized in that, Obtaining the real-time deviation profile includes: Multiple displacement sensors are deployed upstream of the cutting station to collect lateral deviation data points of the strip edge; The lateral deviation data points are digitally filtered to obtain a smooth and continuous deviation profile function. .

3. The intelligent correction method for strip cutting according to claim 2, characterized in that, The local variation energy The specific calculation method is as follows: ； in This represents the deviation profile function; Indicates the location The second derivative at point; This indicates the length of the local window; For window The arithmetic mean of the inner second derivatives.

4. The intelligent correction method for strip steel cutting according to claim 1, characterized in that, The feature scale The specific calculation method is as follows: ； in Indicates the location of the energy peak; For position Local variation energy at the location; For Herveside step function; The threshold coefficient is a value that ranges from 0.2 to 0.

8.

5. The intelligent correction method for strip steel cutting according to claim 1, characterized in that, The range of the adaptive analysis window is: ,in The peak position of the macroscopic deviation, half the window width. The specific calculation method is as follows: ;in This indicates a proportionality coefficient greater than 1; This represents the characteristic scale associated with the peak location.

6. The intelligent correction method for strip steel cutting according to claim 5, characterized in that, The proportionality coefficient The value range is 1.2 to 2.

0.

7. The intelligent correction method for strip steel cutting according to claim 1, characterized in that, The polynomial fitting is a third-order polynomial regression performed using the least squares method, specifically: ； in This represents the deviation profile after fitting. This represents the peak position of the macroscopic deviation; The fitting coefficient is denoted as .

8. The intelligent correction method for strip cutting according to claim 7, characterized in that, In the equation obtained by the third-order polynomial regression, the zero-order, first-order, second-order, and third-order coefficients correspond to the translation, inclination, C-shaped bending degree, and S-shaped bending degree of the strip, respectively.

9. The intelligent correction method for strip steel cutting according to claim 1, characterized in that, After receiving the fitting coefficients, the correction controller drives the correction roller actuator to perform a compensation action to correct the macroscopic deviation.

10. An intelligent deviation correction system for strip steel cutting, characterized in that, It includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, it implements the intelligent correction method for strip cutting as described in any one of claims 1 to 9.

Images