A method and system for detecting the edge drop of strip steel based on the multi-head coordination of an edge drop meter.

Abstract

This application relates to the fields of mechanical engineering and automated inspection technology, and discloses a method and system for detecting the edge drop of strip steel based on the collaborative operation of multiple measuring heads of edge drop gauges. It aims to solve the technical problems in existing technologies, such as spatial misalignment of data from multiple thickness gauges, distortion of the edge drop curve reconstruction, and inability to accurately reflect the thickness distribution of the same cross-section caused by the high-speed operation and speed fluctuations of the strip steel. The method includes: synchronously acquiring thickness data using multiple non-contact thickness gauges arranged in a linear array; obtaining the actual displacement of the strip steel using a high-resolution rotary encoder to establish a spatiotemporal mapping relationship between the thickness data and the physical cross-section; reconstructing a distortion-free cross-sectional thickness distribution curve using a dynamic interpolation algorithm based on a preset sampling interval, and calculating the edge drop value. This application significantly improves the accuracy and dynamic response capability of edge drop detection through a synchronization mechanism and interpolation reconstruction based on the actual displacement of the strip steel, providing high-fidelity data support for strip shape control.

Description

Technical Field

[0001] This invention belongs to the field of mechanical engineering and automated testing technology, specifically relating to a method and system for detecting the edge drop of strip steel based on the collaborative operation of multiple measuring heads of an edge drop meter. Background Technology

[0002] In the strip rolling process, edge thinning (i.e., edge drop) is one of the key factors affecting product quality. To monitor and control edge drop online, multiple thickness gauges are typically arranged along the width of the strip to obtain cross-sectional thickness distribution data.

[0003] Current mainstream methods are based on time-synchronized detection logic, assuming that data collected by each thickness gauge at the same moment corresponds to the same physical cross-section of the strip. However, in actual high-speed rolling, the strip speed fluctuates, and the probes have a fixed spacing in the rolling direction. This means that under variable speed conditions, the data collected by each probe at the same "moment" actually comes from different "positions" on the strip. Existing technologies directly stitch together this time-synchronized but spatially misaligned data, and the reconstructed thickness curve cannot accurately reflect the actual shape of any specific cross-section, resulting in detection distortion.

[0004] However, despite improvements in single-point measurement accuracy and environmental adaptability, existing edge drop meter detection solutions still suffer from data coordination issues among multiple measuring heads when facing high-speed and variable-speed rolling. The core deficiency lies in the lack of a high-precision tracking and data synchronization mechanism for the actual physical displacement of the strip, making it impossible to accurately spatially locate and align the data from each measuring head when the speed changes. This limits the accuracy and reliability of edge drop detection under high-speed and variable-speed rolling conditions, making it difficult to meet the high fidelity requirements of high-precision strip shape closed-loop control for input data. Summary of the Invention

[0005] This invention provides a method and system for detecting strip edge drop based on multi-head collaborative edge drop gauges. It aims to solve the technical problems in existing technologies, such as spatial misalignment of data from multiple thickness gauges, distortion of edge drop curve reconstruction, and inability to accurately reflect the thickness distribution of the same cross-section caused by the high-speed movement and speed fluctuations of the strip. This invention introduces a spatiotemporal synchronization mechanism based on the actual displacement of the strip, combined with high-precision displacement encoding and dynamic interpolation reconstruction algorithms, to achieve accurate, real-time, and distortion-free detection of strip edge drop.

[0006] According to one aspect of the present invention, a method for detecting the edge drop of strip steel based on the coordinated operation of multiple measuring heads of an edge drop meter is provided, comprising: S1. Multiple non-contact thickness gauges arranged in a linear array along the width direction of the strip are used to synchronously collect thickness data at multiple transverse positions of the strip based on a unified clock source, and output a thickness data stream with time stamps. S2. Real-time sensing of the linear displacement of the strip through a high-resolution rotary encoder, and outputting an absolute displacement data stream with time stamp that is synchronized with the thickness data stream; S3. Spatiotemporal coordinate mapping: Match each thickness sampling point in the thickness data stream with the corresponding displacement in the absolute displacement data stream according to the timestamp, and assign a vertical physical coordinate to each thickness sampling point to form spatiotemporal marker data; S4. Generate a sequence of equally spaced target cross-section sampling points on the longitudinal physical coordinate axis of the strip steel at preset fixed physical distances. S5. For each target cross-sectional sampling point, retrieve the nearest valid measurement point pair upstream and downstream of the sampling point from the spatiotemporal marker data for each lateral position; S6. Based on the effective measurement point pair, the thickness estimate at the sampling point of the target cross section at each lateral position is obtained by interpolation calculation; S7. Based on the thickness estimates at the sampling points of the target cross-section at each transverse position, generate a continuous thickness distribution curve covering the entire width of the strip. S8. Calculate and output the edge reduction value based on the thickness distribution curve.

[0007] Furthermore, in step S1, synchronous acquisition is achieved in the following way: a unified hardware trigger pulse is generated by the master clock source of the central data acquisition module and distributed to all thickness gauges to force synchronous sampling. Alternatively, each thickness gauge can be configured with a local clock that is phase-locked with the master clock source. Each thickness gauge can autonomously sample and return timestamped data, which is then timestamped by the data acquisition system.

[0008] Furthermore, in step S2, the high-resolution rotary encoder is coaxially mounted with a roller that is in contact with the surface of the strip steel via a rigid coupling; The output quadrature pulse signal is multiplied and counted by a high-speed counter module, and the count value is timestamped based on a clock synchronized with the unified clock source, thereby calculating the absolute displacement of the strip.

[0009] Furthermore, step S3 specifically includes: comparing the timestamp of the thickness sampling point with the timestamp recorded in the absolute displacement data stream; when the timestamp difference is less than a preset threshold, directly matching the displacement amount; otherwise, performing linear interpolation between two adjacent displacement records to obtain the corresponding longitudinal physical coordinates.

[0010] Furthermore, in step S4, the preset fixed physical distance value is determined comprehensively based on the typical strip rolling speed, the thickness gauge sampling frequency, and the required edge drop detection spatial resolution, and its value is not greater than the natural sampling interval obtained by dividing the rolling speed by the sampling frequency.

[0011] Furthermore, in step S5, the range of the neighborhood is dynamically determined based on the real-time rolling speed and the sampling frequency of the thickness gauge, and its radius is no greater than half of the natural physical distance between consecutive samples at the same measuring point.

[0012] Furthermore, in step S6, the interpolation calculation employs either linear interpolation or cubic spline interpolation algorithms; When using cubic spline interpolation, at least four valid measurement points must be retrieved in the extended neighborhood for each lateral position to construct the spline function.

[0013] Furthermore, step S6 also includes a data verification and fault tolerance step: before interpolating the measurement point pair, verify the rationality of its thickness value and status flag. If the verification fails or there are not enough valid points, the data preservation strategy is enabled, and the thickness value of the most recently successfully calculated thickness value at the corresponding horizontal position is used as the current output.

[0014] Furthermore, prior to step S1, an initialization calibration step for the thickness gauge array is included: performing background noise calibration to determine the zero offset of each thickness gauge, and using a standard template for gain and linearity calibration.

[0015] According to another aspect of the present invention, a strip edge drop detection system based on multi-head coordination of an edge drop meter is provided, comprising: The multi-thickness gauge array module consists of multiple non-contact thickness gauges arranged linearly along the width direction of the strip and aligned with the rolling direction. The high-precision displacement sensing module includes a high-resolution rotary encoder, an installation roller that rotates synchronously with the strip, and a high-speed counter module; The central synchronization and acquisition module has a built-in unified master clock source and is connected to the multi-thickness gauge array module and the high-precision displacement sensing module respectively, so as to realize data synchronous acquisition and time reference alignment. The data processing and analysis module includes: Spatiotemporal coordinate mapping unit is used to bind thickness data with displacement data to generate spatiotemporal marker data; The sampling planning module is used to generate a sequence of sampling points for the target cross-section based on a fixed physical distance. The neighborhood retrieval and interpolation reconstruction unit is used to retrieve valid measurement point pairs for each target point and perform interpolation calculations. The curve generation and edge drop calculation unit is used to generate the full-width thickness distribution curve and calculate the edge drop value. The output module is used to output the edge degradation value to the plate shape control system.

[0016] In summary, this application includes at least one of the following beneficial technical effects: 1. This invention introduces a spatiotemporal synchronization mechanism based on the actual physical displacement of the strip. A high-precision encoder tracks the actual displacement, strictly anchoring each thickness measurement value to its true longitudinal coordinate. Combined with a dynamic interpolation reconstruction algorithm based on physical coordinates, this ensures that each reconstructed cross-sectional thickness curve accurately and without distortion reflects the thickness distribution at the same physical location on the strip. This eliminates the error source caused by speed variations in the detection principle, realizing a shift from "approximate estimation" to "true restoration" in edge drop detection.

[0017] 2. This invention constructs a complete technology chain, from high-precision synchronous acquisition, strict displacement-timestamp alignment, and circular buffer management, to intelligent neighborhood retrieval and fault-tolerant interpolation. The system incorporates fault-tolerant mechanisms such as data validity verification, outlier removal, and preservation of historical values ​​when data is missing, effectively resisting sudden speed changes, signal interference, or momentary sensor failures in industrial settings. Ultimately, under high-speed continuous production conditions, the system can output thickness distribution curves with constant and comparable spatial resolution with extremely low latency. This provides stable, reliable, and high-fidelity input data for downstream shape control systems (such as bending rolls and shifting rolls), enabling control commands to be accurately generated based on real cross-sectional information, significantly improving control performance.

[0018] 3. This invention achieves dynamic, continuous, and uninterrupted perception of the thickness distribution along the entire length of the strip by reconstructing a continuous cross-sectional sequence based on displacement. This not only enables accurate calculation of edge reduction values ​​but also fully reflects the dynamic evolution of the strip shape during the rolling process. This end-to-end, high-fidelity data acquisition capability allows quality control to go beyond simply evaluating the final result. Instead, it enables a deeper understanding and optimization of the forming process, providing unprecedented data support and analytical foundation for process optimization, defect prevention, and improved quality consistency across the entire industry chain. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the architecture of a strip edge drop detection method based on the collaboration of multiple measuring heads of an edge drop meter; Figure 2 This is a schematic diagram of a spatiotemporal synchronization mechanism and dynamic interpolation reconstruction algorithm based on the actual displacement of the strip steel. Figure 3 This is a flowchart of the process for multi-thickness gauge data acquisition, spatiotemporal coordinate mapping, and cross-sectional sampling planning. Detailed Implementation

[0020] This invention provides a method and system for detecting the edge drop of strip steel based on the collaboration of multiple measuring heads of an edge drop meter. Its core lies in the use of a spatiotemporal synchronization mechanism based on the actual physical displacement of the strip steel, combined with a dynamic interpolation reconstruction algorithm, to achieve distortion-free reconstruction of the thickness distribution of the strip steel cross-section under high-speed operation, thereby accurately calculating the edge drop value.

[0021] Firstly, in conjunction with the accompanying drawings, the following will provide a detailed description of the specific implementation steps of the strip edge reduction detection method based on the multi-head collaboration of the edge reduction instrument, supplemented by the necessary disclosure of the system structure.

[0022] The first step, S1, aims to construct the data acquisition front-end for strip edge drop detection. This is achieved by deploying multiple non-contact thickness gauges on the mill exit side and establishing a high-precision synchronous acquisition system to obtain thickness data at multiple points along the strip width. Step S1 provides an accurate and synchronous raw data foundation for subsequent spatiotemporal mapping and cross-sectional reconstruction based on the actual strip displacement. The specific implementation process is as follows: S101: On the mill exit side, a non-contact thickness gauge array is installed along the strip width direction (i.e., perpendicular to the rolling direction). Multiple thickness gauges in this array are fixed by a rigid, highly stable mechanical mounting beam, ensuring that the measuring probes of all thickness gauges are precisely aligned with the same physical position line in the rolling direction, thus forming a linear array arrangement. The total number of thickness gauges in the array is no less than three.

[0023] During installation, first determine the centerline of the strip's movement, align the optical path or beam center of at least one thickness gauge with this centerline, and configure it as a center thickness gauge.

[0024] The remaining thickness gauges are installed symmetrically on both sides of the center line, using the central thickness gauge as a reference and according to the pre-set lateral spacing, to cover the edge drop characteristic area of ​​the strip. X-ray thickness gauges can be used as the type of thickness gauge.

[0025] Each thickness gauge has a preset sampling frequency of no less than 2000 Hz within the system, and its thickness measurement accuracy, as specified by the manufacturer and guaranteed by subsequent system calibration, is better than ±0.5 μm or ±0.1% (whichever is greater).

[0026] S102: All thickness gauges’ data signal output terminals are hardwired to the central data acquisition module via shielded cables in a star or bus topology.

[0027] The data acquisition system integrates a high-precision, high-stability crystal oscillator master clock source. The system adopts a master-slave synchronization mechanism, in which the master clock source periodically generates a unified hardware trigger pulse signal, which is distributed to each thickness gauge through a dedicated synchronization signal line, forcing all thickness gauges to perform thickness sampling at the same time.

[0028] As an alternative, the system can also equip each thickness gauge with a local clock that is phase-locked with the master clock source. Each thickness gauge samples autonomously based on this clock, but the timestamp containing the high-precision clock count needs to be returned along with the data, and the data acquisition system will perform timestamp alignment processing at the back end.

[0029] By using any of the above synchronization mechanisms, it can be ensured that the time reference error of the sampling data obtained from all thickness gauges is less than 1 millisecond.

[0030] S103: Each thickness gauge, under its internal microcontroller or a communication protocol agreed upon with the data acquisition system, performs analog-to-digital conversion on the analog thickness obtained from each sample and packages it into a fixed-format digital data frame. This data frame contains at least the following fields: a 32-bit floating-point number representing the original thickness measurement value, a 16-bit integer representing the device's unique identifier, and a 64-bit high-precision counter value representing the absolute timestamp of this sample.

[0031] These data frames are uploaded to the central data acquisition module in real time via a high-speed fieldbus or industrial Ethernet interface. The main control unit of the data acquisition system maintains a first-in-first-out circular buffer in memory as a raw data buffer. All arriving data frames are immediately parsed and stored in this circular buffer. The storage record includes all fields of the data frame and a system timestamp added by the data acquisition system host to indicate the time of receipt.

[0032] The size of the buffer is configured based on the maximum allowable data processing latency and the number of thickness gauges to ensure that no overflow occurs under peak data traffic.

[0033] S104: After roller replacement, maintenance, or system restart on the production line, a fully automated thickness gauge array initialization and calibration process is executed. This process is controlled by the main control program of the data acquisition system.

[0034] First, background noise calibration is performed: After ensuring that there is no strip steel or other objects in the measurement area, the system sends instructions to all thickness gauges to continuously collect data for several cycles. The system calculates the average value of the readings of each thickness gauge as its electronic zero offset and stores it.

[0035] Then, gain and linearity calibration is performed: an operator or auxiliary drive device passes a standard sample of the same material as the strip steel and of known thickness through the measurement area of ​​the thickness gauge array at a constant low speed.

[0036] The system controls all thickness gauges to perform synchronous scanning measurements as the template passes through, acquiring a series of thickness readings. The system then performs least-squares fitting of the data sequence measured by each thickness gauge with the standard thickness value of the template to calculate the gain correction coefficient and linear compensation parameter for that individual thickness gauge.

[0037] All calibration parameters (zero offset, gain coefficient, etc.) are stored in the non-volatile memory of the data acquisition system. During subsequent online testing, the system reads these parameters in real time and applies the formula "calibrated thickness = (original reading - zero offset) × gain coefficient" to correct each original thickness measurement value in real time, thereby ensuring the uniformity and accuracy of the measurement reference of the entire thickness gauge array.

[0038] In summary, step S1 achieves synchronous acquisition and preprocessing of multi-point thickness data in the width direction of the strip. Its output multi-channel thickness sequence with time stamp provides accurate and reliable input for the spatiotemporal coordinate mapping based on high-precision displacement sensing in the subsequent step S2.

[0039] Step S2, based on the time-stamped multi-channel thickness data already implemented in Step S1, aims to establish precise displacement sensing of the strip in the rolling direction. Its core is to track the strip's movement in real time using a high-precision rotary encoder and convert pulse signals into continuous displacement values, thereby providing a spatial reference for subsequently mapping the thickness data to the actual physical cross-section. The specific implementation process is as follows: S201: A high-resolution rotary encoder with a resolution of no less than 100,000 pulses per revolution is selected. The encoder is directly coaxially mounted on the shaft end of the tension roller or looper roller via a rigid coupling. The roller contacts the surface of the strip and rotates synchronously with it by friction.

[0040] The encoder should be installed in a way that ensures its rotation axis is strictly aligned with the rotation axis of the roller, so as to ensure that there is a definite proportional relationship between the number of angular displacement pulses output by the encoder and the linear displacement of the strip, that is, a strict linear correspondence.

[0041] S202: When the encoder is working, it outputs two incremental orthogonal square wave signals (labeled as phase A and phase B) with a 90-degree phase difference. These two differential signals are connected to the differential input port of a dedicated high-speed counter module via shielded twisted-pair cable. The counter module contains logic circuitry capable of quadrupling the frequency of the A and B phase signals and determining their direction.

[0042] It detects the edges (rising and falling edges) of the A and B phase pulses and their order of occurrence. When each edge event is triggered, it increments or decrements an internal 32-bit or 64-bit counter based on the direction determination result, thereby completing the pulse accumulation count with motion direction recognition in real time.

[0043] S203: To ensure that the displacement calculated subsequently is strictly aligned with the thickness data collected by the thickness gauge in time, it is necessary to ensure that the reading clock of the pulse counter module is synchronized with the main clock source of the central data acquisition module described in S102.

[0044] The specific way to achieve this hard synchronization is: the clock input pin of the counter module is directly connected to the main clock output line of the data acquisition system, so as to use the same physical clock source for driving; Alternatively, the counter module receives a synchronous clock signal with a fixed phase relationship, distributed by the master clock source of the data acquisition system via a clock driver.

[0045] With this hardware connection method, the moment when the counter module captures and updates the pulse count value can be marked by the data acquisition system with a homogeneous, high-precision timestamp.

[0046] S204: The system software periodically (e.g., every 1 millisecond) reads the latest cumulative pulse count value from the dedicated counter module. Based on this count value, the effective circumference of the roller on which the encoder is mounted (obtained through calibration during system initialization), and the encoder nameplate parameter "pulses per revolution", the absolute linear displacement of the strip since a fixed reference zero point set by the system is calculated in real time.

[0047] The calculation formula is: Absolute displacement = (Cumulative number of pulses × Roller circumference) / Number of pulses per revolution. The calculated displacement is expressed as a floating-point number in millimeters (mm).

[0048] S205: The latest absolute displacement obtained from each calculation, together with the precise timestamp corresponding to that calculation, is written in real time to a shared data area or circular queue in the memory of the data acquisition system as a "displacement-time" data pair.

[0049] The system maintains a unified time reference service. When any thickness gauge data frame arrives in step S103, the system immediately queries the latest "displacement-time" data stored in the shared data area at that moment, and assigns a precise longitudinal physical coordinate (i.e., displacement) to the thickness data based on the proximity of the timestamps (requiring the time difference to be less than a very small threshold, such as 0.5 milliseconds), through linear interpolation or direct matching.

[0050] Ultimately, each thickness data record will be expanded into a complete spatiotemporal information unit containing "thickness value, lateral position (implied by the device identifier), vertical physical coordinates, and timestamp" for use by subsequent modules.

[0051] In summary, step S2 establishes a high-precision spatial coordinate axis based on the actual physical displacement of the strip. The time-series thickness data output from step S1 is transformed into spatial sequence data with clear physical positional information in the rolling direction. This transformation is crucial to overcoming data misalignment caused by strip speed fluctuations and provides spatial reference information for establishing accurate spatiotemporal mapping relationships in subsequent step S3 and for reconstructing distortion-free cross-sectional thickness curves in steps S4 to S7.

[0052] For step S3, building upon the results of step S2, which established a high-precision displacement sequence of the strip in the rolling direction and aligned it with the thickness gauge data in time to form a timestamped "displacement-time" pair, step S3 utilizes the precise displacement information to assign a true spatial location attribute to the thickness data collected in step S1, which originally only had time attributes. This establishes a one-to-one spatiotemporal mapping relationship between the thickness measurement value and the physical cross-section of the strip. The specific implementation process is as follows: S301: The system allocates and maintains a circular buffer in its running memory. The total capacity of this buffer is pre-calculated to ensure that it can continuously store raw thickness data frames reported by all thickness gauges at the highest sampling frequency within the last 10 seconds, as well as displacement-time data pairs generated periodically by step S205.

[0053] The buffer's data structure is designed so that each record corresponds to a unique spatiotemporal data point. The system manages the writing and overwriting of the buffer by maintaining a pair of read and write pointers, ensuring that new data always overwrites the oldest data, thus achieving sliding window management of the data. This buffer serves as the core data pool for the entire spatiotemporal mapping module.

[0054] S302: When the data acquisition system receives a digital data frame from any thickness gauge in real time through the high-speed communication link established in step S103, the system immediately calls the preset parsing program to process the data frame.

[0055] The parsing process accurately extracts three core fields from the data frame according to the fixed format defined in S103: a 32-bit floating-point number representing the original thickness measurement value, a 16-bit integer device ID used to identify the specific thickness gauge, and a 64-bit high-precision timestamp characterizing the sampling time of the thickness gauge itself.

[0056] S303: After successfully parsing the thickness measurement data and obtaining its precise sampling timestamp, the system immediately uses this timestamp as the query key to initiate a fast search in the "displacement-time" data queue maintained in step S205. The goal of the search is to locate the displacement record in the displacement data queue whose timestamp has the smallest difference from the thickness measurement sampling timestamp.

[0057] Since the thickness gauge and the displacement counter share the same master clock source, their timestamps are essentially from the same source. The system calculates the absolute difference between the thickness measurement timestamp and the displacement recording timestamp.

[0058] If the difference is less than a preset synchronization fault tolerance threshold (e.g., 0.5 milliseconds), the displacement record is considered valid and the absolute displacement stored therein can be used directly.

[0059] If no perfect match is found but there are two displacement records before and after each other, and the thickness measurement timestamp is between the timestamps of these two records and the difference is within the fault tolerance threshold, then the system uses a linear interpolation algorithm to calculate the precise interpolated displacement corresponding to the thickness measurement sampling time based on the time and displacement of these two displacement records.

[0060] S304: The precise absolute displacement determined in step S303 that strictly corresponds to the thickness measurement sampling time is directly defined as the longitudinal physical coordinate of the strip corresponding to the thickness measurement point.

[0061] The system uses this longitudinal physical coordinate value as a new data field and associates it with the original thickness value and device identifier obtained from step S302, combining them into an enhanced data record containing spatial location information.

[0062] S305: Package the enhanced data record generated in step S304 into a complete spatiotemporal tagged record according to a predefined structure. This record must contain at least: thickness value (before or after calibration), lateral position index (derived from device ID), longitudinal physical coordinates, and the timestamp of data generation.

[0063] Subsequently, the system writes this spatiotemporal marker record to the current write position of the annular buffer managed by step S301 and updates the write pointer. This storage action ensures that each piece of raw thickness measurement data is permanently bound to its actual two-dimensional physical position on the strip (the lateral position is determined by the fixedly installed probe, and the longitudinal position is determined by the encoder displacement), thus completing the transformation from "time series data" to "spatial distribution data".

[0064] Steps S301 to S303 collectively address the "when and where" problem, finding the precise spatial location (displacement) for each thickness sampling point through timestamp alignment and interpolation. Steps S304 and S305 then complete the data reorganization and persistence, generating data units with complete spatiotemporal information. This completely changes the traditional method's approach where data only has time-series attributes.

[0065] In summary, step S3 transforms the time-based multi-channel thickness sampling stream into a spatial data point set with the physical displacement of the strip as the coordinates. This transformation allows all subsequent processing to be performed based on physical spatial coordinates, eliminating the need for traditional splicing logic that relies on unstable time intervals.

[0066] For step S4, the location of the "target cross-section" for the final thickness curve reconstruction is planned and generated. The core of this step is to define a fixed physical distance in the rolling direction as the sampling interval, and generate an equidistant sequence of sampling points accordingly. This discretizes the continuous strip space into a series of physical cross-sections to be analyzed, specifically including the following steps: S401: During system initialization or process parameter changes, a preset cross-sectional sampling interval is set for the system via a human-machine interface or a preset configuration file, based on the strip steel product specifications and testing requirements. This interval is a fixed physical distance value in the rolling direction, used to define the center distance between two adjacent strip steel physical cross-sections that the system needs to reconstruct in subsequent steps. In this embodiment, the typical value range of this preset interval is between 5 mm and 50 mm.

[0067] S402: The specific value of the above sampling interval is determined comprehensively based on three factors: the typical rolling speed of the strip, the fixed sampling frequency of the thickness gauge array, and the required spatial resolution for edge drop detection. The core principle is that the sampling interval determines the spatial density of the output thickness curve.

[0068] As a calculation benchmark, the natural physical spacing of the same measuring point on the strip between consecutive samples is first evaluated. The calculation formula is: Natural physical spacing = Rolling speed / Thickness gauge sampling frequency.

[0069] For example, when the rolling speed is 30 m / s and the thickness gauge sampling frequency is 2000 Hz, the natural spacing is 15 mm. If the sampling interval is set to this natural spacing value, the system will reconstruct the cross-section near the spatial location corresponding to each original measurement point.

[0070] To obtain denser thickness distribution information (i.e., higher spatial resolution) than the original measurement points, the sampling interval should be set to a value smaller than the natural spacing (e.g., 5 mm). The smaller the interval, the more spatial points the reconstructed thickness curve will have, but it will require more reliance on subsequent interpolation algorithms to calculate the thickness value of the intermediate points from the original data. This process is called spatial oversampling.

[0071] S403: During the strip rolling process, the system uses the real-time updated absolute displacement coordinate axis of the strip established in steps S2 and S3 as the spatial reference. Starting from the head of the strip entering the detection area or from a manually set absolute displacement zero point, the system continuously generates a sequence of strictly equidistant cross-sectional sampling points in the rolling direction according to the fixed physical sampling interval set in step S401.

[0072] The specific generation method is as follows: divide the current absolute displacement of the strip by the sampling interval, round the quotient down, and then multiply it by the sampling interval to determine the theoretical coordinates of the next sampling point to be generated at an integer multiple interval. The system dynamically maintains this sequence to ensure that the sequence always contains several future sampling points that will soon enter the processing window.

[0073] Each sample point generated in this way represents the precise longitudinal center position of a "target cross-section" to be reconstructed. This dynamically updated sequence of sample points is stored in system memory as spatial location input to drive subsequent steps S5 to S7 for data retrieval and interpolation calculations.

[0074] The equidistant physical cross-section sampling point sequence generated in step S4 provides a spatial target framework for subsequent data processing. This framework is based on physical displacement and is independent of changes in the strip running speed. This allows subsequent steps S5 and S6 to operate stably in a fixed, time-invariant spatial coordinate system, ensuring that the reconstructed cross-sectional thickness curves at different times have a consistent and comparable spatial reference system.

[0075] For step S5, for each target sampling point, the most relevant original measurement data that can be used to calculate the thickness value of that point is quickly and accurately identified from the massive spatiotemporal marker data established in step S3 and stored in the circular buffer. The specific implementation process is as follows: S501: For the target cross-section sampling point to be processed, the system first determines an upstream and downstream search neighborhood centered on its longitudinal physical coordinates. The range of this neighborhood (i.e., the distance from the center point to the upstream or downstream boundary) is dynamically set according to the data spatial density of the strip in the rolling direction.

[0076] Specifically, the system calculates a reference value called "natural sampling interval of the thickness gauge" based on the real-time strip rolling speed and the fixed sampling frequency of the thickness gauge. The calculation formula is: Natural sampling interval = Current rolling speed / Thickness gauge sampling frequency.

[0077] This value represents the typical spatial distance between two consecutive samples of the same measurement point on the strip. The system sets the radius of the search neighborhood to no more than half of this natural sampling interval.

[0078] By limiting the search range to this neighborhood, it is ensured that the candidate data points found for the target point are closely adjacent to it in space, which satisfies the basic requirement of data locality for interpolation algorithms, while avoiding the introduction of irrelevant or outdated data due to an excessively large search range.

[0079] S502: The system initiates a spatial retrieval of the circular buffer established and maintained in step S301, using the longitudinal coordinates of the target point and the neighborhood range calculated in step S501 as query conditions.

[0080] All spatiotemporal marker records stored in this circular buffer contain vertical physical coordinates, thickness values, device IDs, etc. When written or through a background process, they are sorted in ascending order according to their vertical physical coordinate values, or an efficient spatial index structure similar to a B-tree is established.

[0081] The system utilizes this ordered structure or index to quickly locate all data records in the buffer whose vertical coordinates fall within the upstream and downstream neighborhoods of the target point. The retrieval process is performed independently for each thickness gauge channel, aiming to find all candidate data points falling within the spatial neighborhood for each lateral position.

[0082] S503: For multiple candidate data points belonging to the same thickness gauge channel retrieved through S502, the system first performs a validity check. The check is based on whether the preset quality flag bit in the data record and the thickness value are within a preset reasonable physical range (e.g., greater than 0 and less than the maximum possible thickness of the strip). Data points that pass the check are marked as valid.

[0083] Subsequently, the system calculates the absolute difference between the vertical coordinate of each valid data point and the target point coordinate. Based on the sign of the difference (coordinates greater than the target point are downstream, and coordinates less than the target point are upstream), all valid data points are divided into upstream point sets and downstream point sets.

[0084] Next, in the upstream and downstream point sets, find the data point with the smallest absolute difference from the target point. These two points are the selected "upstream nearest point" and "downstream nearest point", which are located on both sides of the target point and are the closest to it.

[0085] This filtering logic ensures that subsequent interpolation calculations are based on the most reliable measurements of the two nearest neighbors to the target point, thus avoiding the amplification of errors that may result from unilateral extrapolation.

[0086] S504: The system checks whether step S503 successfully filtered out the "upstream nearest point" and "downstream nearest point" for the current thickness gauge channel. If both exist, these two data points constitute a valid data pair and are output to the subsequent S6 interpolation calculation module.

[0087] If, for a certain thickness gauge channel, no valid upstream or downstream point (or both) is found in the neighborhood, the system determines that the channel has "missing data" at the current target cross-section position. For channels with missing data, the system skips the interpolation calculation for that channel in this round of reconstruction.

[0088] If all thickness gauge channels are determined to have "missing data" at the current target point, the system determines that the overall reconstruction of the target cross-section sampling point has failed. The system will generate and record an exception event log. At the same time, in order to maintain the input continuity of the downstream plate shape control system, the system will retain the thickness value of the same lateral position (channel) output in the previous successful calculation cycle as the thickness output value of that channel in the current cycle.

[0089] Step S5 uses spatial retrieval, filtering, and fault-tolerant logic to prepare the raw data pairs required for interpolation at each preset physical cross-sectional location. This ensures that the dynamic interpolation calculation in the subsequent step S6 is based on the nearest and most reliable real measurement data, and is a crucial data preprocessing step for ultimately generating a high-fidelity, distortion-free thickness distribution curve.

[0090] Step S6: Based on the data point pairs prepared in step S5, interpolation calculations are performed to determine the precise thickness value of the target point at its corresponding lateral position in the channel. This is a crucial calculation step in synthesizing discrete measurement data into a continuous thickness distribution curve. The specific implementation process is as follows: S601: Before formally performing the interpolation calculation, the system performs a final validity check on each data point pair (i.e., the "upstream nearest point" and the "downstream nearest point") output in step S5.

[0091] The verification operation is completed by accessing predefined fields in each data point record and applying logical judgment rules, specifically including: Read the thickness value field and determine whether its value is within the preset physical reasonable range (e.g., the lower limit is greater than 0 and the upper limit is less than the maximum thickness value allowed by the strip steel process specification).

[0092] Read the data status flag field to check for predefined abnormal flags such as sensor hardware failure, communication timeout, and data verification error.

[0093] Read the timestamp fields of the two data points, calculate their difference, and verify the time order (the upstream timestamp should be less than the downstream timestamp) and whether the time interval is within the reasonable physical motion time range estimated based on the current strip speed.

[0094] S602: If a data point is found to be abnormal during the re-verification, the system will remove that point. After removal, the system will determine the number of remaining valid data points.

[0095] If there are at least two valid points remaining, the subsequent interpolation steps will continue; if there are fewer than two valid points, the system will be unable to perform reliable interpolation calculations.

[0096] At this point, the system activates a data retention strategy. Specifically, the system maintains a storage unit for each thickness gauge channel (horizontal position) to record the thickness value output by the most recent successful interpolation calculation for that channel.

[0097] When an interpolation fails due to insufficient data, the system reads the previous successful calculation result for the corresponding channel from the storage unit and uses it as the output value for the current target cross-sectional position and the current channel thickness. This strategy ensures the continuity of the output curve even when individual data points are momentarily abnormal.

[0098] S603: When the system is configured in linear interpolation mode and the data is valid, the system performs linear interpolation calculations. Let the vertical physical coordinates of the target point be... The coordinates of the upstream valid point are Thickness value The coordinates of the downstream effective point are Thickness value Then the formula for calculating the thickness h of the target point is:

[0099] The system performs floating-point calculations based on this formula to obtain an estimated thickness value for the target point. This calculation is based on the assumption that the thickness varies linearly between two points, and it features simple calculation and fast response.

[0100] S604: Perform cubic spline interpolation calculation (optional mode) When a higher degree of smoothness is required for the thickness distribution curve, the system can adopt a cubic spline interpolation mode. In this mode, the system not only needs the pair of nearest neighbors provided in step S5, but also needs to retrieve and verify a total of four valid data points for this channel from the S301 buffer within a larger neighborhood (e.g., extended to 1 times the natural sampling interval) centered on the target point x. These four points are required to be sorted by coordinates and distributed as evenly as possible on both sides of the target point.

[0101] The system is based on the coordinates of these four points. (in The thickness and z values ​​are constructed using a standard cubic spline interpolation algorithm. Specifically, the system solves for a spline function S(z) consisting of piecewise cubic polynomials, which satisfies: In each sub-interval The above is a cubic polynomial; 2) Through all given data points, i.e. ; 3) Throughout the entire interval It has continuous first and second derivatives. Natural boundary conditions (where the second derivative is zero at the endpoints) or given boundary conditions are typically added to determine a unique solution. After construction, the value of S(x) is calculated, which is the thickness estimate h at the target point x.

[0102] This method can generate smoother thickness curves, but the computational cost is relatively high.

[0103] In summary, step S6 transforms the spatially discrete measurement data provided in step S5 into continuous thickness estimates at each lateral position on each target cross-section. After the thickness value of each channel at each target point is calculated one by one, it will be integrated and arranged by the subsequent step S7 to finally generate the cross-sectional thickness distribution curve, thus laying the foundation for accurate edge drop calculation.

[0104] For step S7, the specific details are as follows: S701: The system first acquires the thickness interpolation results of all thickness gauge channels at the longitudinal coordinate of the current target cross-section, calculated in step S6. The interpolation result of each channel represents the thickness estimate at a fixed lateral position point of that channel. The system sorts these discrete thickness values, which correspond one-to-one with each channel, according to their known installation lateral coordinates, initially forming a discrete two-dimensional thickness data point set corresponding to the current longitudinal position.

[0105] S702: Because the thickness gauges are discretely arranged along the width of the strip, there are transverse gap areas between adjacent thickness gauges that are not directly measured. In order to obtain a thickness distribution curve that continuously covers the entire width of the strip, the system needs to perform transverse interpolation filling on these gap areas.

[0106] The system iterates through all adjacent thickness gauge channel pairs. For any pair of adjacent channels, the system reads their known lateral installation coordinates. and ( < ), and the corresponding thickness value obtained from step S701. and The system, based on a preset lateral interpolation density, in and A series of virtual horizontal coordinate points are inserted at equal intervals between them.

[0107] S703: For each virtual transverse coordinate point Xi generated in step S702, the system uses a linear interpolation algorithm to calculate its corresponding thickness value. The specific calculation formula is as follows: The system processes all adjacent channels one by one until the thickness value at all preset virtual lateral coordinate points is calculated, thereby filling all measurement gaps.

[0108] S704: For the extended areas on the outer sides of the strip, i.e., the area between the outermost thickness gauge mounting point and the actual physical edge of the strip, the system uses an extrapolation algorithm for processing. Based on the measurement point of the outermost thickness gauge and the measurement point of the adjacent inner thickness gauge, the system calculates outwards to a preset position on the edge of the strip, such as 0 mm from the edge of the strip, according to the thickness variation trend of this area.

[0109] S705: After completing the interpolation and extrapolation calculations for all lateral positions, the system integrates data points from three parts: direct interpolation results from all thickness gauge channels, lateral interpolation points between all adjacent channels, and extrapolation points from the two edges. The system uniformly sorts all data points according to the ascending order of lateral coordinates, generating a sequence containing continuous lateral coordinates and corresponding thickness values.

[0110] S706: The system normalizes the sequence according to the preset final output curve resolution requirements. These requirements are typically reflected in two indicators: first, the total number of data points in the entire curve must be no less than 100; second, the horizontal spacing between data points must remain constant. Based on the aforementioned sequence, the system resamples or performs secondary interpolation across the entire width of the strip at a constant horizontal spacing to generate the final standard output thickness distribution curve. The horizontal coordinates of this curve cover the entire range from the left edge to the right edge of the strip, and its spatial resolution is determined by the preset cross-sectional sampling interval and is independent of the strip's running speed.

[0111] S707: The system stores the final generated complete thickness distribution curve corresponding to the longitudinal coordinates of the current target cross-section in the designated output buffer and marks it as ready. At the same time, the system repeats steps S5 to S6 to obtain new longitudinal interpolation results for each channel for the next preset cross-section sampling point coordinates, and executes the process of steps S701 to S706 again to achieve uninterrupted reconstruction of the thickness distribution of the continuous cross-section over the entire strip length.

[0112] Step S7 completes the reconstruction process from multi-channel discrete longitudinal interpolation results to a complete two-dimensional cross-sectional thickness distribution curve. First, steps S701-S703 associate the longitudinal information with fixed transverse coordinates and use linear interpolation to fill the transverse measurement gaps, ensuring the continuity of the curve in the width direction. Then, steps S704-S706 process the edge regions and normalize and resample the full-width data, outputting a standard thickness distribution curve with a fixed high density and equidistant data points. The core of this process lies in its comprehensive use of interpolation techniques in both longitudinal and transverse dimensions on a longitudinal cross-section determined by actual physical displacement coordinates, transforming measurement data from a finite number of fixed points into a continuous full-width thickness distribution. This completely eliminates distortion-prone methods that rely on time-stitching, ensuring that each reconstructed cross-section is a real physical cross-section on the strip steel, and its curve is true and free of geometric distortion.

[0113] This complete and high-fidelity thickness distribution curve provides a direct and reliable data basis for the accurate calculation of edge drop values ​​in the subsequent S8 step, enabling the assessment of edge drop to truly reflect the thickness distribution differences within the same physical cross-section.

[0114] Finally, in step S8, the edge reduction value is calculated based on the reconstructed thickness distribution curve and output to the control system: S801: The system acquires the complete thickness distribution curve data corresponding to the current target physical cross-section, output from step S7. This data is a set of ordered data points, each containing a transverse coordinate value and a precise thickness value, continuously covering the full width of the strip.

[0115] S802: Determine the calculation area. Based on the known strip width, the system first locates its centerline. The center area is defined as all data points within 50 mm on either side of the centerline. The left edge area is defined as data points within 25 mm to 100 mm from the left edge of the strip. The right edge area is defined as data points within 25 mm to 100 mm from the right edge of the strip.

[0116] S803: Calculate the average thickness. The system calculates the arithmetic mean of the thickness values ​​of all data points in the central region, the left edge region, and the right edge region, respectively, to obtain the average thickness of the three regions, denoted as _____. , and .

[0117] S804: Calculate edge degradation. The system calculates the degradation of the left edge separately. and the right-side decrease These two differences quantify the amount of thickness reduction at the edges relative to the central region.

[0118] S805: Result Output and Closed-Loop Control. The system will calculate the... and The data, along with a timestamp, is packaged into a standard data message and sent in real time to the rolling mill's automatic shape control system via an industrial communication interface. Based on this, the system generates adjustment commands to dynamically control the work roll bending or shifting mechanism, achieving closed-loop control of the edge drop.

[0119] Step S8 defines the specific region and calculation method for extracting the edge drop feature value from the high-fidelity thickness curve, and feeds the result back to the control system in real time through a standard interface. This process transforms the accurate cross-sectional thickness distribution obtained from the previous steps into key parameters that directly drive strip shape adjustment, thus forming a complete closed loop from detection and analysis to control. This ensures that the edge drop control command is based on real, distortion-free cross-sectional information, ultimately improving the uniformity of strip thickness.

[0120] At the system level, the edge drop detection system of the present invention includes a multi-thickness gauge array module, a high-precision displacement sensing module, a spatiotemporal coordinate mapping module, a cross-sectional sampling planning module, a neighborhood data retrieval module, a dynamic interpolation reconstruction module, an edge drop curve generation module, and an edge drop value calculation module.

[0121] The multi-thickness gauge array module consists of no fewer than three non-contact thickness gauges, which are fixedly installed along the width of the strip steel, and their sampling frequency and accuracy meet the aforementioned requirements.

[0122] The high-precision displacement sensing module consists of a high-resolution rotary encoder and its signal processing circuit. It is synchronized with the master clock through hard wiring to ensure the high fidelity of the displacement signal.

[0123] The spatiotemporal coordinate mapping module integrates a circular buffer and a timestamp alignment engine, responsible for binding the original thickness data with the displacement. The cross-section sampling planning module dynamically configures the sampling interval based on process parameters and generates a sequence of sampling points.

[0124] The neighborhood data retrieval module enables efficient spatial indexing and data filtering, supporting millisecond-level response.

[0125] The dynamic interpolation and reconstruction module has a built-in validity check and interpolation algorithm library, supporting both linear and cubic spline modes. The edge descent curve generation module coordinates the workflow of each module and outputs a standardized thickness distribution curve.

[0126] The edge reduction calculation module performs regional average and difference calculations and pushes the results to the control system.

[0127] The entire system runs on an industrial real-time operating system, with all modules interacting via shared memory and message queues. Upon system startup, it first completes hardware self-tests and parameter initialization, including thickness gauge calibration, encoder zero-point calibration, and sampling interval setting. During rolling, the system cyclically performs data acquisition, coordinate mapping, interpolation reconstruction, and edge reduction calculations at microsecond-level timing, ensuring output latency is less than twenty milliseconds. An anomaly handling mechanism is implemented throughout the entire process, including sensor failure detection, data loss compensation, and communication interruption recovery, ensuring the system's robustness in harsh industrial environments.

[0128] In summary, this embodiment solves the distortion problem of strip edge drop detection under high-speed variable working conditions through strict spatiotemporal synchronization, dynamic interpolation reconstruction and closed-loop feedback mechanism, realizes the true restoration of the thickness distribution of the physical cross section, and provides a reliable data foundation for high-precision strip shape control.

[0129] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention. Therefore, the embodiments should be regarded as exemplary and non-limiting in all respects.

[0130] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A method for detecting the edge drop of strip steel based on the coordinated operation of multiple measuring heads of an edge drop meter, characterized in that, include: S1. Multiple non-contact thickness gauges arranged in a linear array along the width direction of the strip are used to synchronously collect thickness data at multiple transverse positions of the strip based on a unified clock source, and output a thickness data stream with time stamps. S2. Real-time sensing of the linear displacement of the strip using a high-resolution rotary encoder, and outputting an absolute displacement data stream with time stamps that is synchronized with the thickness data stream; S3. Spatiotemporal coordinate mapping: Match each thickness sampling point in the thickness data stream with the corresponding displacement in the absolute displacement data stream according to the timestamp, and assign a vertical physical coordinate to each thickness sampling point to form spatiotemporal marker data; S4. Generate a sequence of equally spaced target cross-section sampling points on the longitudinal physical coordinate axis of the strip steel at preset fixed physical distances. S5. For each target cross-sectional sampling point, retrieve the nearest valid measurement point pair upstream and downstream of the sampling point from the spatiotemporal marker data for each lateral position; S6. Based on the effective measurement point pair, the thickness estimate at the sampling point of the target cross section at each lateral position is obtained by interpolation calculation; S7. Based on the thickness estimates at the sampling points of the target cross-section at each lateral position, generate a continuous thickness distribution curve covering the entire width of the strip. S8. Calculate and output the edge reduction value based on the thickness distribution curve.

2. The method according to claim 1, characterized in that, In step S1, synchronous acquisition is achieved in the following way: a unified hardware trigger pulse is generated by the master clock source of the central data acquisition module and distributed to all thickness gauges to force synchronous sampling. Alternatively, each thickness gauge can be configured with a local clock that is phase-locked with the master clock source. Each thickness gauge can autonomously sample and return timestamped data, which is then timestamped by the data acquisition system.

3. The method according to claim 1, characterized in that, In step S2, the high-resolution rotary encoder is coaxially mounted with a roller that is in contact with the surface of the strip steel via a rigid coupling; The output quadrature pulse signal is multiplied and counted by a high-speed counter module, and the count value is timestamped based on a clock synchronized with the unified clock source, thereby calculating the absolute displacement of the strip.

4. The method according to claim 1, characterized in that, Step S3 specifically includes: comparing the timestamp of the thickness sampling point with the timestamp recorded in the absolute displacement data stream; when the timestamp difference is less than a preset threshold, directly matching the displacement amount; otherwise, performing linear interpolation between two adjacent displacement records to obtain the corresponding longitudinal physical coordinates.

5. The method according to claim 1, characterized in that, In step S4, the preset fixed physical distance value is determined comprehensively based on the typical rolling speed of strip steel, the sampling frequency of the thickness gauge, and the required edge drop detection spatial resolution, and its value is not greater than the natural sampling interval obtained by dividing the rolling speed by the sampling frequency.

6. The method according to claim 1, characterized in that, In step S5, the range of the neighborhood is dynamically determined based on the real-time rolling speed and the sampling frequency of the thickness gauge, and its radius is no greater than half of the natural physical distance between consecutive samples at the same measuring point.

7. The method according to claim 1, characterized in that, In step S6, the interpolation calculation uses either linear interpolation or cubic spline interpolation algorithms; When using cubic spline interpolation, at least four valid measurement points must be retrieved in the extended neighborhood for each lateral position to construct the spline function.

8. The method according to claim 7, characterized in that, Step S6 also includes data verification and fault tolerance steps: before interpolating the measurement point pairs, verify the rationality of their thickness values ​​and status flags; If the verification fails or there are not enough valid points, the data preservation strategy is enabled, and the thickness value of the most recently successfully calculated thickness value at the corresponding horizontal position is used as the current output.

9. The method according to claim 1, characterized in that, Before step S1, there is also an initialization calibration step for the thickness gauge array: performing background noise calibration to determine the zero offset of each thickness gauge, and using a standard template to perform gain and linearity calibration.

10. A strip edge drop detection system based on multi-head collaborative edge drop measuring instrument, used to implement the method of any one of claims 1 to 9, characterized in that, include: The multi-thickness gauge array module consists of multiple non-contact thickness gauges arranged linearly along the width direction of the strip and aligned with the rolling direction. The high-precision displacement sensing module includes a high-resolution rotary encoder, an installation roller that rotates synchronously with the strip, and a high-speed counter module; The central synchronization and acquisition module has a built-in unified master clock source and is connected to the multi-thickness gauge array module and the high-precision displacement sensing module respectively to realize data synchronous acquisition and time reference alignment. The data processing and analysis module includes: Spatiotemporal coordinate mapping unit is used to bind thickness data with displacement data to generate spatiotemporal marker data; The sampling planning module is used to generate a sequence of sampling points for the target cross-section based on a fixed physical distance. The neighborhood retrieval and interpolation reconstruction unit is used to retrieve valid measurement point pairs for each target point and perform interpolation calculations. The curve generation and edge drop calculation unit is used to generate the full-width thickness distribution curve and calculate the edge drop value. The output module is used to output the edge degradation value to the plate shape control system.

Images