Online measurement method for thickness and flatness in heavy steel rolling process

Abstract

The invention relates to the technical field of online detection in a metal material rolling process, in particular to an online measurement method for thickness and flatness in a heavy steel rolling process, which comprises the following steps of: S1, mounting a laser distance measuring sensor array and an ultrasonic thickness sensor on the outlet side of a rolling mill, the laser ranging sensor array and the ultrasonic thickness sensor are initialized through a calibration program; s2, in the rolling process, a laser distance measuring sensor array and an ultrasonic thickness sensor are synchronously triggered to conduct real-time data collection; s3, performing data preprocessing on the distance data set and the original thickness data; s4, calculating the thickness value of the steel; s5, the flatness of the steel is calculated; and S6, performing data fusion on the final thickness value and the flatness index to generate a thickness and flatness measurement result. Through a multi-sensor data fusion and self-adaptive processing method, the thickness and flatness in the heavy steel rolling process can be accurately measured and fed back in real time, and therefore the rolling technology is optimized.

Description

Technical Field

[0001] This invention relates to the field of online detection technology in the rolling process of metal materials, and in particular to an online measurement method for thickness and flatness in the rolling process of heavy steel. Background Technology

[0002] The heavy steel rolling process refers to the plastic deformation of high-temperature steel billets using a rolling mill during heavy steel production to obtain steel products with the required thickness and flatness. In this process, thickness and flatness are key quality indicators that directly affect the mechanical properties and performance of the steel. Currently, thickness and flatness are primarily measured using offline methods or simple online sensor methods.

[0003] Offline measurement methods typically involve moving steel samples to a laboratory environment after the rolling process is complete, and then measuring them using contact measuring instruments (such as micrometers or thickness gauges) or non-contact instruments (such as laser rangefinders). This method suffers from significant lag, failing to provide real-time feedback on thickness and flatness changes during the rolling process, leading to delayed production adjustments and a high risk of batch defects. Furthermore, contact measurements are susceptible to instrument wear and measurement errors due to the high temperatures and surface oxide scale on the steel.

[0004] Online measurement methods attempt to measure thickness and flatness in real time during the rolling process, but existing online technologies have several problems. For example, some technologies use a single type of sensor (such as a laser sensor or an ultrasonic sensor) for measurement, but due to the harsh environment of heavy steel rolling (including high temperature, vibration, water mist, and dust interference), a single sensor is easily affected by environmental interference, resulting in unstable measurement data and low accuracy. In addition, existing online measurement methods often lack effective fusion and processing of multi-source data, leading to simple thickness and flatness calculation models that cannot compensate for errors caused by nonlinear factors such as temperature drift and mechanical vibration. Some methods attempt to measure flatness through image processing technology, but the high temperature radiation and oxide scale on the surface of heavy steel affect image quality, and the processing algorithms are complex and have poor real-time performance.

[0005] Therefore, the main problems with existing technologies are: low accuracy and poor reliability of online measurement, and significant susceptibility to interference from the rolling environment, failing to meet the real-time quality control requirements of the Chongqing Steel rolling process. To address these issues, this invention proposes an online measurement method based on multi-sensor data fusion and adaptive processing to improve the accuracy and robustness of thickness and flatness measurements. Summary of the Invention

[0006] To achieve the above objectives, the present invention provides an online measurement method for thickness and flatness during the rolling process of heavy steel, the online measurement method comprising the following steps: S1: Install a laser rangefinder array and an ultrasonic thickness sensor on the mill exit side, and initialize the laser rangefinder array and ultrasonic thickness sensor through a calibration procedure. The calibration procedure includes obtaining a reference value on a standard test block to compensate for the inherent error of the sensor. S2: During the rolling process, the laser ranging sensor array and the ultrasonic thickness sensor are simultaneously triggered to collect data in real time. The laser ranging sensor array collects distance data from multiple points on the steel surface to form a distance dataset, and the ultrasonic thickness sensor collects the raw thickness data of the steel. The acquisition frequency is adaptively adjusted according to the rolling speed. S3: Data preprocessing is performed on the distance dataset and the original thickness data. Data preprocessing includes noise filtering and temperature compensation. Noise filtering uses an adaptive median filtering algorithm to filter the distance dataset and the original thickness data to remove impulse noise. The window size of the adaptive median filtering is dynamically adjusted according to the degree of data fluctuation. The degree of data fluctuation is determined by calculating the variance of the data. Temperature compensation is performed on the distance dataset and the original thickness data based on the surface temperature of the steel. The surface temperature of the steel is acquired in real time by an infrared temperature sensor. Temperature compensation uses a temperature-expansion coefficient relationship model. S4: Calculate the thickness value of the steel based on the preprocessed thickness data. The thickness calculation includes taking the arithmetic mean of the thickness data collected by multiple ultrasonic thickness sensors to obtain the preliminary thickness value, and calculating the standard deviation of the thickness data according to the statistical distribution of historical thickness data to remove outliers that exceed the standard deviation range. The average value is then recalculated to obtain the final thickness value. S5: Calculate the flatness of steel based on the distance dataset after data preprocessing. The flatness calculation includes selecting multiple reference points from the distance dataset and fitting the reference plane using the least squares method. Calculate the distance deviation value from each measurement point to the reference plane. The root mean square value of all distance deviation values ​​is used as the flatness index. S6: The final thickness value and flatness index are fused to generate the measurement results of thickness and flatness. The data fusion adopts a weighted average algorithm, and the weight value is determined according to the sensor confidence level. The sensor confidence level is calculated by the accuracy of historical data. The measurement results are output to the rolling mill control system in real time and stored in the database.

[0007] Preferably, in S1, the laser ranging sensor array consists of at least three laser ranging sensors arranged linearly at equal intervals, covering the entire width direction of the steel. The ultrasonic thickness sensor includes two ultrasonic thickness sensors, installed at the edge and center of the steel, respectively. During installation, the distance between the laser ranging sensor array and the ultrasonic thickness sensor and the steel surface is ensured to be within the effective measurement range by adjusting the bracket. The calibration procedure specifically includes: placing a standard test block at the measurement position, the standard test block having a known thickness value and a known flatness value; collecting the measurement data of the laser ranging sensor array and the ultrasonic thickness sensor on the standard test block; calculating the deviation value between the measurement data and the known thickness value and the known flatness value; storing the deviation value as a compensation parameter; and applying the compensation parameter in real time to correct the measurement data in subsequent measurements.

[0008] Preferably, in S2, real-time data acquisition is initiated by a trigger signal from the rolling mill control system. The adaptive adjustment process of the acquisition frequency includes: measuring the rolling speed through an encoder, calculating the ratio of the rolling speed to a preset reference speed, and linearly adjusting the acquisition frequency according to the ratio to ensure that the number of data points acquired per unit length of steel is constant. The distance dataset and the original thickness data are temporarily stored in a buffer, and the size of the buffer is dynamically allocated according to the acquisition frequency to avoid data overflow.

[0009] Preferably, in S3, the adaptive median filtering algorithm for noise filtering specifically includes: setting an initial filtering window size, calculating the variance of the distance dataset and the original thickness data within the sliding window, the variance being obtained by dividing the sum of the squares of the data points and the average value within the window by the number of data points; if the variance is greater than a preset threshold, increasing the filtering window size; otherwise, keeping the filtering window size unchanged; and outputting smooth data after filtering. The temperature-expansion coefficient relationship model for temperature compensation is a linear model, which is established through laboratory calibration. Laboratory calibration includes measuring the thickness change of a standard test block at different temperatures, fitting the linear relationship between the temperature value and the expansion coefficient, and inputting the steel surface temperature value into the linear model to obtain the compensation coefficient. The compensation coefficient is multiplied by the original data to obtain the compensated data.

[0010] Preferably, in S4, the specific process of thickness calculation and verification includes: retrieving historical thickness data from the database for the most recent multiple measurement cycles, calculating the average value and standard deviation of the historical thickness data, obtaining the preliminary thickness value by taking the arithmetic mean of the thickness data collected by multiple ultrasonic thickness sensors in the current cycle, and marking it as an outlier and removing it if the difference between the current thickness data and the average value exceeds three times the standard deviation. After removing the outlier, the arithmetic mean of the remaining thickness data is recalculated as the final thickness value, and the final thickness value is stored in the database for subsequent data fusion.

[0011] Preferably, in S5, the specific process of flatness calculation includes: selecting three reference points from the distance dataset, with the three reference points located at the two ends and the center of the distance dataset respectively; fitting the reference plane using the least squares method; obtaining the plane parameters by solving a system of linear equations using the least squares method; constructing the system of linear equations based on the coordinate values ​​of the reference points; calculating the distance deviation value from each measurement point to the reference plane; calculating the distance deviation value using the point-to-plane distance formula; obtaining the root mean square value of all distance deviation values ​​by taking the square root of the sum of the squares of the distance deviation values ​​divided by the number of measurement points; and outputting the flatness index to the data fusion step.

[0012] Preferably, in S6, the weighted average algorithm for data fusion specifically includes: the sensor confidence level is calculated by comparing historical measurement values ​​with standard values ​​to determine the accuracy rate, the accuracy rate is the proportion of historical measurement values ​​that are consistent with the standard values, the weight value is proportional to the accuracy rate, and the final thickness value and flatness index are multiplied by their corresponding weight values ​​and then summed to obtain the fused measurement result. The measurement result is output to the rolling mill control system in real time to adjust the rolling parameters and stored in the database for quality traceability.

[0013] Preferably, in S3, a temperature-compensated infrared temperature sensor is installed near the laser rangefinder array. The acquisition frequency of the infrared temperature sensor is synchronized with the acquisition frequency of the laser rangefinder array and the ultrasonic thickness sensor. The surface temperature value of the steel is acquired in a non-contact manner by the infrared temperature sensor. The temperature-expansion coefficient relationship model is updated periodically. The update process includes recalibrating the infrared temperature sensor and the temperature-expansion coefficient relationship model using a standard temperature source during the rolling interval.

[0014] Preferably, in S4 and S5, historical thickness data and historical distance data are retrieved from the database in real time. The database storage period is the data of the most recent multiple rolling batches. The calculation of the statistical distribution includes rolling updates of the mean and standard deviation. The rolling update is implemented through a sliding window. The size of the sliding window is dynamically adjusted according to the rolling speed to ensure that the statistical distribution reflects the current rolling status.

[0015] Preferably, in S6, the data fusion also includes a reliability check, which includes calculating the coefficient of variation of the final thickness value and flatness index. If the coefficient of variation exceeds a preset threshold, an alarm is triggered and the sensor system is reinitialized. The reinitialization process includes repeating the calibration procedure in S1 to ensure the reliability of the measurement results.

[0016] The beneficial effects of this invention are: 1. This invention utilizes real-time synchronous measurement via a laser rangefinder sensor array and an ultrasonic thickness sensor. This not only avoids the lag inherent in offline measurements but also enables real-time acquisition of steel thickness and surface flatness data during the rolling process. This online measurement method allows for more timely data collection and feedback during production, enabling rapid identification of quality fluctuations, timely adjustment of rolling parameters, effective reduction of scrap rates, and improved production efficiency.

[0017] 2. By adopting non-contact laser rangefinders and ultrasonic thickness sensors instead of traditional contact measuring instruments, instrument wear and measurement errors caused by high temperatures, oxide scale, and uneven steel surfaces are eliminated. This avoids damage to the instrument caused by contact measurements and improves measurement stability and reliability over long-term operation.

[0018] 3. This invention overcomes the problem of single-sensor susceptibility to environmental interference by combining multi-sensor data fusion from laser rangefinders and ultrasonic thickness sensors. Laser sensors are less affected by vibration, dust, and water mist, while ultrasonic sensors can accurately measure the thickness of steel. This multi-sensor system works collaboratively, using weighted averaging and adaptive algorithms for data fusion, further improving measurement accuracy and stability, and enhancing the system's robustness in complex environments.

[0019] 4. This invention employs temperature compensation and an adaptive median filtering algorithm to address the effects of high temperatures and mechanical vibrations. In practical applications, temperature and mechanical vibrations can cause deviations in sensor measurement results. This invention compensates for the temperature effect using a temperature-expansion coefficient relationship model, while simultaneously utilizing an adaptive median filtering algorithm to remove noise and maintain data stability, thereby effectively reducing measurement errors caused by nonlinear factors.

[0020] 5. This invention calculates the flatness of steel by using a sensor-based physical measurement method, avoiding the problem of limited image quality. Furthermore, by using the least squares method for flatness calculation, it can improve calculation accuracy while ensuring real-time performance.

[0021] 6. This invention acquires historical thickness and distance data in real time, and combines this with statistical analysis methods such as standard deviation and mean to eliminate outliers. Finally, a weighted average algorithm is used for data fusion, further improving the accuracy of the measurement results. This not only effectively avoids the problems of lack of data fusion and historical data utilization in traditional methods, but also improves the reliability of the data, ensuring high precision and reliability of the measurement results.

[0022] 7. The final thickness and flatness measurement results are obtained through a data fusion algorithm and output to the rolling mill control system in real time, enabling precise control of the production process. This system can dynamically adjust rolling parameters based on the measurement results to ensure that the steel quality consistently meets production standards, thereby achieving comprehensive quality control. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in this invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart of the steps of the method of the present invention; Figure 2 This is a flowchart illustrating the steps of the adaptive median filtering algorithm for noise filtering in the method of this invention. Figure 3 This is a flowchart illustrating the specific steps of the thickness calculation and verification process of the method of the present invention. Detailed Implementation

[0025] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. It should also be noted that, to make the embodiments more comprehensive, the following embodiments are the best and preferred embodiments, and those skilled in the art can use other alternative methods to implement some well-known technologies; moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0026] Please see Figures 1-3 This invention provides an online measurement method for thickness and flatness during the rolling process of heavy steel, which is suitable for real-time quality monitoring in the production process of heavy steel.

[0027] The online measurement method of this invention achieves real-time measurement of thickness and flatness through multi-sensor data acquisition, adaptive filtering, and data fusion. The method includes six main steps: sensor system installation and initialization, real-time data acquisition, data preprocessing, thickness calculation and verification, flatness calculation, and data fusion and output.

[0028] In this embodiment, the sensor system is installed and initialized: The laser rangefinder array consists of multiple laser rangefinders arranged linearly, covering the entire width of the steel surface. The laser rangefinders utilize the principle of laser triangulation, enabling non-contact distance measurement. During installation, the distance between the sensor array and the steel surface must be maintained within the effective measurement range (e.g., 50-500 mm, depending on sensor specifications). Initial calibration uses a standard test block with known thickness and flatness values, for example, a thickness of 20 mm and a flatness value of 0.1 mm / m. During calibration, data from the sensors on the test block is collected, the deviation between the measured values ​​and the standard values ​​is calculated, and stored as compensation parameters. These compensation parameters are applied in real-time in subsequent measurements to eliminate inherent sensor errors (such as zero drift and linearity errors).

[0029] Ultrasonic thickness sensors are based on the pulse-echo principle, calculating thickness by measuring the propagation time of ultrasonic waves in steel. During installation, the sensor must be perpendicular to the steel surface, and a coupling agent (a high-temperature coupling agent is used in high-temperature environments) is used to ensure effective sound wave transmission. Initial calibration also uses a standard test block to obtain thickness deviation values ​​as compensation parameters.

[0030] The sensor is fixed to the exit side of the rolling mill by a bracket that is adjustable to accommodate different steel sizes and rolling conditions. The bracket is made of a high-temperature resistant alloy to resist the heat radiation of the rolling environment.

[0031] Real-time data acquisition: Data acquisition is initiated by a trigger signal from the rolling mill control system, which is synchronized with the rolling speed. The trigger frequency is adaptively adjusted according to the rolling speed. Specifically, the rolling speed (in m / s) is measured by an encoder, and the ratio to a preset reference speed (e.g., 1 m / s) is calculated. The acquisition frequency is then adjusted proportionally (e.g., the acquisition frequency doubles when the rolling speed is 2 m / s). This ensures a constant number of data points collected per unit length of steel, avoiding data that is too dense or too sparse.

[0032] The acquired data is temporarily stored in a buffer, the size of which is dynamically allocated. For example, the buffer size is calculated based on the acquisition frequency and data type: each sensor data point occupies a fixed number of bytes, and the buffer size is set to store data for at least one rolling cycle (e.g., 10 seconds) to prevent data overflow.

[0033] Data preprocessing: An adaptive median filtering algorithm is used to process the distance dataset and the raw thickness data. The algorithm first sets the initial filter window size (e.g., a 3×3 window), and then calculates the variance of the data within the sliding window. The variance is calculated using the formula: Variance = Σ(Data points - Mean). 2 / Number of data points. If the variance value is greater than a preset threshold (e.g., 0.1), it indicates significant noise, so the window size is increased (e.g., to 5×5); otherwise, the original window size is maintained. After filtering, smooth data is output.

[0034] The surface temperature of the steel is acquired in real time using an infrared temperature sensor. The infrared temperature sensor is installed near the laser rangefinder array, and its acquisition frequency is synchronized with the main sensor. The temperature-coefficient of thermal expansion relationship model is linear, established through laboratory calibration: the thickness change of a standard specimen is measured at different temperature points (e.g., 100°C to 800°C, in 50°C intervals), and a linear relationship between temperature and the coefficient of thermal expansion is fitted; for example, coefficient of thermal expansion = a × temperature + b, where a and b are fitting parameters. During compensation, the acquired temperature value is input into the model to obtain the compensation coefficient, which is then multiplied by the original data to complete the compensation.

[0035] The temperature-expansion coefficient relationship model is updated regularly (e.g., once per shift), and the infrared temperature sensor and model are recalibrated using a standard temperature source (such as a blackbody radiation source) during rolling breaks.

[0036] Thickness calculation and verification: The arithmetic mean of the thickness data collected by multiple ultrasonic thickness sensors is calculated. For example, using three sensors located on the left, middle, and right sides of the steel, the average value is calculated as (left thickness + middle thickness + right thickness) / 3.

[0037] The standard deviation is calculated based on the statistical distribution of historical thickness data. Historical data is retrieved from the database for the most recent N measurement periods (e.g., N=100), and the mean and standard deviation are calculated. If the difference between the current thickness data and the mean exceeds three times the standard deviation (based on the 3σ principle), it is considered an outlier and removed. After removal, the average of the remaining data is recalculated as the final thickness value.

[0038] Historical thickness data is stored in a database and updated using a rolling mechanism. The sliding window size is dynamically adjusted based on the rolling speed: at higher rolling speeds, the window size decreases to ensure data timeliness.

[0039] Flatness calculation: Reference plane establishment: Three reference points are selected from the distance dataset (e.g., the left, center, and right endpoints of the array), and a reference plane is fitted using the least squares method. The coefficients of the plane equation Ax + By + C = z are solved using the least squares method, where (x, y) are the coordinates of the reference points, and z is the distance value. The solution process is accomplished by constructing a system of linear equations and using matrix operations.

[0040] Calculate the distance deviation from each measurement point to the reference plane using the point-to-plane distance formula: Deviation = |Ax + By + Cz| / √(A 2 +B 2+1). The root mean square (RMS) value of all deviations is used as a smoothness index, RMS = √(Σ deviation). 2 (Number of measurement points).

[0041] Data fusion and output: Sensor confidence is calculated using historical data accuracy. Accuracy is defined as the proportion of historical measurements that match a standard value (e.g., the percentage of consistent measurements out of the last 100). Weights are proportional to accuracy; for example, weight = accuracy / Σ (accuracy of all sensors). The final thickness and flatness indices are multiplied by their respective weights and then summed to obtain the fusion result.

[0042] Calculate the coefficient of variation (standard deviation / mean) for the final thickness value and flatness index. If the coefficient of variation exceeds a preset threshold (e.g., 0.05), trigger an alarm and reinitialize the sensor system. Reinitialization includes repeating the calibration procedure in S1.

[0043] The measurement results are output to the rolling mill control system in real time to adjust the roll gap or rolling speed; at the same time, they are stored in the database to support quality traceability.

[0044] Example: The example was conducted on a heavy steel rolling production line, using high-temperature steel billets (approximately 800°C) as the rolling material, with a rolling speed ranging from 0.5 to 2 m / s. The production line environment was subject to vibration, water mist, and dust interference. The measurement targets were the thickness (nominal value 20 mm) and flatness (required ≤ 0.5 mm / m) of the rolled steel.

[0045] S1: Sensor System Installation and Initialization A laser rangefinder array is installed, consisting of three laser rangefinders (model: KeyenceIL-300) arranged linearly at equal intervals, covering the width of the steel section (1500mm). The sensor installation height is 200mm, adjustable via a bracket.

[0046] Two ultrasonic thickness sensors (model: Panametrics-NDT38DL) are installed at the edge and center of the steel, respectively, with the installation angle perpendicular.

[0047] Calibration is performed using a standard test block (20.00 mm thickness, flatness 0.10 mm / m). Sensor data is collected, and deviations are calculated. For example, a laser rangefinder sensor measures 199.5 mm (standard value 200.0 mm), with a deviation of -0.5 mm; an ultrasonic thickness sensor measures 20.05 mm, with a deviation of +0.05 mm. The deviation values ​​are stored as compensation parameters.

[0048] S2: Real-time data acquisition: The data acquisition is triggered by the rolling mill control system, and the rolling speed is measured by the encoder (current speed 1.5m / s). Data acquisition frequency calculation: The reference speed of 1m / s corresponds to a frequency of 100Hz, and the current frequency = 100 × (1.5 / 1) = 150Hz.

[0049] Data buffer size setting: Each data point occupies 4 bytes, data volume per second = 150 × 4 × number of sensors (5) = 3000 bytes, buffer allocation 30KB (stores 10 seconds of data).

[0050] S3: Data Preprocessing Adaptive median filtering application. Initial window size 3×3. Calculate variance: for example, if the variance of the distance dataset is 0.15 (greater than the threshold of 0.1), increase the window size to 5×5. After filtering, the data smoothness is improved.

[0051] An infrared temperature sensor (model: Fluke62Max+) collects the surface temperature of the steel (current value 750°C). The temperature-expansion coefficient relationship model is: Expansion coefficient = 0.000012 × Temperature + 0.000001 (fitted through laboratory calibration). The compensation coefficient = 1 + Expansion coefficient × (Current temperature - Reference temperature 20°C) = 1.0088. The original thickness data is multiplied by the compensation coefficient to complete the compensation.

[0052] S4: Thickness Calculation and Verification Preliminary thickness values ​​were calculated: the data from the three ultrasonic thickness sensors were 20.02 mm, 20.05 mm, and 19.98 mm, with an average value of 20.016 mm.

[0053] Outlier removal: Historical data (last 100 periods) was retrieved, with a mean of 20.00 mm and a standard deviation of 0.03 mm. The current data differs from the mean by less than 3 × standard deviation (0.09 mm), indicating no outliers. The final thickness value is 20.016 mm.

[0054] S5: Flatness calculation: Reference points were selected: distances from the left end (0,0,199.5mm), center (750,0,200.0mm), and right end (1500,0,199.8mm) of the dataset. The least squares method was used to fit the plane: the equation of the plane was obtained as z = -0.0002x + 0.0001y + 199.6.

[0055] Calculate the deviation: For example, the deviation of point (500, 0, 199.7 mm) = |-0.0002×500+0.0001×0+199.6-199.7| / √((-0.0002)) 2 +(0.0001) 2+1)≈0.05mm. The RMS of all deviations is 0.08mm / m.

[0056] S6: Data Fusion and Output Weighting calculation: The ultrasonic thickness sensor has a historical accuracy of 95%, and the laser rangefinder array has an accuracy of 90%, with weights of 0.53 and 0.47 respectively. The fused thickness value = 20.016 × 0.53 + (thickness estimate based on distance calculation) × 0.47.

[0057] Reliability check: Thickness variation coefficient 0.01 (less than threshold 0.05), no alarm. Output results to the rolling mill control system and store in the database.

[0058] It is understandable that: Variance threshold (0.1): This value is determined experimentally. In the Chongqing Iron and Steel rolling environment, a variance exceeding 0.1 indicates significant noise, requiring an increase in the filtering window. The threshold can be adjusted through on-site testing.

[0059] Temperature-expansion coefficient model: The linear model simplifies calculations and ensures real-time performance. Laboratory calibration uses least squares fitting, with a goodness-of-fit R0. 2 A value >0.99 indicates that the model is reliable.

[0060] Coefficient of variation threshold (0.05): This value is based on statistical quality control principles. A coefficient of variation exceeding 0.05 indicates measurement instability and requires intervention.

[0061] The 3σ principle: Under the assumption of normal distribution, cover 99.7% of the data to ensure effective outlier removal.

[0062] Handling multiple situations: If the rolling speed varies greatly, the acquisition frequency is adaptively adjusted to prevent data loss.

[0063] If the temperature fluctuates drastically, the model is updated periodically to maintain accuracy.

[0064] If a sensor fails, a reliability check triggers a reinitialization to ensure system robustness.

[0065] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0066] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An online measurement method for thickness and flatness during heavy steel rolling, characterized in that, The online measurement method includes the following steps: S1: Install a laser rangefinder array and an ultrasonic thickness sensor on the mill exit side, and initialize the laser rangefinder array and ultrasonic thickness sensor through a calibration procedure. The calibration procedure includes obtaining a reference value on a standard test block to compensate for the inherent error of the sensor. S2: During the rolling process, the laser ranging sensor array and the ultrasonic thickness sensor are simultaneously triggered to collect data in real time. The laser ranging sensor array collects distance data from multiple points on the steel surface to form a distance dataset, and the ultrasonic thickness sensor collects the raw thickness data of the steel. The acquisition frequency is adaptively adjusted according to the rolling speed. S3: Data preprocessing is performed on the distance dataset and the original thickness data. Data preprocessing includes noise filtering and temperature compensation. Noise filtering uses an adaptive median filtering algorithm to filter the distance dataset and the original thickness data to remove impulse noise. The window size of the adaptive median filtering is dynamically adjusted according to the degree of data fluctuation. The degree of data fluctuation is determined by calculating the variance of the data. Temperature compensation is performed on the distance dataset and the original thickness data based on the surface temperature of the steel. The surface temperature of the steel is acquired in real time by an infrared temperature sensor. Temperature compensation uses a temperature-expansion coefficient relationship model. S4: Calculate the thickness value of the steel based on the preprocessed thickness data. The thickness calculation includes taking the arithmetic mean of the thickness data collected by multiple ultrasonic thickness sensors to obtain the preliminary thickness value, and calculating the standard deviation of the thickness data according to the statistical distribution of historical thickness data to remove outliers that exceed the standard deviation range. The average value is then recalculated to obtain the final thickness value. S5: Calculate the flatness of steel based on the distance dataset after data preprocessing. The flatness calculation includes selecting multiple reference points from the distance dataset and fitting the reference plane using the least squares method. Calculate the distance deviation value from each measurement point to the reference plane. The root mean square value of all distance deviation values ​​is used as the flatness index. S6: The final thickness value and flatness index are fused to generate the measurement results of thickness and flatness. The data fusion adopts a weighted average algorithm, and the weight value is determined according to the sensor confidence level. The sensor confidence level is calculated by the accuracy of historical data. The measurement results are output to the rolling mill control system in real time and stored in the database.

2. The online measurement method for thickness and flatness in the rolling process of heavy steel as described in claim 1, characterized in that, In S1, the laser ranging sensor array consists of at least three laser ranging sensors arranged linearly at equal intervals, covering the entire width direction of the steel. The ultrasonic thickness sensor includes two ultrasonic thickness sensors, installed at the edge and center of the steel, respectively. During installation, the distance between the laser ranging sensor array and the ultrasonic thickness sensor and the steel surface is ensured to be within the effective measurement range by adjusting the bracket. The calibration procedure specifically includes: placing a standard test block at the measurement position, the standard test block having known thickness and known flatness values; collecting measurement data of the laser ranging sensor array and the ultrasonic thickness sensor on the standard test block; calculating the deviation value between the measurement data and the known thickness and known flatness values; storing the deviation value as a compensation parameter; and applying the compensation parameter in real time to correct the measurement data in subsequent measurements.

3. The online measurement method for thickness and flatness in the rolling process of heavy steel as described in claim 1, characterized in that, In S2, real-time data acquisition is initiated by the trigger signal of the rolling mill control system. The adaptive adjustment process of the acquisition frequency includes: measuring the rolling speed through the encoder, calculating the ratio of the rolling speed to the preset reference speed, and linearly adjusting the acquisition frequency according to the ratio to ensure that the number of data points collected per unit length of steel is constant. The distance dataset and the original thickness data are temporarily stored in the buffer, and the size of the buffer is dynamically allocated according to the acquisition frequency to avoid data overflow.

4. The online measurement method for thickness and flatness in the rolling process of heavy steel as described in claim 1, characterized in that, In S3, the adaptive median filtering algorithm for noise filtering specifically includes: setting the initial filtering window size, calculating the variance of the distance dataset and the original thickness data within the sliding window. The variance is obtained by dividing the sum of the squares of the data points and the average value within the window by the number of data points. If the variance is greater than a preset threshold, the filtering window size is increased; otherwise, the filtering window size remains unchanged. Smooth data is output after filtering. The temperature-expansion coefficient relationship model for temperature compensation is a linear model. The linear model is established through laboratory calibration, which includes measuring the thickness change of a standard test block at different temperatures and fitting the linear relationship between the temperature value and the expansion coefficient. In the compensation process, the surface temperature value of the steel is input into the linear model to obtain the compensation coefficient. The compensation coefficient is multiplied by the original data to obtain the compensated data.

5. The online measurement method for thickness and flatness in the rolling process of heavy steel as described in claim 1, characterized in that, In S4, the specific process of thickness calculation and verification includes: retrieving historical thickness data from the database for the most recent multiple measurement cycles, calculating the average and standard deviation of the historical thickness data, obtaining the preliminary thickness value by taking the arithmetic mean of the thickness data collected by multiple ultrasonic thickness sensors in the current cycle, marking the current thickness data as an outlier and removing it after removing outliers, recalculating the arithmetic mean of the remaining thickness data as the final thickness value, and storing the final thickness value in the database for subsequent data fusion.

6. The online measurement method for thickness and flatness in the rolling process of heavy steel according to claim 1, characterized in that, In S5, the specific process of flatness calculation includes: selecting three reference points from the distance dataset, with the three reference points located at the two ends and the center of the distance dataset respectively; fitting the reference plane using the least squares method; obtaining the plane parameters by solving a system of linear equations using the least squares method; constructing the system of linear equations based on the coordinate values ​​of the reference points; calculating the distance deviation value from each measurement point to the reference plane; calculating the distance deviation value using the point-to-plane distance formula; and obtaining the root mean square value of all distance deviation values ​​by taking the square root of the sum of the squares of the distance deviation values ​​divided by the number of measurement points. The flatness index is then output to the data fusion step.

7. The online measurement method for thickness and flatness in the rolling process of heavy steel according to claim 1, characterized in that, In S6, the weighted average algorithm for data fusion specifically includes: the sensor confidence level is calculated by comparing historical measurement values ​​with standard values ​​to determine the accuracy rate. The accuracy rate is the proportion of historical measurement values ​​that are consistent with the standard values. The weight value is proportional to the accuracy rate. Finally, the thickness value and flatness index are multiplied by their respective weight values ​​and then summed to obtain the fused measurement result. The measurement result is output to the rolling mill control system in real time to adjust the rolling parameters and stored in the database for quality traceability.

8. The online measurement method for thickness and flatness in the rolling process of heavy steel as described in claim 1, characterized in that, In S3, a temperature-compensated infrared temperature sensor is installed near the laser rangefinder array. The acquisition frequency of the infrared temperature sensor is synchronized with the acquisition frequency of the laser rangefinder array and the ultrasonic thickness sensor. The surface temperature value of the steel is acquired in a non-contact manner through the infrared temperature sensor. The temperature-expansion coefficient relationship model is updated periodically. The update process includes recalibrating the infrared temperature sensor and the temperature-expansion coefficient relationship model using a standard temperature source during the rolling interval.

9. The online measurement method for thickness and flatness in the rolling process of heavy steel according to claim 1, characterized in that, In S4 and S5, historical thickness data and historical distance data are retrieved from the database in real time. The database storage period is the data of the most recent rolling batches. The calculation of the statistical distribution includes rolling updates of the mean and standard deviation. The rolling update is implemented through a sliding window. The size of the sliding window is dynamically adjusted according to the rolling speed to ensure that the statistical distribution reflects the current rolling status.

10. The online measurement method for thickness and flatness in the rolling process of heavy steel according to claim 1, characterized in that, In S6, data fusion also includes a reliability check, which involves calculating the coefficient of variation of the final thickness value and flatness index. If the coefficient of variation exceeds a preset threshold, an alarm is triggered and the sensor system is reinitialized. The reinitialization process includes repeating the calibration procedure of S1 to ensure the reliability of the measurement results.

Images