A method and apparatus for detecting the thickness of water film on the surface of stainless steel after pickling.

Abstract

This application relates to a method and apparatus for detecting the thickness of a water film on the surface of stainless steel after pickling, comprising: uniformly conveying a stainless steel plate after pickling and washing to a detection area, triggering a position sensor to generate a start signal; the control system receiving the signal and activating an upper spectral confocal displacement sensor array, aligning it vertically with the steel plate; the sensor array emitting a broadband light beam and receiving light signals of different wavelengths reflected from the upper surface of the water film and the metal substrate; the signal processing unit calculating the current point water film thickness in real time based on the difference between two characteristic wavelengths and the refractive index model of water; generating a two-dimensional distribution map of the water film thickness on the steel plate surface based on the current point water film thickness combined with the steel plate position recorded by the encoder and the lateral scanning position of the sensor; comparing the two-dimensional distribution map of the water film thickness on the steel plate surface with a preset threshold, and adjusting the process parameters in real time for areas exceeding the threshold.

Description

[0001] Technical Field This application relates to the field of data processing technology, specifically to a method and apparatus for detecting the thickness of a water film on the surface of stainless steel after pickling.

[0002] Background Technology: Stainless steel pickling is a key process for removing oxide scale and contaminants from its surface. After pickling, it is usually rinsed with water to remove residual acid. After rinsing, a water film will be covered on the stainless steel surface. The thickness and uniformity of the water film directly affect the quality of subsequent passivation treatment. If the water film is too thick or unevenly distributed, it may cause local dilution or uneven retention of the passivation solution, resulting in uneven passivation film thickness and reduced corrosion resistance. If the water film is too thin or even locally dry, it may cause secondary oxidation on the surface, which will also impair the passivation effect. Currently, the quality control of the surface condition of stainless steel after pickling mainly focuses on the offline detection of solid film layers such as the integrity, elemental composition and corrosion resistance of the passivation film after pickling. Auger electron spectroscopy, X-ray photoelectron spectroscopy or atomic force microscopy are used. Although these methods are highly accurate, the equipment is expensive, the operation is complicated and time-consuming. They cannot achieve online, non-contact real-time monitoring, and it is even more difficult to quickly and non-destructively assess the thickness of the water film, a liquid transient layer, during the water washing stage after pickling. There is a lack of effective methods and devices in the current technology for directly and in real time detecting the thickness and uniformity of residual water film on the surface of stainless steel after pickling.

[0003] Summary of the Invention In order to solve the problems existing in the prior art, the purpose of this application is to provide a method and device for detecting the thickness of water film on the surface of stainless steel after pickling.

[0004] The method for detecting the surface water film thickness of stainless steel after pickling as described in this application includes: S101. The pickled and washed stainless steel plate is conveyed to the detection area at a uniform speed, triggering the position sensor to generate a start signal. S102. The control system receives the signal and activates the upper spectral confocal displacement sensor array, making it vertically aligned with the steel plate. S103, the sensor array emits a broadband light beam and receives light signals of different wavelengths reflected from the upper surface of the water film and the metal substrate; S104. The signal processing unit calculates the water film thickness at the current point in real time based on the difference between the two characteristic wavelengths and the refractive index model of water. S105. Based on the current water film thickness and the steel plate position recorded by the encoder and the sensor's lateral scanning position, generate a two-dimensional distribution map of the water film thickness on the steel plate surface. S106. Compare the two-dimensional distribution map of water film thickness on the steel plate surface with the preset threshold, and adjust the process parameters in real time for areas exceeding the threshold.

[0005] Preferably, in step S101, the position information of the stainless steel plate on the conveyor belt is acquired in real time, and the position sensor detects whether it has entered the detection area. After the sensor outputs a start signal, it collects the current speed of the conveyor belt, calculates the estimated time for the steel plate to arrive at the center point of the detection area, and starts the image acquisition device in advance before arrival to obtain an image of the steel plate surface. The image is used to identify and classify defects using a convolutional neural network, and the corresponding defect category results are output.

[0006] Preferably, in step S102, after the control system receives the signal that the steel plate is in place, it drives the upper laser triangulation sensor group to start, and performs multi-point synchronous acquisition of the height of the steel plate surface to obtain multi-point height data. Based on the data, calculate the deviation of each point relative to the reference plane. If the deviation exceeds the preset range, mark the corresponding area as an abnormal height area and obtain its lateral coordinates. The contour boundary of the region is extracted by edge detection to determine the geometric center position. Based on this, the focal length of the linear scan camera is adjusted and the shooting is triggered to obtain a grayscale image of the surface. After performing brightness equalization on the image, a convolutional neural network is used to classify the surface state and output the final classification result.

[0007] Preferably, in step S103, the sensor array emits a broadband light beam to collect the reflected light signals from the upper surface of the water film and the metal substrate and determine the initial data; Different wavelength light signals are separated by wavelength differentiation method to obtain light signal group. The intensity of reflected light on the upper surface of water film is extracted. If it exceeds the threshold, it is marked as an abnormal reflection area. The reflected signal from the metal substrate was then extracted to determine its wavelength distribution characteristics, and the differential data of optical signals between the water film and the substrate were analyzed and obtained. By using convolutional neural networks to classify differential data, determine the state category of the water film surface, generate state description data, and complete the state assessment.

[0008] Preferably, in step S104, the signal processing unit acquires and processes characteristic wavelength data from the sensor array to obtain wavelength comparison results; The difference between the two characteristic wavelengths is calculated and standardized. The refractive index parameters at the point are obtained by matching the refractive index database with the water refractive index model. After model calculation and correction, the initial water film thickness value is determined. If it exceeds the threshold, the original wavelength data is traced back for secondary verification to obtain the corrected thickness. The information processing module records the archive thickness and location information, and then transmits it to the subsequent monitoring system through the data interface to complete real-time analysis and data sharing.

[0009] Preferably, in step S105, the acquisition module obtains the longitudinal position data of the encoder's steel plate and the lateral position data of the sensor's lateral scanning mechanism, and the thickness calculation module obtains the water film thickness value at the corresponding position. Based on the longitudinal and transverse position data, a coordinate relationship is established to obtain the two-dimensional coordinate points of each thickness value, and a preliminary thickness distribution matrix is ​​generated through mapping processing. The blank cells in the matrix are filled using nearest neighbor interpolation to obtain the complete matrix; Then, the color depth is determined according to the thickness value of each unit, and a two-dimensional distribution map of the water film thickness on the steel plate surface is generated according to the color mapping rules.

[0010] Preferably, in step S106, a two-dimensional distribution map of the thickness of the water film on the steel plate is obtained and compared with a preset threshold to identify areas where the thickness exceeds the limit and to extract their coordinates. By combining the steel plate running speed and coolant flow rate data, the time window corresponding to each over-limit area is calculated, the flow rate during that period is extracted and the adjustment range of the nozzle group is determined, and the flow control command is obtained to modify the set value. If there are still areas exceeding the limits, the distribution map is re-acquired, the traffic data is refreshed according to the updated traffic settings, the time window is recalculated, and the control process is repeated.

[0011] A device for detecting the thickness of water film on the surface of stainless steel after pickling includes a main body, a variable frequency speed control motor is arranged on the left side of the main body, a detection area is arranged at the bottom of the variable frequency speed control motor, and a conveyor belt is arranged at the bottom of the detection area. A sensor detector is installed on the right side of the main body, a lifting adjustment bracket is installed at the bottom of the sensor detector, a laser triangulation sensor is installed on the right side of the outside of the lifting adjustment bracket, and a spectral confocal displacement sensor is installed on the left side of the outside of the lifting adjustment bracket. An encoder is installed in the middle of the body, and a linear displacement sensor is installed to the right of the encoder.

[0012] Preferably, the variable frequency speed control motor is fixedly connected to the main body, and the detection area is fastened to the conveyor belt; The main body is fixedly connected to the encoder, and the encoder is embedded and fixedly connected to the linear displacement sensor. The lifting and adjusting bracket is fixedly connected to the sensor detector, laser triangulation sensor, and spectral confocal displacement sensor.

[0013] Preferably, in a device for detecting the thickness of a water film on the surface of stainless steel after pickling, the working principle is as follows: The variable frequency speed control motor is electrically connected to the conveyor belt. The stainless steel plate after pickling and washing is placed directly on the conveyor belt surface. The variable frequency speed control motor drives the conveyor belt to run at a constant linear speed of 0.35m / s as required, which is used to realize the uniform transmission of the steel plate to the inspection area and to provide a basis for fixed-point inspection. The sensor detector is fixed directly above the main body, and the detection optical path of the detection area is horizontally aligned with the steel plate conveying surface of the conveyor belt, and maintains a non-contact distance with the steel plate surface. The optical path height is matched with the steel plate conveying height. When the leading edge of the steel plate blocks the optical path, the sensor detector generates a start signal of 2.5V according to the file threshold setting, providing the trigger condition for activation. The laser triangulation sensor and the spectral confocal displacement sensor are fixed on the lifting and adjusting bracket; The lifting and adjusting bracket can be vertically fine-tuned to ensure vertical alignment with the steel plate surface. The detection ends of the sensors all face the conveyor belt surface, maintaining a non-contact detection distance from the steel plate surface, for multi-point synchronous data acquisition; The encoder, connected to the conveyor belt, collects longitudinal position data of the steel plate in real time, providing a basis for matching the thickness value with the longitudinal coordinate. The linear displacement sensor is fixed in the middle of the body and works in conjunction with the sensor detector, moving synchronously to collect the transverse scanning position data of the sensor detector in real time, which provides a basis for matching the thickness value with the transverse coordinate. The signal output terminals of the spectral confocal displacement sensor array, laser triangulation sensor group, encoder, and linear displacement sensor are all connected via dedicated signal transmission harnesses.

[0014] The method and apparatus for detecting the thickness of the water film on the surface of stainless steel after pickling described in this application have the advantage that, during the uniform conveying process of stainless steel plates after pickling and washing, traditional methods are difficult to achieve online, rapid, and non-contact accurate detection of the thickness distribution of the residual water film, which can easily lead to process problems such as uneven drying, residual liquid corrosion, or coating quality defects. This invention triggers a start signal via a position sensor, activating the upper spectral confocal displacement sensor array in real time, so that it is vertically aligned with the moving steel plate; By simultaneously acquiring characteristic wavelength signals reflected from the upper surface of the water film and the metal substrate using a broadband beam, and combining the refractive index model of water with high-precision wavelength difference calculation, the thickness of the water film at each detection point can be calculated instantaneously. By fusing the longitudinal position of the steel plate recorded by the encoder with the transverse scanning position of the sensor, a complete two-dimensional distribution map of the water film thickness on the steel plate surface is constructed. The distribution map is compared with the preset thickness threshold, and the parameters of the preceding pickling, rinsing or purging process are adjusted in real time for areas with out-of-tolerance conditions, thereby achieving accurate online monitoring and closed-loop control of water film thickness. This invention significantly improves the consistency of stainless steel surface cleanliness, reduces the risk of residual liquid corrosion, and ensures the stability of subsequent processing quality and product qualification rate. Attached Figure Description

[0015] Figure 1This application describes a method for detecting the surface water film thickness of stainless steel after pickling. Figure 1 ; Figure 2 This application describes a method for detecting the surface water film thickness of stainless steel after pickling. Figure 2 ; Figure 3 This application describes a three-dimensional device for detecting the surface water film thickness of stainless steel after pickling. Figure 1 ; Figure 4 This application describes a three-dimensional device for detecting the surface water film thickness of stainless steel after pickling. Figure 2 ; Figure 5 This is a left view of the stainless steel surface water film thickness detection device after pickling as described in this application; 1-Body, 2-Variable frequency speed control motor, 3-Detection area, 4-Conveyor belt, 5-Sensor detector, 6-Lifting and adjusting bracket, 7-Laser triangulation sensor, 8-Spectral confocal displacement sensor, 9-Encoder, 10-Linear displacement sensor. Detailed Implementation

[0016] like Figures 1-2 As shown in this application, a method for detecting the thickness of a water film on the surface of stainless steel after pickling is described.

[0017] like Figures 1-2 As shown, in step S101, the pickled and washed stainless steel plate is uniformly conveyed to the detection area, triggering the position sensor to generate a start signal.

[0018] Further, in step S101, the real-time position information of the stainless steel plate on the conveyor belt is obtained; A position sensor is used to detect whether a stainless steel plate has entered the detection area; If the position sensor outputs a start signal, then the current speed value of the conveyor belt is obtained; Calculate the estimated time for the stainless steel plate to reach the center point of the inspection area based on the current speed of the conveyor belt. Start the image acquisition equipment before the expected time arrives; The surface image of the stainless steel plate when it is located in the detection area is obtained using an image acquisition device; A convolutional neural network is used to classify defects in surface images to obtain defect category results.

[0019] Specifically, in step S101, after the stainless steel plate is pickled and washed in multiple stages, the residual liquid moisture content on the surface is controlled to be below 0.8%. The steel plate is then smoothly fed into the detection area by a conveyor belt driven by a variable frequency speed control motor at a constant linear speed of 0.35 m / s. When the leading edge of the steel plate reaches the detection optical path of the photoelectric through-beam position sensor, the optical signal changes abruptly from high level to low level. The threshold of the comparator inside the sensor is set to 2.5V, generating a falling edge trigger signal. This signal is shaped by a Schmitt trigger and then input into the high-speed counting channel of the PLC. Signal acquisition and de-jitter processing are completed within 10ms. The PLC calls the preset startup subroutine, first reads the current encoder pulse value as the reference position t0, and starts a timer interrupt at a frequency of 1000Hz. In the interrupt service routine, it acquires the steel plate surface image acquired by the line scan camera in real time and stores it in a circular buffer. It uses the Otsu adaptive threshold segmentation algorithm to binarize each frame of the image and calculates the pixel ratio of the defect area. When the percentage of defective pixels in three consecutive frames exceeds 1.2% or the maximum connected area of ​​a single frame is greater than 18 mm², an audible and visual alarm is triggered, and the steel plate number and defect coordinates corresponding to the current time t0 are uploaded to the MES system for traceability via the ModbusTCP protocol. The entire process, from sensor triggering to defect determination, is controlled within 180ms, ensuring no missed detections or duplicate markings under uniform transmission conditions.

[0020] Specifically, in step S101, the estimated time for the steel plate to reach the center point of the detection area is calculated using the following formula: Where t is the estimated time, L is the distance from the leading edge of the steel plate to the center point of the detection area, and v is the constant linear velocity of the conveyor belt, v=0.35m / s; The specific calculation of the percentage of pixels in the defective area is as follows: Where P is the percentage of pixels in the defect area, and S... 缺 S represents the number of pixels in the defective region of the image. 总 This represents the total number of pixels in a single frame of the image.

[0021] like Figures 1-2 As shown, S102, the control system receives the signal and activates the upper spectral confocal displacement sensor array, making it vertically aligned with the steel plate.

[0022] Furthermore, in step S102, the control system receives the steel plate positioning signal and drives the upper laser triangulation sensor group to start, and performs multi-point synchronous acquisition of the steel plate surface height; Multiple height data of the steel plate surface are obtained by using a laser triangulation sensor set; Calculate the deviation of each point relative to the reference plane based on the height data of multiple points on the steel plate surface; If the deviation value at a certain point exceeds the preset range, the area to which that point belongs is marked as an abnormal height area; Obtain the horizontal coordinate value of the corresponding location for areas with abnormal height; Edge detection methods are used to extract the contour boundaries of regions with abnormal height. Determine the geometric center of the abnormal height region by using the contour boundary; Adjust the focal length of the line scan camera based on the geometric center position and trigger the shooting; Acquire grayscale images of the steel plate surface captured by a linear scan camera; A balanced image is obtained by performing brightness equalization processing on the grayscale image of the steel plate surface; A convolutional neural network is used to classify the surface state of a balanced image, and the classification results are obtained.

[0023] Specifically, in step S102, after the control system receives the signal, it immediately activates the upper spectral confocal displacement sensor array and realizes vertical alignment and precise measurement of the steel plate through a preset logic program. The control system sends activation commands via an internal clock at 50-microsecond intervals to adjust the operating frequency of the sensor array to 2000Hz, ensuring that it can respond in real time to minute height changes on the steel plate surface. The vertical alignment angle of the sensor array is calibrated by a built-in gyroscope, with the error controlled within 0.1 degrees. The sensor array collects steel plate surface height data at a sampling rate of 5000 times per second. The data is transmitted to the embedded processor for preprocessing. A sliding window averaging filter algorithm is used to smooth the collected data. The window size is set to 10 sampling points to reduce noise interference. The standard deviation of the filtered data must be less than 0.05mm. Otherwise, the system will automatically adjust the sampling rate to 6000 times / second and re-collect the data. The processor performs deviation analysis based on a preset reference height value of 3.2mm. If a height deviation of more than 0.15mm is detected in a certain area, it is marked as a potential anomaly point, and the three-dimensional coordinate data of the point is stored in a temporary cache in a format of 100 bytes per group. The analysis module based on the K-means clustering algorithm is called to classify outliers. The number of cluster centers is set to 3, and the number of iterations is limited to 15 to distinguish different types of surface unevenness. The classification results are transmitted to the main control unit through the internal bus. If the percentage of outliers exceeds 0.5% within 5 consecutive sampling periods, the system will automatically generate an anomaly report, which includes key parameters such as the distribution density of outliers and the maximum deviation value of 0.18 mm, and upload the report to the data center for analysis via Ethernet protocol. The system is used to assess the uniformity of steel plate thickness. The height data is compared with the standard thickness value of 2.8 mm to calculate the uniformity index. If the index is lower than 0.95, a calibration procedure is triggered to automatically adjust the scanning range of the sensor array to ±5 mm to ensure detection accuracy. The entire process is completed automatically by the system and takes less than 120 milliseconds.

[0024] Specifically, in step S102, the deviation value of the surface point relative to the reference plane is calculated as follows: Where Δh is the height deviation value, h 测 h represents the actual height of the surface point acquired by the laser triangulation sensor. 基 The preset reference plane height; The steel plate thickness uniformity index is calculated as follows: Where K is the uniformity index, h represents the average thickness of the steel plate inspection area. 标 This refers to the standard value for steel plate thickness. The noise assessment in the standard deviation of the sampled data is calculated as follows: Where σ is the standard deviation of the sampling height data, n is the number of sampling points within the sliding window, and h i Let be the height value of the i-th sampling point. This represents the average height of the sampling points within the window.

[0025] like Figures 1-2 As shown, S103 and the sensor array emit a broadband light beam and receive light signals of different wavelengths reflected from the upper surface of the water film and the metal substrate.

[0026] Further, in step S103, a broadband light beam is emitted through a sensor array to acquire the reflected light signals from the upper surface of the water film and the metal substrate, thereby determining the initial light signal data; Based on the initial optical signal data, the reflected optical signals of different wavelengths are separated using a wavelength differentiation method to obtain the separated optical signal groups; For the separated light signal group, the intensity of the reflected light signal corresponding to the upper surface of the water film is obtained, and it is determined whether the signal intensity exceeds the preset threshold range. If it does, it is marked as an abnormal reflection area. By extracting the reflected light signal corresponding to the metal substrate from the data of the abnormal reflection area, the wavelength distribution characteristics of the substrate reflection are determined. Based on the wavelength distribution characteristics of the substrate reflection, the difference in optical signal between the upper surface of the water film and the metal substrate is analyzed to obtain differential data; A convolutional neural network is used to classify the differential data and determine the state category of the water film surface. Based on the state category, obtain the corresponding water film state description data to complete the state assessment process.

[0027] Specifically, in step S103, after the sensor array is powered on, it emits a broadband beam of light containing wavelengths from 400 nm to 1000 nm downwards. The beam is focused by a collimating lens group into a point light source with a diameter of about 20 micrometers and irradiates the water film area on the steel plate surface. The spectrometer inside the spectral confocal system receives in real time short-wavelength light signals reflected from the upper surface of the water film and long-wavelength light signals reflected from the surface of the metal substrate after penetrating the water film. The two reflected lights enter the dispersive element through different optical paths for wavelength separation. The separated spectral data were acquired using a 2048-pixel linear CCD array, with a fixed integration time of 80 microseconds per frame, corresponding to a single measurement cycle of approximately 125 microseconds. The acquired raw spectral signals are first processed by a Gaussian fitting algorithm to locate the peak wavelength position. The peak positioning accuracy is controlled within 0.08 nm. The peak wavelength reflected from the upper surface of the water film is denoted as λ1, and the peak wavelength reflected from the metal substrate is denoted as λ2. The system has a built-in lookup table that directly maps the water film thickness value based on the wavelength difference Δλ=λ2-λ1. The lookup table has a resolution of 0.001mm and corresponds to a thickness measurement range of 0.1mm to 2.5mm. The processor applies an exponentially weighted moving average algorithm to smooth the continuously acquired thickness sequence, with the weighting coefficient α set to 0.12 to balance response speed and stability. If the rate of change of two adjacent thickness readings exceeds 18% / s after smoothing, it is determined to be transient interference, and the system will automatically discard the data set and re-initiate the measurement in the next cycle; The thickness data is compared with the preset standard water film thickness value of 0.8mm in real time to calculate the local thickness deviation. If the absolute value of the deviation at a certain point is greater than 0.09mm, its coordinates and thickness value are combined into a key-value pair and stored in a circular buffer. The buffer capacity is set to 800 sets. Every 20 measurement cycles, the system calls the DBSCAN density clustering algorithm to perform spatial clustering analysis on the deviation points in the buffer. The neighborhood radius ε is 0.6 mm, and the minimum number of samples MinPts is set to 4 to identify continuous water film accumulation or local missing areas. The clustering labels are sent to the host controller via the SPI bus. If the number of points in any cluster exceeds 35 or the maximum thickness deviation within the cluster exceeds 0.22 mm, an abnormal water film thickness event is triggered. The cluster center location, number of points, and average deviation value are packaged into a data frame and pushed to the process monitoring server via TCP / IP protocol. The entire automatic detection and judgment link is designed to take no more than 90 milliseconds.

[0028] Specifically, in step S103, the difference in reflection wavelength between the upper and lower surfaces of the water film is calculated as follows: Where Δλ is the wavelength difference, λ2 is the peak wavelength of reflection from the metal substrate, and λ1 is the peak wavelength of reflection from the upper surface of the water film. The rate of change of thickness reading is calculated as follows: Where r is the rate of change of thickness reading, d2 is the water film thickness value of the second measurement, d1 is the water film thickness value of the first measurement, and Δt is the time interval between the two measurements. Local thickness deviation, specifically calculated as follows: Where Δd is the local thickness deviation, d 测 The measured value of the water film thickness at the test point, d 标 This is the preset standard water film thickness value.

[0029] like Figures 1-2 As shown, S104 and the signal processing unit calculate the water film thickness at the current point in real time based on the difference between the two characteristic wavelengths and the refractive index model of water.

[0030] Further, in step S104, the signal processing unit acquires characteristic wavelength data from the sensor array, performs preliminary processing on the acquired wavelength information, and obtains wavelength comparison results. Based on the wavelength comparison results, the difference data between the two characteristic wavelengths is calculated, and the difference data is standardized using a preset calculation rule to determine the standardized difference information. Based on the standardized difference information, and combined with the pre-established refractive index value database, the corresponding water refractive index model is called for matching to obtain the refractive index parameters of the current point. By using the refractive index parameters at the current location and combining the difference information, the model is calculated, and the calculation results are corrected using real-time analysis methods to determine the preliminary water film thickness value. If the initial water film thickness value exceeds the preset threshold range, the signal processing flow is backtracked to check, the original characteristic wavelength data is obtained for secondary verification, and the corrected thickness data is obtained. Based on the corrected thickness data and the location information of the current point, the thickness calculation results are recorded and archived through the information processing module to complete the thickness analysis process; For the archived thickness analysis data, a data acquisition interface is used to transmit it to the subsequent monitoring system to complete real-time analysis and data sharing.

[0031] Specifically, in step S104, after the spectral acquisition module is started, it uses a multi-channel fiber bundle to project broadband light with different incident angles onto the surface of the water film on the steel plate to be tested. The incident light wavelength is limited to between 450nm and 950nm. The light intensity of each channel is equalized by a light homogenizing plate to form a measurement spot array with a diameter of about 15 micrometers. The reflected light signals are captured by dual-channel receiving lenses integrated in the probe. The paraxial reflected light mainly carries information about the upper surface of the water film, while the off-axis reflected light contains more of the metal substrate reflection component. The two beams are coupled into independent 0.22NA optical fibers and transmitted to the spectrometer's beam splitter. The grating line density is 1200 lines / mm, the blaze wavelength is 750nm, and the spectrum after spectral dispersion is projected onto a 1024×1 back-illuminated sCMOS detector. The single exposure time is set to 60 microseconds, the actual frame rate is about 200Hz, and the original spectral curve is first smoothed by seven points using Savitzky-Golay filtering. The center wavelengths of two significant reflection peaks were extracted by fitting the Lorentzian peak, and the fitting residual was controlled within 0.05 nm. The wavelength of the upper surface reflection peak was labeled as λs, and the wavelength of the substrate reflection peak was labeled as λb. The thickness calculation uses a pre-calibrated dispersion model formula, where n(λ) is the average refractive index of water in this band, which is taken as 1.333, and Δn is the empirical correction value of 0.0042 for the effective refractive index difference of the reflected light from the upper and lower surfaces. The calculation result has an accuracy of 0.0008mm and a range covering 0.05mm to 3.0mm. The continuous thickness data stream is fed into a Kalman filter for noise reduction. The process noise covariance is set to 0.0004 mm², and the measurement noise covariance is set to 0.0012 mm². The filtered output value is compared with the target thickness of 0.75 mm to obtain the deviation sequence. Points with deviations exceeding ±0.085mm are marked and stored along with their spatial coordinates (X,Y) in a FIFO buffer with a capacity of 1000 groups; K-means clustering is triggered every 25 valid bias points accumulated in the buffer. The preset number of clusters is 3, the iterative convergence threshold is 0.001mm, and the optimal cluster division is evaluated by the silhouette coefficient. After clustering is completed, the geometric center coordinates, number of points within the cluster, and root mean square deviation of thickness are calculated for each cluster. If the number of points in any cluster is greater than or equal to 28 or the root mean square deviation is greater than 0.20 mm, an abnormal event report is generated, which includes the cluster number, center position, abnormality type, and statistical characteristic value. The report is uploaded to the production line PLC in real time via Modbus RTU protocol. The entire signal processing to event reporting process is controlled within 75 milliseconds.

[0032] Specifically, in step S104, the core formula of the water film thickness dispersion model is: Where d is the water film thickness, λb is the center wavelength of the reflection peak of the metal substrate, λs is the center wavelength of the reflection peak of the upper surface of the water film, n(λ) is the average refractive index of water in the detection band, and Δn is the empirical correction value for the effective refractive index difference of the reflected light from the upper and lower surfaces. Noise reduction calculations in Kalman filter processes or measurement noise covariance correlation: Process noise covariance: Q = 0.0004 mm 2 ; Measurement noise covariance: R = 0.0012 mm 2 ; Comparison of thickness deviation with target thickness: Where, Δd k The thickness deviation after Kalman filtering, d 滤 d represents the water film thickness value output after filtering. 目 This represents the target thickness value of the water film. Root mean square of thickness deviation: Where RMS is the root mean square of the thickness deviation within the cluster, m is the number of deviation points within the cluster, and Δd j This represents the thickness deviation value at the j-th point within the cluster.

[0033] like Figures 1-2 As shown, in step S105, based on the current water film thickness and the steel plate position recorded by the encoder and the sensor's lateral scanning position, a two-dimensional distribution map of the water film thickness on the steel plate surface is generated.

[0034] Further, in step S105, the longitudinal position data of the steel plate output by the encoder is acquired through the acquisition module; The lateral position data output by the sensor's lateral scanning mechanism is acquired through the acquisition module; The thickness of the water film at the current location is obtained through the thickness calculation module; Establish a positional correspondence based on the longitudinal and transverse positional data of the steel plate to obtain the two-dimensional coordinate point of each thickness value; By mapping two-dimensional coordinate points to water film thickness values, the distribution position of thickness values ​​in the two-dimensional grid is determined, and a preliminary thickness distribution matrix is ​​obtained. If there are blank grid cells in the initial thickness distribution matrix, the nearest neighbor interpolation method is used to obtain the thickness value from the adjacent grid cells to fill the blank positions, thus obtaining the complete thickness distribution matrix; The color depth of the corresponding color block is determined based on the thickness value of each grid cell in the complete thickness distribution matrix, and a two-dimensional distribution map of the water film thickness on the steel plate surface is generated through color mapping rules.

[0035] Specifically, in step S105, after the system receives the pulse signal from the steel plate running encoder and the position feedback from the sensor's lateral stepper motor, it accurately matches the real-time thickness value with the corresponding spatial position. Whenever the encoder moves forward 1mm, it triggers a longitudinal coordinate update. At the same time, the lateral scanning mechanism completes one line of data acquisition with a step size of 0.5mm, forming a grid of regularly spaced dots. All valid measurement points are organized according to the X-axis corresponding to the steel plate movement direction and the Y-axis corresponding to the transverse scanning direction, and the thickness data is directly mapped to the corresponding grid cell. For the collected two-dimensional scatter thickness matrix, the original non-uniform sampling points are first expanded to a uniform resolution grid of 0.2mm×0.2mm using a bilinear interpolation algorithm. The interpolation window is calculated by weighting the nearest 12 effective neighboring points. Gaussian smoothing was then performed on the entire thickness field, with a smoothing kernel size of 5×5 and a pixel standard deviation of 1.2, in order to suppress local random fluctuations while preserving the true thickness gradient features. The deviation field matrix is ​​obtained by subtracting the target thickness of 0.75 mm from the smoothed thickness matrix. The deviation field is then processed by median filtering to eliminate isolated outliers. The median filtering window size is 3×3. Next, the marching squares algorithm is used to extract the deviation contour lines. The main contour line levels are set to ±0.05mm, ±0.10mm, ±0.15mm, and ±0.20mm. Potential abnormal areas are automatically identified based on the degree of contour line closure and the enclosed area. The number of points inside each closed contour and the average deviation value are counted. If the number of points inside the contour exceeds 450 and the average absolute deviation is greater than 0.12mm, the area is marked as a thickness uneven area, and its minimum bounding rectangle center coordinates, aspect ratio and maximum deviation gradient are calculated. The thickness 2D distribution map uses pseudo-color rendering, with the color band mapping range set from 0.30mm to 1.20mm. It uses a jet color map that transitions from blue to red, with dark blue displayed in low-thickness areas and dark red displayed in high-thickness areas. Areas exceeding the measurement range are represented by saturated magenta and cyan, respectively. The final visualization image resolution is fixed at 1600×1200 pixels, and is cached along with key statistical indicators for periodic refresh or archiving.

[0036] Specifically, in step S105, the deviation field matrix is ​​calculated: Where, d 偏 d represents the thickness deviation at coordinates (x, y) in a two-dimensional mesh. 网(x,y) represents the water film thickness at grid cell (x,y), d 目 This represents the target thickness value of the water film.

[0037] like Figures 1-2 As shown in step S106, the two-dimensional distribution map of the water film thickness on the steel plate surface is compared with the preset threshold, and the process parameters in the area exceeding the threshold are adjusted in real time.

[0038] Further, in step S106, a two-dimensional distribution map of the water film thickness on the steel plate surface and a preset thickness threshold are obtained; Local areas with excessive thickness were identified through comparison; Extract the coordinates of each out-of-limit location from the local area; Obtain the current running speed data of the steel plate and the coolant injection flow rate data; Calculate the time window corresponding to each over-limit area based on the location coordinates and running speed data; Extract coolant injection flow rate data within the time window; Determine the adjustment range for the corresponding nozzle group based on the coolant injection flow rate data; The flow control command for each nozzle group is generated by adjusting the amplitude. Modify the flow rate setting value of the corresponding nozzle group using the flow control command; If areas with excessive thickness still exist, repeat the acquisition of the two-dimensional distribution map of water film thickness on the steel plate surface; Update the coolant injection flow rate data according to the modified flow rate setting; The time window is recalculated based on the updated coolant injection flow rate data.

[0039] Specifically, in step S106, during the process of comparing the two-dimensional distribution map of the water film thickness on the steel plate surface with the preset threshold and adjusting the process parameters in real time, the system first compares the thickness value of each grid cell with the preset threshold range of 0.70mm to 0.80mm point by point through the constructed thickness distribution matrix, automatically identifies abnormal areas with thickness values ​​lower than 0.70mm or higher than 0.80mm, and records the coordinate range of these areas. For example, a certain area spans 50.5mm to 62.3mm in the X-axis direction and 30.2mm to 35.8mm in the Y-axis direction. Using a region segmentation algorithm, the boundaries of the abnormal regions are fitted to a continuous curve using the least squares method. The coordinates of the geometric center point of each abnormal region are calculated, such as the center point being (56.4mm, 33.0mm). The average thickness deviation within the region is also calculated, for example, the average deviation is +0.09mm. Based on the direction and magnitude of the deviation in the abnormal area, and combined with the pre-established process parameter mapping model, the process parameter values ​​that need to be adjusted are automatically calculated. For example, when the thickness is too high, the water spray pressure parameter is reduced from the default 3.5MPa to 3.2MPa. The adjustment range is based on the linear regression relationship between the deviation value and the pressure, and the regression coefficient is set to 0.03MPa / mm. At the same time, the adjusted parameter values ​​are transmitted to the process control module through the internal communication protocol, and the parameter change logs before and after the adjustment are recorded for subsequent traceability. Next, the changes in the adjusted thickness distribution are monitored in real time. Using a moving average algorithm with a time window of 5 minutes, the thickness values ​​are analyzed to see if they return to the threshold range. For example, if the average thickness of a certain area drops from 0.89 mm to 0.78 mm within 5 minutes after the adjustment, the adjustment is confirmed to be effective. Finally, the processing results of abnormal areas and parameter adjustment records are integrated into structured data, automatically archived into the database, and a brief report is generated for subsequent process optimization reference, ensuring that the entire feedback process forms a closed-loop control.

[0040] Specifically, in step S106, the water spray pressure adjustment amount is calculated: Where ΔP is the water spray pressure adjustment amount, and k is the linear regression coefficient between the deviation value and the pressure. The average thickness deviation of the region exceeding the limit; The adjusted water spray pressure is calculated as follows: Among them, P 调 P is the adjusted water spray pressure. 原 This is the original setting value for the water spray pressure. If the thickness is too high, use "-"; if the thickness is too low, use "+". Calculation of time window for the out-of-limit region: Where T is the time window corresponding to the over-limit area, X2 and X1 are the start and end coordinates of the over-limit area on the X-axis in the direction of steel plate movement, and v is the constant linear velocity of the conveyor belt. Thickness sliding average calculation: in, The thickness is the moving average within the time window, where N is the number of thickness measurement points within the time window, and d is the moving average thickness. k The thickness value is the value measured in the kth measurement.

[0041] like Figures 3-5As shown, a device for detecting the thickness of water film on the surface of stainless steel after pickling includes a main body (1), a variable frequency speed control motor (2) is provided on the left side of the main body (1), a detection area (3) is provided at the bottom of the variable frequency speed control motor (2), and a conveyor belt (4) is provided at the bottom of the detection area (3). A sensor detector (5) is provided on the right side of the main body (1), a lifting adjustment bracket (6) is provided at the bottom of the sensor detector (5), a laser triangulation sensor (7) is provided on the right side of the lifting adjustment bracket (6), and a spectral confocal displacement sensor (8) is provided on the left side of the lifting adjustment bracket (6). An encoder (9) is installed in the middle of the body (1), and a linear displacement sensor (10) is installed on the right side of the encoder (9).

[0042] Furthermore, the variable frequency speed control motor (2) is fixedly connected to the main body (1), and the detection area (3) is snapped together with the conveyor belt (4); The main body (1) is fixedly connected to the encoder (9), and the encoder (9) is embedded and fixedly connected to the linear displacement sensor (10); The lifting and adjusting bracket (6) is fixedly connected to the sensor detector (5), the laser triangulation sensor (7), and the spectral confocal displacement sensor (8) through a through-hole connection.

[0043] A device for detecting the thickness of water film on the surface of stainless steel after pickling, the working principle of which is as follows: The variable frequency speed control motor is electrically connected to the conveyor belt. The stainless steel plate after pickling and washing is placed directly on the conveyor belt surface. The variable frequency speed control motor drives the conveyor belt to run at a constant linear speed of 0.35m / s as required, which is used to realize the uniform transmission of the steel plate to the inspection area and to provide a basis for fixed-point inspection. The sensor detector is fixed directly above the main body, and the detection optical path of the detection area is horizontally aligned with the steel plate conveying surface of the conveyor belt, and maintains a non-contact distance with the steel plate surface. The optical path height is matched with the steel plate conveying height. When the leading edge of the steel plate blocks the optical path, the sensor detector generates a start signal of 2.5V according to the file threshold setting, providing the trigger condition for activation. The laser triangulation sensor and the spectral confocal displacement sensor are fixed on the lifting and adjusting bracket; The lifting and adjusting bracket can be vertically fine-tuned to ensure vertical alignment with the steel plate surface. The detection ends of the sensors all face the conveyor belt surface, maintaining a non-contact detection distance from the steel plate surface, for multi-point synchronous data acquisition; The encoder, connected to the conveyor belt, collects longitudinal position data of the steel plate in real time, providing a basis for matching the thickness value with the longitudinal coordinate. The linear displacement sensor is fixed in the middle of the body and works in conjunction with the sensor detector, moving synchronously to collect the transverse scanning position data of the sensor detector in real time, which provides a basis for matching the thickness value with the transverse coordinate. The signal output terminals of the spectral confocal displacement sensor array, laser triangulation sensor group, encoder, and linear displacement sensor are all connected via dedicated signal transmission harnesses.

[0044] For those skilled in the art, various other corresponding changes and modifications can be made based on the technical solutions and concepts described above, and all such changes and modifications should fall within the protection scope of the claims of this application.

Claims

1. A method for detecting the thickness of a water film on the surface of stainless steel after pickling, characterized in that, include: S101. The pickled and washed stainless steel plate is conveyed to the detection area at a uniform speed, triggering the position sensor to generate a start signal. S102. The control system receives the signal and activates the upper spectral confocal displacement sensor array, making it vertically aligned with the steel plate. S103, the sensor array emits a broadband light beam and receives light signals of different wavelengths reflected from the upper surface of the water film and the metal substrate; S104. The signal processing unit calculates the water film thickness at the current point in real time based on the difference between the two characteristic wavelengths and the refractive index model of water. S105. Based on the current water film thickness and the steel plate position recorded by the encoder and the sensor's lateral scanning position, generate a two-dimensional distribution map of the water film thickness on the steel plate surface. S106. Compare the two-dimensional distribution map of water film thickness on the steel plate surface with the preset threshold, and adjust the process parameters in real time for areas exceeding the threshold.

2. The method for detecting the thickness of a water film on the surface of stainless steel after pickling, as described in claim 1, is characterized in that, In step S101, the position information of the stainless steel plate on the conveyor belt is acquired in real time, and the position sensor detects whether it has entered the detection area. After the position sensor outputs a start signal, it collects the current speed of the conveyor belt, calculates the estimated time for the steel plate to arrive at the center point of the detection area based on the current speed of the conveyor belt, and starts the image acquisition device in advance before arrival to obtain an image of the steel plate surface. A convolutional neural network is used to identify and classify defects in steel plate surface images, and the corresponding defect category results are output.

3. The method for detecting the thickness of a water film on the surface of stainless steel after pickling, as described in claim 1, is characterized in that... In S102, after the control system receives the signal that the steel plate is in place, it drives the laser triangulation sensor group to start, and performs multi-point synchronous acquisition of the height of the steel plate surface to obtain multi-point height data. Based on multi-point height data, the deviation value of each point relative to the reference plane is calculated. If it exceeds the preset range, the corresponding area is marked as an abnormal height area, and its lateral coordinates are obtained. The contour boundary of the abnormal height region is extracted by edge detection to determine the geometric center position. The focal length of the linear scan camera is adjusted according to the geometric center position and the shooting is triggered to obtain the surface grayscale image. After performing brightness equalization processing on the grayscale image of the surface, a convolutional neural network is used to classify the surface state and output the final classification result.

4. The method for detecting the thickness of a water film on the surface of stainless steel after pickling, as described in claim 1, is characterized in that... In S103, the sensor array emits a broadband light beam to collect the reflected light signals from the upper surface of the water film and the metal substrate and determine the initial data. Different wavelength light signals are separated by wavelength differentiation method to obtain light signal group. The intensity of reflected light on the upper surface of water film is extracted. If it exceeds the threshold, it is marked as an abnormal reflection area. Then, the reflection signal from the metal substrate is extracted to determine its wavelength distribution characteristics. The differential optical signals between the water film and the substrate were analyzed and obtained; By using a convolutional neural network to classify differential data, determine the state category of the water film surface, obtain state description data, and complete the state assessment.

5. The method for detecting the thickness of a water film on the surface of stainless steel after pickling, as described in claim 1, is characterized in that... In step S104, the signal processing unit acquires and organizes characteristic wavelength data from the sensor array to obtain wavelength comparison results. The difference between the two characteristic wavelengths is calculated and standardized. The refractive index parameters at the point are obtained by matching the refractive index database with the water refractive index model. The initial water film thickness value is determined through calculation and correction. If it exceeds the threshold, the original wavelength data is traced back for secondary verification to obtain the corrected thickness. The information processing module records the archive thickness and location information, and then transmits it to the subsequent monitoring system through the data interface for real-time analysis and data sharing.

6. The method for detecting the thickness of a water film on the surface of stainless steel after pickling, as described in claim 1, is characterized in that... In step S105, the acquisition module obtains the longitudinal position data of the steel plate of the encoder and the lateral position data of the transverse scan of the sensor, and the thickness calculation module obtains the water film thickness value at the corresponding position. Based on the longitudinal and transverse position data, a coordinate relationship is established to obtain the two-dimensional coordinate points of each thickness value. After mapping processing, a preliminary thickness distribution matrix is ​​obtained. The blank cells in the matrix are filled using nearest neighbor interpolation to obtain the complete matrix; Then, the color depth is determined according to the thickness value of each unit, and a two-dimensional distribution map of the water film thickness on the steel plate surface is generated according to the color mapping rules.

7. The method for detecting the thickness of a water film on the surface of stainless steel after pickling according to claim 1, characterized in that, In step S106, a two-dimensional distribution map of the thickness of the water film on the steel plate is obtained and compared with a preset thickness threshold. The region with excessive thickness is identified and its coordinates are extracted. By combining the steel plate running speed and coolant flow rate data, the time window corresponding to each over-limit area is calculated, the flow rate during that period is extracted and the adjustment range of the nozzle group is determined, and the flow control command is obtained to modify the set value. If there are still areas exceeding the limits, the distribution map is re-acquired, the traffic data is refreshed according to the updated traffic settings, the time window is recalculated, and the control process is repeated.

8. An apparatus for performing the method for detecting the thickness of a water film on the surface of stainless steel after pickling according to any one of claims 1-7, characterized in that, Includes a main body (1), a variable frequency speed control motor (2) is provided on the left side of the main body (1), a detection area (3) is provided at the bottom of the variable frequency speed control motor (2), and a conveyor belt (4) is provided at the bottom of the detection area (3); A sensor detector (5) is provided on the right side of the main body (1), a lifting adjustment bracket (6) is provided at the bottom of the sensor detector (5), a laser triangulation sensor (7) is provided on the right side of the lifting adjustment bracket (6), and a spectral confocal displacement sensor (8) is provided on the left side of the lifting adjustment bracket (6). An encoder (9) is installed in the middle of the body (1), and a linear displacement sensor (10) is installed on the right side of the encoder (9).

Images